2023-04-27 12:19:53,946 INFO [train.py:976] (0/8) Training started 2023-04-27 12:19:53,948 INFO [train.py:986] (0/8) Device: cuda:0 2023-04-27 12:19:53,950 INFO [train.py:995] (0/8) {'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 12:19:53,951 INFO [train.py:997] (0/8) About to create model 2023-04-27 12:19:54,602 INFO [zipformer.py:178] (0/8) 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 12:19:54,626 INFO [train.py:1001] (0/8) Number of model parameters: 70369391 2023-04-27 12:19:57,412 INFO [train.py:1016] (0/8) Using DDP 2023-04-27 12:19:58,277 INFO [multidataset.py:46] (0/8) About to get multidataset train cuts 2023-04-27 12:19:58,277 INFO [multidataset.py:49] (0/8) Loading LibriSpeech in lazy mode 2023-04-27 12:19:58,304 INFO [multidataset.py:65] (0/8) Loading GigaSpeech 1998 splits in lazy mode 2023-04-27 12:20:00,752 INFO [multidataset.py:72] (0/8) Loading CommonVoice in lazy mode 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:230] (0/8) Enable MUSAN 2023-04-27 12:20:00,755 INFO [asr_datamodule.py:231] (0/8) About to get Musan cuts 2023-04-27 12:20:03,004 INFO [asr_datamodule.py:255] (0/8) Enable SpecAugment 2023-04-27 12:20:03,005 INFO [asr_datamodule.py:256] (0/8) Time warp factor: 80 2023-04-27 12:20:03,005 INFO [asr_datamodule.py:266] (0/8) Num frame mask: 10 2023-04-27 12:20:03,005 INFO [asr_datamodule.py:279] (0/8) About to create train dataset 2023-04-27 12:20:03,005 INFO [asr_datamodule.py:306] (0/8) Using DynamicBucketingSampler. 2023-04-27 12:20:07,640 INFO [asr_datamodule.py:321] (0/8) About to create train dataloader 2023-04-27 12:20:07,642 INFO [asr_datamodule.py:435] (0/8) About to get dev-clean cuts 2023-04-27 12:20:07,644 INFO [asr_datamodule.py:442] (0/8) About to get dev-other cuts 2023-04-27 12:20:07,645 INFO [asr_datamodule.py:352] (0/8) About to create dev dataset 2023-04-27 12:20:07,885 INFO [asr_datamodule.py:369] (0/8) About to create dev dataloader 2023-04-27 12:20:25,635 INFO [train.py:904] (0/8) Epoch 1, batch 0, loss[loss=7.31, simple_loss=6.616, pruned_loss=6.922, over 17039.00 frames. ], tot_loss[loss=7.31, simple_loss=6.616, pruned_loss=6.922, over 17039.00 frames. ], batch size: 41, lr: 2.50e-02, grad_scale: 2.0 2023-04-27 12:20:25,636 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 12:20:32,881 INFO [train.py:938] (0/8) Epoch 1, validation: loss=6.911, simple_loss=6.238, pruned_loss=6.721, over 944034.00 frames. 2023-04-27 12:20:32,882 INFO [train.py:939] (0/8) Maximum memory allocated so far is 12865MB 2023-04-27 12:20:36,375 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:20:51,540 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:21:17,139 INFO [train.py:904] (0/8) Epoch 1, batch 50, loss[loss=1.353, simple_loss=1.192, pruned_loss=1.436, over 17021.00 frames. ], tot_loss[loss=2.171, simple_loss=1.966, pruned_loss=1.961, over 740198.30 frames. ], batch size: 55, lr: 2.75e-02, grad_scale: 2.0 2023-04-27 12:21:46,544 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:21:54,086 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=8.75 vs. limit=2.0 2023-04-27 12:22:02,793 WARNING [train.py:894] (0/8) Grad scale is small: 0.001953125 2023-04-27 12:22:02,793 INFO [train.py:904] (0/8) Epoch 1, batch 100, loss[loss=1.238, simple_loss=1.05, pruned_loss=1.47, over 16698.00 frames. ], tot_loss[loss=1.639, simple_loss=1.459, pruned_loss=1.619, over 1307739.12 frames. ], batch size: 57, lr: 3.00e-02, grad_scale: 0.00390625 2023-04-27 12:22:13,674 INFO [optim.py:368] (0/8) 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 12:22:20,966 WARNING [optim.py:388] (0/8) Scaling gradients by 0.0112030990421772, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:21,083 INFO [optim.py:450] (0/8) 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 12:22:27,952 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=18.23 vs. limit=2.0 2023-04-27 12:22:43,780 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:22:46,129 WARNING [optim.py:388] (0/8) Scaling gradients by 0.0022801109589636326, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:46,270 INFO [optim.py:450] (0/8) 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 12:22:49,650 WARNING [optim.py:388] (0/8) Scaling gradients by 0.04246773198246956, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:49,758 INFO [optim.py:450] (0/8) 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 12:22:51,355 WARNING [optim.py:388] (0/8) Scaling gradients by 0.000716241542249918, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:51,490 INFO [optim.py:450] (0/8) 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 12:22:53,004 INFO [train.py:904] (0/8) Epoch 1, batch 150, loss[loss=0.9579, simple_loss=0.8103, pruned_loss=1.061, over 16828.00 frames. ], tot_loss[loss=1.398, simple_loss=1.226, pruned_loss=1.448, over 1754857.65 frames. ], batch size: 90, lr: 3.25e-02, grad_scale: 0.00390625 2023-04-27 12:22:53,729 WARNING [optim.py:388] (0/8) Scaling gradients by 0.049951765686273575, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:53,833 INFO [optim.py:450] (0/8) Parameter Dominanting tot_sumsq module.encoder.encoder_embed.conv.0.weight with proportion 0.92, where dominant_sumsq=(grad_sumsq*orig_rms_sq)=3.845e+08, grad_sumsq = 8.968e+09, orig_rms_sq=4.287e-02 2023-04-27 12:22:58,545 WARNING [optim.py:388] (0/8) Scaling gradients by 0.00609818659722805, model_norm_threshold=1019.0284423828125 2023-04-27 12:22:58,675 INFO [optim.py:450] (0/8) 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 12:23:16,874 WARNING [optim.py:388] (0/8) Scaling gradients by 0.059935860335826874, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:17,015 INFO [optim.py:450] (0/8) 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] (0/8) Scaling gradients by 0.060559310019016266, model_norm_threshold=1019.0284423828125 2023-04-27 12:23:28,448 INFO [optim.py:450] (0/8) 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 12:23:31,460 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=20.10 vs. limit=2.0 2023-04-27 12:23:40,756 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=8.77 vs. limit=2.0 2023-04-27 12:23:42,817 WARNING [train.py:894] (0/8) Grad scale is small: 0.00390625 2023-04-27 12:23:42,818 INFO [train.py:904] (0/8) Epoch 1, batch 200, loss[loss=0.9703, simple_loss=0.8218, pruned_loss=0.9966, over 16165.00 frames. ], tot_loss[loss=1.263, simple_loss=1.095, pruned_loss=1.325, over 2108520.78 frames. ], batch size: 165, lr: 3.50e-02, grad_scale: 0.0078125 2023-04-27 12:23:51,004 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 2.076e+02 2.707e+02 4.762e+02 1.423e+06, threshold=5.415e+02, percent-clipped=11.0 2023-04-27 12:23:51,004 WARNING [optim.py:388] (0/8) Scaling gradients by 0.002041660714894533, model_norm_threshold=541.4743041992188 2023-04-27 12:23:51,110 INFO [optim.py:450] (0/8) 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 12:24:00,576 WARNING [optim.py:388] (0/8) Scaling gradients by 0.02974529005587101, model_norm_threshold=541.4743041992188 2023-04-27 12:24:00,687 INFO [optim.py:450] (0/8) 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 12:24:01,480 WARNING [optim.py:388] (0/8) Scaling gradients by 0.01955481991171837, model_norm_threshold=541.4743041992188 2023-04-27 12:24:01,590 INFO [optim.py:450] (0/8) 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 12:24:30,815 INFO [train.py:904] (0/8) Epoch 1, batch 250, loss[loss=0.888, simple_loss=0.7489, pruned_loss=0.8739, over 16808.00 frames. ], tot_loss[loss=1.164, simple_loss=1.002, pruned_loss=1.214, over 2374199.32 frames. ], batch size: 124, lr: 3.75e-02, grad_scale: 0.0078125 2023-04-27 12:24:33,601 WARNING [optim.py:388] (0/8) Scaling gradients by 0.057925041764974594, model_norm_threshold=541.4743041992188 2023-04-27 12:24:33,739 INFO [optim.py:450] (0/8) 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 12:25:16,008 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:25:19,910 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:25:21,112 WARNING [train.py:894] (0/8) Grad scale is small: 0.0078125 2023-04-27 12:25:21,112 INFO [train.py:904] (0/8) Epoch 1, batch 300, loss[loss=0.8398, simple_loss=0.7038, pruned_loss=0.8023, over 16674.00 frames. ], tot_loss[loss=1.093, simple_loss=0.935, pruned_loss=1.122, over 2578531.96 frames. ], batch size: 89, lr: 4.00e-02, grad_scale: 0.015625 2023-04-27 12:25:30,081 INFO [optim.py:368] (0/8) 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 12:25:41,583 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5363, 4.3114, 4.3606, 4.3845, 4.3641, 4.3819, 4.2340, 4.0291], device='cuda:0'), covar=tensor([0.0663, 0.0090, 0.0048, 0.0022, 0.0035, 0.0032, 0.0107, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009, 0.0009], device='cuda:0'), out_proj_covar=tensor([9.0275e-06, 8.9292e-06, 8.8654e-06, 8.8594e-06, 8.8205e-06, 9.0260e-06, 8.8186e-06, 9.1037e-06], device='cuda:0') 2023-04-27 12:25:44,465 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=4.60 vs. limit=2.0 2023-04-27 12:25:57,798 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=3.91 vs. limit=2.0 2023-04-27 12:26:05,418 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=48.29 vs. limit=5.0 2023-04-27 12:26:12,608 INFO [train.py:904] (0/8) Epoch 1, batch 350, loss[loss=0.8905, simple_loss=0.7418, pruned_loss=0.829, over 16952.00 frames. ], tot_loss[loss=1.041, simple_loss=0.8849, pruned_loss=1.051, over 2740170.10 frames. ], batch size: 116, lr: 4.25e-02, grad_scale: 0.015625 2023-04-27 12:26:18,629 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=357.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:26:52,678 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=387.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:27:06,013 INFO [train.py:904] (0/8) Epoch 1, batch 400, loss[loss=0.8588, simple_loss=0.7041, pruned_loss=0.8035, over 16563.00 frames. ], tot_loss[loss=1.006, simple_loss=0.8489, pruned_loss=0.9992, over 2868386.97 frames. ], batch size: 68, lr: 4.50e-02, grad_scale: 0.03125 2023-04-27 12:27:17,883 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 7.937e+01 1.110e+02 1.413e+02 1.837e+02 3.136e+02, threshold=2.826e+02, percent-clipped=0.0 2023-04-27 12:27:46,076 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=439.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:27:56,256 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=448.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:27:59,558 INFO [train.py:904] (0/8) Epoch 1, batch 450, loss[loss=1.013, simple_loss=0.8211, pruned_loss=0.9401, over 17144.00 frames. ], tot_loss[loss=0.9805, simple_loss=0.8215, pruned_loss=0.9586, over 2974107.18 frames. ], batch size: 49, lr: 4.75e-02, grad_scale: 0.03125 2023-04-27 12:28:07,878 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=28.54 vs. limit=5.0 2023-04-27 12:28:51,185 INFO [train.py:904] (0/8) Epoch 1, batch 500, loss[loss=0.8794, simple_loss=0.7179, pruned_loss=0.7726, over 16235.00 frames. ], tot_loss[loss=0.9641, simple_loss=0.8015, pruned_loss=0.928, over 3061719.97 frames. ], batch size: 165, lr: 4.99e-02, grad_scale: 0.0625 2023-04-27 12:29:01,372 INFO [optim.py:368] (0/8) 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 12:29:21,023 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=8.78 vs. limit=2.0 2023-04-27 12:29:44,940 INFO [train.py:904] (0/8) Epoch 1, batch 550, loss[loss=0.8855, simple_loss=0.7104, pruned_loss=0.7845, over 17226.00 frames. ], tot_loss[loss=0.9495, simple_loss=0.7841, pruned_loss=0.8982, over 3112843.04 frames. ], batch size: 44, lr: 4.98e-02, grad_scale: 0.0625 2023-04-27 12:29:58,148 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:03,857 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=568.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:30:16,595 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=5.05 vs. limit=2.0 2023-04-27 12:30:26,937 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=590.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:30:38,175 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=600.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:30:38,754 INFO [train.py:904] (0/8) Epoch 1, batch 600, loss[loss=0.9659, simple_loss=0.7778, pruned_loss=0.8233, over 16636.00 frames. ], tot_loss[loss=0.9395, simple_loss=0.7715, pruned_loss=0.8719, over 3152782.74 frames. ], batch size: 62, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:30:48,387 INFO [optim.py:368] (0/8) 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 12:31:02,475 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=623.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:31:07,553 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=629.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:31:28,006 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=648.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:31:31,010 INFO [train.py:904] (0/8) Epoch 1, batch 650, loss[loss=0.8215, simple_loss=0.661, pruned_loss=0.6823, over 16795.00 frames. ], tot_loss[loss=0.9262, simple_loss=0.7573, pruned_loss=0.8418, over 3184713.12 frames. ], batch size: 39, lr: 4.98e-02, grad_scale: 0.125 2023-04-27 12:31:31,407 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=651.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:31:32,146 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=652.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:31:32,367 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=2.07 vs. limit=2.0 2023-04-27 12:31:38,718 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=8.71 vs. limit=5.0 2023-04-27 12:31:43,337 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=6.21 vs. limit=5.0 2023-04-27 12:32:22,402 INFO [train.py:904] (0/8) Epoch 1, batch 700, loss[loss=0.838, simple_loss=0.6747, pruned_loss=0.6778, over 16803.00 frames. ], tot_loss[loss=0.9132, simple_loss=0.7454, pruned_loss=0.8096, over 3210079.30 frames. ], batch size: 39, lr: 4.98e-02, grad_scale: 0.25 2023-04-27 12:32:28,248 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.83 vs. limit=2.0 2023-04-27 12:32:31,781 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.097e+02 2.100e+02 3.098e+02 3.988e+02 9.500e+02, threshold=6.196e+02, percent-clipped=26.0 2023-04-27 12:32:41,291 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.32 vs. limit=5.0 2023-04-27 12:33:00,811 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=739.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:33:05,889 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=743.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:33:14,017 INFO [train.py:904] (0/8) Epoch 1, batch 750, loss[loss=0.8423, simple_loss=0.6984, pruned_loss=0.628, over 17003.00 frames. ], tot_loss[loss=0.8981, simple_loss=0.7341, pruned_loss=0.7732, over 3235222.07 frames. ], batch size: 55, lr: 4.97e-02, grad_scale: 0.25 2023-04-27 12:33:51,298 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=787.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:34:06,259 INFO [train.py:904] (0/8) Epoch 1, batch 800, loss[loss=0.7936, simple_loss=0.6713, pruned_loss=0.5577, over 17073.00 frames. ], tot_loss[loss=0.8689, simple_loss=0.7136, pruned_loss=0.7244, over 3251348.20 frames. ], batch size: 53, lr: 4.97e-02, grad_scale: 0.5 2023-04-27 12:34:17,533 INFO [optim.py:368] (0/8) 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 12:34:54,109 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=847.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:34:58,594 INFO [train.py:904] (0/8) Epoch 1, batch 850, loss[loss=0.6934, simple_loss=0.5907, pruned_loss=0.4729, over 16851.00 frames. ], tot_loss[loss=0.8361, simple_loss=0.6918, pruned_loss=0.6726, over 3277566.87 frames. ], batch size: 90, lr: 4.96e-02, grad_scale: 0.5 2023-04-27 12:35:50,499 INFO [train.py:904] (0/8) Epoch 1, batch 900, loss[loss=0.659, simple_loss=0.5659, pruned_loss=0.4364, over 16871.00 frames. ], tot_loss[loss=0.8019, simple_loss=0.6691, pruned_loss=0.6227, over 3283984.32 frames. ], batch size: 90, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:35:57,038 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=908.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:36:00,426 INFO [optim.py:368] (0/8) 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 12:36:08,850 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=918.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:36:14,429 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=924.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:36:37,796 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=946.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:36:42,346 INFO [train.py:904] (0/8) Epoch 1, batch 950, loss[loss=0.671, simple_loss=0.5987, pruned_loss=0.4066, over 17022.00 frames. ], tot_loss[loss=0.7698, simple_loss=0.6478, pruned_loss=0.5775, over 3287522.65 frames. ], batch size: 50, lr: 4.96e-02, grad_scale: 1.0 2023-04-27 12:36:44,109 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=952.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:37:33,570 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1944, 4.3094, 4.0239, 3.8597, 4.2971, 4.4049, 3.9307, 4.4056], device='cuda:0'), covar=tensor([0.3263, 0.3435, 0.3346, 0.4087, 0.3394, 0.2332, 0.3771, 0.2641], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0078, 0.0080, 0.0083, 0.0078, 0.0066, 0.0092, 0.0077], device='cuda:0'), out_proj_covar=tensor([6.3195e-05, 7.2426e-05, 7.4588e-05, 7.4783e-05, 7.2775e-05, 6.4000e-05, 8.3850e-05, 6.7637e-05], device='cuda:0') 2023-04-27 12:37:35,475 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1000.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:37:36,003 INFO [train.py:904] (0/8) Epoch 1, batch 1000, loss[loss=0.5783, simple_loss=0.5169, pruned_loss=0.3465, over 17246.00 frames. ], tot_loss[loss=0.7361, simple_loss=0.6253, pruned_loss=0.5337, over 3290761.67 frames. ], batch size: 43, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:37:39,955 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 12:37:46,296 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.400e+02 3.913e+02 4.932e+02 6.154e+02 1.349e+03, threshold=9.864e+02, percent-clipped=6.0 2023-04-27 12:37:51,292 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([6.6536, 6.3843, 6.6123, 6.6583, 6.6574, 6.6563, 6.6580, 6.6509], device='cuda:0'), covar=tensor([0.0164, 0.0782, 0.0285, 0.0161, 0.0130, 0.0224, 0.0167, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0044, 0.0067, 0.0056, 0.0044, 0.0042, 0.0046, 0.0049, 0.0044], device='cuda:0'), out_proj_covar=tensor([3.9092e-05, 5.8056e-05, 4.8526e-05, 3.6184e-05, 3.7448e-05, 3.7438e-05, 4.4674e-05, 3.7584e-05], device='cuda:0') 2023-04-27 12:38:21,629 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1043.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:38:29,495 INFO [train.py:904] (0/8) Epoch 1, batch 1050, loss[loss=0.5879, simple_loss=0.5201, pruned_loss=0.3569, over 16944.00 frames. ], tot_loss[loss=0.7081, simple_loss=0.6074, pruned_loss=0.4963, over 3301621.67 frames. ], batch size: 116, lr: 4.95e-02, grad_scale: 1.0 2023-04-27 12:39:12,047 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1091.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:39:21,805 INFO [train.py:904] (0/8) Epoch 1, batch 1100, loss[loss=0.5578, simple_loss=0.5114, pruned_loss=0.3144, over 16949.00 frames. ], tot_loss[loss=0.6794, simple_loss=0.589, pruned_loss=0.4603, over 3306260.48 frames. ], batch size: 41, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:39:33,208 INFO [optim.py:368] (0/8) 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 12:39:44,220 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.98 vs. limit=2.0 2023-04-27 12:40:17,545 INFO [train.py:904] (0/8) Epoch 1, batch 1150, loss[loss=0.6057, simple_loss=0.5601, pruned_loss=0.3346, over 17064.00 frames. ], tot_loss[loss=0.6527, simple_loss=0.5719, pruned_loss=0.4281, over 3312569.32 frames. ], batch size: 53, lr: 4.94e-02, grad_scale: 1.0 2023-04-27 12:40:19,531 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1153.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:41:04,204 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-04-27 12:41:10,400 INFO [train.py:904] (0/8) Epoch 1, batch 1200, loss[loss=0.5778, simple_loss=0.5069, pruned_loss=0.3483, over 16640.00 frames. ], tot_loss[loss=0.6293, simple_loss=0.5565, pruned_loss=0.4007, over 3323178.63 frames. ], batch size: 134, lr: 4.93e-02, grad_scale: 2.0 2023-04-27 12:41:12,653 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1203.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 12:41:21,561 INFO [optim.py:368] (0/8) 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 12:41:25,709 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1214.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:41:30,047 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1218.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:41:34,769 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1224.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:41:57,159 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1246.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:42:03,017 INFO [train.py:904] (0/8) Epoch 1, batch 1250, loss[loss=0.5421, simple_loss=0.5135, pruned_loss=0.2854, over 17266.00 frames. ], tot_loss[loss=0.6091, simple_loss=0.5436, pruned_loss=0.3775, over 3320725.03 frames. ], batch size: 52, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:42:17,898 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1266.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:42:24,721 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:49,319 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1294.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:42:56,584 INFO [train.py:904] (0/8) Epoch 1, batch 1300, loss[loss=0.5072, simple_loss=0.4702, pruned_loss=0.277, over 15474.00 frames. ], tot_loss[loss=0.5917, simple_loss=0.5322, pruned_loss=0.358, over 3321552.39 frames. ], batch size: 190, lr: 4.92e-02, grad_scale: 2.0 2023-04-27 12:43:07,780 INFO [optim.py:368] (0/8) 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 12:43:51,778 INFO [train.py:904] (0/8) Epoch 1, batch 1350, loss[loss=0.5476, simple_loss=0.5246, pruned_loss=0.2827, over 17018.00 frames. ], tot_loss[loss=0.5752, simple_loss=0.5224, pruned_loss=0.3395, over 3318758.96 frames. ], batch size: 50, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:44:49,673 INFO [train.py:904] (0/8) Epoch 1, batch 1400, loss[loss=0.4866, simple_loss=0.4733, pruned_loss=0.2451, over 17196.00 frames. ], tot_loss[loss=0.5606, simple_loss=0.513, pruned_loss=0.3242, over 3308611.76 frames. ], batch size: 46, lr: 4.91e-02, grad_scale: 2.0 2023-04-27 12:45:00,104 INFO [optim.py:368] (0/8) 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 12:45:27,135 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1434.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:45:46,441 INFO [train.py:904] (0/8) Epoch 1, batch 1450, loss[loss=0.5014, simple_loss=0.4643, pruned_loss=0.2729, over 16535.00 frames. ], tot_loss[loss=0.5458, simple_loss=0.5033, pruned_loss=0.3097, over 3309037.44 frames. ], batch size: 75, lr: 4.90e-02, grad_scale: 2.0 2023-04-27 12:46:33,633 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9713, 4.8567, 5.0671, 5.3201, 4.9123, 5.1812, 4.6208, 4.6909], device='cuda:0'), covar=tensor([0.0929, 0.1032, 0.0709, 0.0526, 0.1528, 0.0612, 0.1232, 0.2290], device='cuda:0'), in_proj_covar=tensor([0.0031, 0.0032, 0.0034, 0.0025, 0.0029, 0.0031, 0.0026, 0.0025], device='cuda:0'), out_proj_covar=tensor([2.2173e-05, 2.3471e-05, 2.5155e-05, 1.8752e-05, 2.1984e-05, 2.2718e-05, 1.9441e-05, 2.0247e-05], device='cuda:0') 2023-04-27 12:46:34,552 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1495.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 12:46:41,361 INFO [train.py:904] (0/8) Epoch 1, batch 1500, loss[loss=0.475, simple_loss=0.4444, pruned_loss=0.2544, over 16783.00 frames. ], tot_loss[loss=0.5323, simple_loss=0.4945, pruned_loss=0.2968, over 3318453.84 frames. ], batch size: 83, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:46:44,254 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1503.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:46:51,068 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1509.0, num_to_drop=2, layers_to_drop={1, 3} 2023-04-27 12:46:52,707 INFO [optim.py:368] (0/8) 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 12:46:57,601 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3968, 5.2002, 5.2354, 5.4724, 4.8774, 5.2742, 5.4026, 4.6462], device='cuda:0'), covar=tensor([0.0665, 0.0274, 0.0524, 0.0234, 0.0414, 0.0469, 0.0282, 0.0506], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0060, 0.0075, 0.0061, 0.0066, 0.0067, 0.0058, 0.0064], device='cuda:0'), out_proj_covar=tensor([6.2524e-05, 4.9741e-05, 7.2090e-05, 5.6954e-05, 6.2351e-05, 6.0371e-05, 5.3960e-05, 5.7800e-05], device='cuda:0') 2023-04-27 12:47:38,778 INFO [train.py:904] (0/8) Epoch 1, batch 1550, loss[loss=0.5179, simple_loss=0.5078, pruned_loss=0.2595, over 16620.00 frames. ], tot_loss[loss=0.5226, simple_loss=0.4896, pruned_loss=0.2864, over 3319189.50 frames. ], batch size: 62, lr: 4.89e-02, grad_scale: 2.0 2023-04-27 12:47:39,033 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:47:56,079 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1566.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:48:35,347 INFO [train.py:904] (0/8) Epoch 1, batch 1600, loss[loss=0.4872, simple_loss=0.4784, pruned_loss=0.2442, over 17099.00 frames. ], tot_loss[loss=0.5152, simple_loss=0.4856, pruned_loss=0.2786, over 3325364.19 frames. ], batch size: 47, lr: 4.88e-02, grad_scale: 4.0 2023-04-27 12:48:44,805 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0554, 5.0715, 4.4716, 5.1904, 5.0039, 5.2153, 5.0266, 5.1180], device='cuda:0'), covar=tensor([0.0262, 0.0270, 0.0694, 0.0557, 0.0581, 0.0227, 0.0356, 0.0345], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0071, 0.0088, 0.0072, 0.0074, 0.0068, 0.0069, 0.0071], device='cuda:0'), out_proj_covar=tensor([6.5461e-05, 6.4403e-05, 8.7673e-05, 7.1555e-05, 6.8592e-05, 6.6574e-05, 6.3996e-05, 6.6183e-05], device='cuda:0') 2023-04-27 12:48:47,168 INFO [optim.py:368] (0/8) 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 12:48:47,520 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1611.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 12:48:53,587 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5012, 5.3825, 4.8838, 5.4093, 5.4657, 3.1336, 5.3780, 4.4682], device='cuda:0'), covar=tensor([0.0137, 0.0083, 0.0514, 0.0097, 0.0151, 0.1502, 0.0127, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0032, 0.0024, 0.0030, 0.0029, 0.0021, 0.0032, 0.0026, 0.0022], device='cuda:0'), out_proj_covar=tensor([2.7181e-05, 1.9873e-05, 2.6067e-05, 2.2437e-05, 1.8596e-05, 2.7845e-05, 2.0839e-05, 1.9713e-05], device='cuda:0') 2023-04-27 12:49:05,645 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1627.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:49:32,291 INFO [train.py:904] (0/8) Epoch 1, batch 1650, loss[loss=0.4401, simple_loss=0.447, pruned_loss=0.2107, over 16965.00 frames. ], tot_loss[loss=0.5104, simple_loss=0.4838, pruned_loss=0.2729, over 3323036.42 frames. ], batch size: 41, lr: 4.87e-02, grad_scale: 4.0 2023-04-27 12:49:57,913 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1672.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:50:29,827 INFO [train.py:904] (0/8) Epoch 1, batch 1700, loss[loss=0.3772, simple_loss=0.3944, pruned_loss=0.174, over 16865.00 frames. ], tot_loss[loss=0.5046, simple_loss=0.4819, pruned_loss=0.2664, over 3321882.47 frames. ], batch size: 42, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:50:41,830 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.189e+02 5.218e+02 6.714e+02 8.049e+02 1.427e+03, threshold=1.343e+03, percent-clipped=1.0 2023-04-27 12:50:48,596 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-27 12:51:28,774 INFO [train.py:904] (0/8) Epoch 1, batch 1750, loss[loss=0.4682, simple_loss=0.4734, pruned_loss=0.2275, over 16540.00 frames. ], tot_loss[loss=0.4961, simple_loss=0.4775, pruned_loss=0.2589, over 3317463.38 frames. ], batch size: 68, lr: 4.86e-02, grad_scale: 4.0 2023-04-27 12:51:39,995 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3411, 4.2025, 3.8109, 4.5289, 4.3822, 4.5056, 4.2781, 4.4404], device='cuda:0'), covar=tensor([0.0394, 0.0488, 0.1159, 0.0477, 0.0595, 0.0359, 0.0670, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0081, 0.0103, 0.0082, 0.0082, 0.0079, 0.0075, 0.0081], device='cuda:0'), out_proj_covar=tensor([7.5335e-05, 7.4393e-05, 1.0433e-04, 8.0628e-05, 7.7742e-05, 7.5985e-05, 7.1564e-05, 7.8320e-05], device='cuda:0') 2023-04-27 12:52:15,580 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1790.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 12:52:27,941 INFO [train.py:904] (0/8) Epoch 1, batch 1800, loss[loss=0.4053, simple_loss=0.4081, pruned_loss=0.1987, over 16801.00 frames. ], tot_loss[loss=0.4906, simple_loss=0.4757, pruned_loss=0.2533, over 3319113.46 frames. ], batch size: 39, lr: 4.85e-02, grad_scale: 4.0 2023-04-27 12:52:37,455 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=1809.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:52:38,892 INFO [optim.py:368] (0/8) 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 12:53:10,969 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1838.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:25,093 INFO [train.py:904] (0/8) Epoch 1, batch 1850, loss[loss=0.4688, simple_loss=0.4847, pruned_loss=0.2234, over 16712.00 frames. ], tot_loss[loss=0.4842, simple_loss=0.473, pruned_loss=0.2476, over 3329419.03 frames. ], batch size: 57, lr: 4.84e-02, grad_scale: 4.0 2023-04-27 12:53:32,843 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=1857.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:53:33,002 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1857.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:53:48,599 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=1870.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:54:21,395 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1899.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:54:23,735 INFO [train.py:904] (0/8) Epoch 1, batch 1900, loss[loss=0.4518, simple_loss=0.4462, pruned_loss=0.2278, over 16865.00 frames. ], tot_loss[loss=0.4734, simple_loss=0.4673, pruned_loss=0.2392, over 3313445.26 frames. ], batch size: 96, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:54:36,246 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.249e+02 5.423e+02 6.685e+02 8.506e+02 1.861e+03, threshold=1.337e+03, percent-clipped=8.0 2023-04-27 12:54:43,909 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1918.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 12:54:49,111 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1922.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 12:55:00,056 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=1931.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 12:55:22,825 INFO [train.py:904] (0/8) Epoch 1, batch 1950, loss[loss=0.4448, simple_loss=0.4499, pruned_loss=0.2192, over 16506.00 frames. ], tot_loss[loss=0.4645, simple_loss=0.4631, pruned_loss=0.2322, over 3327509.33 frames. ], batch size: 146, lr: 4.83e-02, grad_scale: 4.0 2023-04-27 12:55:42,118 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=1967.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:55:53,413 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9247, 5.0137, 4.6704, 5.3522, 5.0342, 5.3063, 5.2771, 5.0522], device='cuda:0'), covar=tensor([0.0336, 0.0378, 0.0693, 0.0536, 0.0521, 0.0276, 0.0279, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0095, 0.0095, 0.0124, 0.0100, 0.0095, 0.0096, 0.0087, 0.0100], device='cuda:0'), out_proj_covar=tensor([9.0226e-05, 9.1946e-05, 1.3017e-04, 1.0592e-04, 9.3855e-05, 9.5339e-05, 8.5127e-05, 1.0105e-04], device='cuda:0') 2023-04-27 12:55:59,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2029, 5.2512, 4.8854, 5.4647, 5.2978, 5.3966, 5.4543, 5.2681], device='cuda:0'), covar=tensor([0.0219, 0.0242, 0.0557, 0.0378, 0.0323, 0.0259, 0.0196, 0.0292], device='cuda:0'), in_proj_covar=tensor([0.0095, 0.0095, 0.0124, 0.0100, 0.0094, 0.0096, 0.0087, 0.0099], device='cuda:0'), out_proj_covar=tensor([8.9885e-05, 9.1649e-05, 1.2987e-04, 1.0551e-04, 9.4068e-05, 9.5317e-05, 8.5268e-05, 1.0097e-04], device='cuda:0') 2023-04-27 12:56:19,053 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-2000.pt 2023-04-27 12:56:23,568 INFO [train.py:904] (0/8) Epoch 1, batch 2000, loss[loss=0.4578, simple_loss=0.4625, pruned_loss=0.2266, over 15459.00 frames. ], tot_loss[loss=0.4573, simple_loss=0.4588, pruned_loss=0.2272, over 3331382.49 frames. ], batch size: 191, lr: 4.82e-02, grad_scale: 8.0 2023-04-27 12:56:36,804 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.491e+02 5.646e+02 7.242e+02 8.659e+02 1.834e+03, threshold=1.448e+03, percent-clipped=2.0 2023-04-27 12:57:27,607 INFO [train.py:904] (0/8) Epoch 1, batch 2050, loss[loss=0.4272, simple_loss=0.4293, pruned_loss=0.2126, over 16281.00 frames. ], tot_loss[loss=0.4473, simple_loss=0.4531, pruned_loss=0.2203, over 3322081.24 frames. ], batch size: 165, lr: 4.81e-02, grad_scale: 8.0 2023-04-27 12:58:04,134 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3033, 5.2236, 4.8679, 5.4888, 5.3591, 5.4378, 5.4804, 5.3613], device='cuda:0'), covar=tensor([0.0213, 0.0224, 0.0640, 0.0296, 0.0340, 0.0222, 0.0210, 0.0268], device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0096, 0.0129, 0.0105, 0.0097, 0.0101, 0.0089, 0.0101], device='cuda:0'), out_proj_covar=tensor([9.2637e-05, 9.5216e-05, 1.3788e-04, 1.1260e-04, 1.0017e-04, 1.0124e-04, 8.8449e-05, 1.0589e-04], device='cuda:0') 2023-04-27 12:58:19,057 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2090.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 12:58:32,227 INFO [train.py:904] (0/8) Epoch 1, batch 2100, loss[loss=0.4291, simple_loss=0.4506, pruned_loss=0.2038, over 16628.00 frames. ], tot_loss[loss=0.4392, simple_loss=0.4483, pruned_loss=0.2147, over 3324229.90 frames. ], batch size: 57, lr: 4.80e-02, grad_scale: 16.0 2023-04-27 12:58:45,937 INFO [optim.py:368] (0/8) 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 12:59:19,750 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2138.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 12:59:35,905 INFO [train.py:904] (0/8) Epoch 1, batch 2150, loss[loss=0.4515, simple_loss=0.4472, pruned_loss=0.2279, over 16504.00 frames. ], tot_loss[loss=0.4343, simple_loss=0.446, pruned_loss=0.211, over 3324723.70 frames. ], batch size: 146, lr: 4.79e-02, grad_scale: 16.0 2023-04-27 12:59:42,749 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3317, 3.5459, 3.7045, 3.6654, 3.5139, 3.0528, 3.5511, 3.8718], device='cuda:0'), covar=tensor([0.0289, 0.0221, 0.0154, 0.0264, 0.0154, 0.0414, 0.0188, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0044, 0.0045, 0.0039, 0.0043, 0.0044, 0.0040, 0.0041, 0.0037], device='cuda:0'), out_proj_covar=tensor([3.7070e-05, 3.8529e-05, 3.2018e-05, 3.5084e-05, 3.7702e-05, 3.2430e-05, 3.2757e-05, 3.1098e-05], device='cuda:0') 2023-04-27 13:00:03,377 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3991, 4.3761, 4.2277, 4.6362, 4.5075, 4.7550, 4.6147, 4.6369], device='cuda:0'), covar=tensor([0.0295, 0.0315, 0.0701, 0.0413, 0.0541, 0.0235, 0.0279, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0098, 0.0137, 0.0111, 0.0101, 0.0104, 0.0091, 0.0104], device='cuda:0'), out_proj_covar=tensor([1.0010e-04, 1.0008e-04, 1.4710e-04, 1.2072e-04, 1.0786e-04, 1.0473e-04, 9.2154e-05, 1.1048e-04], device='cuda:0') 2023-04-27 13:00:31,474 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:40,262 INFO [train.py:904] (0/8) Epoch 1, batch 2200, loss[loss=0.376, simple_loss=0.4208, pruned_loss=0.1655, over 17038.00 frames. ], tot_loss[loss=0.4285, simple_loss=0.4428, pruned_loss=0.2069, over 3324729.87 frames. ], batch size: 50, lr: 4.78e-02, grad_scale: 16.0 2023-04-27 13:00:53,197 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.805e+02 5.075e+02 6.122e+02 8.359e+02 1.658e+03, threshold=1.224e+03, percent-clipped=12.0 2023-04-27 13:00:56,685 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2213.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:00:56,890 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2023, 4.3328, 4.7516, 4.6734, 4.2201, 4.7679, 4.5221, 4.7965], device='cuda:0'), covar=tensor([0.1318, 0.0146, 0.0163, 0.0145, 0.0140, 0.0139, 0.0165, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0057, 0.0035, 0.0031, 0.0025, 0.0022, 0.0029, 0.0037, 0.0031], device='cuda:0'), out_proj_covar=tensor([5.6988e-05, 3.0539e-05, 2.6974e-05, 2.3558e-05, 2.2395e-05, 2.5231e-05, 3.3073e-05, 2.5193e-05], device='cuda:0') 2023-04-27 13:01:07,620 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2222.0, num_to_drop=2, layers_to_drop={1, 2} 2023-04-27 13:01:11,883 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2226.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:01:44,911 INFO [train.py:904] (0/8) Epoch 1, batch 2250, loss[loss=0.4153, simple_loss=0.4531, pruned_loss=0.1887, over 17063.00 frames. ], tot_loss[loss=0.4242, simple_loss=0.4411, pruned_loss=0.2035, over 3308122.72 frames. ], batch size: 53, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:01:55,634 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3485, 5.1914, 4.2761, 4.8420, 5.3304, 5.4569, 4.7508, 5.3121], device='cuda:0'), covar=tensor([0.0205, 0.0332, 0.0379, 0.0270, 0.0093, 0.0129, 0.0187, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0048, 0.0048, 0.0065, 0.0061, 0.0046, 0.0051, 0.0060, 0.0059], device='cuda:0'), out_proj_covar=tensor([4.3379e-05, 4.7584e-05, 6.9913e-05, 5.9873e-05, 3.8974e-05, 4.5851e-05, 6.1245e-05, 5.8181e-05], device='cuda:0') 2023-04-27 13:02:05,802 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2267.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:02:08,818 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2270.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:02:49,162 INFO [train.py:904] (0/8) Epoch 1, batch 2300, loss[loss=0.3923, simple_loss=0.4122, pruned_loss=0.1863, over 16411.00 frames. ], tot_loss[loss=0.4177, simple_loss=0.4374, pruned_loss=0.1988, over 3311498.28 frames. ], batch size: 146, lr: 4.77e-02, grad_scale: 16.0 2023-04-27 13:03:01,868 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.985e+02 4.330e+02 5.680e+02 7.295e+02 1.284e+03, threshold=1.136e+03, percent-clipped=1.0 2023-04-27 13:03:07,889 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2315.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:03:20,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1710, 3.4448, 3.2973, 2.8770, 2.4437, 2.8587, 3.5825, 3.3327], device='cuda:0'), covar=tensor([0.2037, 0.1232, 0.0926, 0.1716, 0.2296, 0.2328, 0.0718, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0025, 0.0026, 0.0033, 0.0034, 0.0035, 0.0037, 0.0022, 0.0023], device='cuda:0'), out_proj_covar=tensor([2.3865e-05, 2.1730e-05, 2.8541e-05, 2.9286e-05, 2.9065e-05, 3.0961e-05, 1.8353e-05, 2.0801e-05], device='cuda:0') 2023-04-27 13:03:39,616 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2340.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:03:53,146 INFO [train.py:904] (0/8) Epoch 1, batch 2350, loss[loss=0.4274, simple_loss=0.4346, pruned_loss=0.2101, over 16888.00 frames. ], tot_loss[loss=0.4113, simple_loss=0.4335, pruned_loss=0.1945, over 3321444.29 frames. ], batch size: 116, lr: 4.76e-02, grad_scale: 16.0 2023-04-27 13:04:00,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9146, 4.8460, 4.0149, 4.4760, 4.8056, 4.8842, 4.3118, 4.7753], device='cuda:0'), covar=tensor([0.0159, 0.0235, 0.0295, 0.0242, 0.0111, 0.0162, 0.0202, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0047, 0.0046, 0.0063, 0.0059, 0.0045, 0.0049, 0.0058, 0.0057], device='cuda:0'), out_proj_covar=tensor([4.3662e-05, 4.5456e-05, 7.0563e-05, 5.8931e-05, 3.8655e-05, 4.5020e-05, 6.1200e-05, 5.7550e-05], device='cuda:0') 2023-04-27 13:04:23,067 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-27 13:04:35,650 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2320, 4.1799, 4.4200, 4.4671, 4.8294, 4.2574, 4.1833, 4.6468], device='cuda:0'), covar=tensor([0.0471, 0.0384, 0.0514, 0.0477, 0.0352, 0.0427, 0.0536, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0095, 0.0111, 0.0111, 0.0105, 0.0101, 0.0105, 0.0090], device='cuda:0'), out_proj_covar=tensor([9.7099e-05, 1.0720e-04, 1.1896e-04, 1.0961e-04, 1.1318e-04, 1.0510e-04, 1.0487e-04, 8.5553e-05], device='cuda:0') 2023-04-27 13:04:54,414 INFO [train.py:904] (0/8) Epoch 1, batch 2400, loss[loss=0.4347, simple_loss=0.4511, pruned_loss=0.2092, over 16361.00 frames. ], tot_loss[loss=0.4079, simple_loss=0.4318, pruned_loss=0.1919, over 3329043.58 frames. ], batch size: 165, lr: 4.75e-02, grad_scale: 16.0 2023-04-27 13:04:55,578 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2401.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 13:05:07,299 INFO [optim.py:368] (0/8) 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 13:05:37,583 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-27 13:05:55,899 INFO [train.py:904] (0/8) Epoch 1, batch 2450, loss[loss=0.4114, simple_loss=0.4354, pruned_loss=0.1937, over 16495.00 frames. ], tot_loss[loss=0.4048, simple_loss=0.4313, pruned_loss=0.189, over 3325105.11 frames. ], batch size: 146, lr: 4.74e-02, grad_scale: 16.0 2023-04-27 13:06:51,093 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2494.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:06:58,993 INFO [train.py:904] (0/8) Epoch 1, batch 2500, loss[loss=0.3514, simple_loss=0.3966, pruned_loss=0.1531, over 17177.00 frames. ], tot_loss[loss=0.3999, simple_loss=0.428, pruned_loss=0.1859, over 3316820.01 frames. ], batch size: 46, lr: 4.73e-02, grad_scale: 16.0 2023-04-27 13:07:09,740 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3538, 3.9288, 3.9224, 4.4090, 3.6298, 4.3326, 3.4790, 3.9634], device='cuda:0'), covar=tensor([0.1812, 0.0145, 0.0231, 0.0095, 0.0242, 0.0157, 0.0304, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0037, 0.0034, 0.0026, 0.0023, 0.0031, 0.0042, 0.0034], device='cuda:0'), out_proj_covar=tensor([7.2890e-05, 3.4027e-05, 3.2714e-05, 2.5992e-05, 2.5426e-05, 2.9116e-05, 3.9901e-05, 2.9754e-05], device='cuda:0') 2023-04-27 13:07:11,504 INFO [optim.py:368] (0/8) 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 13:07:14,731 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:07:50,937 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2542.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:01,515 INFO [train.py:904] (0/8) Epoch 1, batch 2550, loss[loss=0.378, simple_loss=0.4354, pruned_loss=0.1603, over 17139.00 frames. ], tot_loss[loss=0.3969, simple_loss=0.4259, pruned_loss=0.1839, over 3318113.08 frames. ], batch size: 48, lr: 4.72e-02, grad_scale: 16.0 2023-04-27 13:08:15,981 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2561.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:08:32,315 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=2574.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:09:07,702 INFO [train.py:904] (0/8) Epoch 1, batch 2600, loss[loss=0.4027, simple_loss=0.4425, pruned_loss=0.1815, over 16500.00 frames. ], tot_loss[loss=0.3926, simple_loss=0.4232, pruned_loss=0.1809, over 3322737.91 frames. ], batch size: 68, lr: 4.71e-02, grad_scale: 16.0 2023-04-27 13:09:20,093 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.935e+02 4.794e+02 6.148e+02 7.560e+02 1.171e+03, threshold=1.230e+03, percent-clipped=4.0 2023-04-27 13:10:11,898 INFO [train.py:904] (0/8) Epoch 1, batch 2650, loss[loss=0.3664, simple_loss=0.423, pruned_loss=0.1549, over 17136.00 frames. ], tot_loss[loss=0.387, simple_loss=0.4209, pruned_loss=0.1765, over 3330506.84 frames. ], batch size: 48, lr: 4.70e-02, grad_scale: 16.0 2023-04-27 13:10:39,229 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0228, 5.3243, 5.0441, 5.4144, 4.8502, 5.2432, 4.9540, 5.3133], device='cuda:0'), covar=tensor([0.0409, 0.0538, 0.0428, 0.0234, 0.0580, 0.0314, 0.0431, 0.0273], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0138, 0.0119, 0.0093, 0.0125, 0.0108, 0.0116, 0.0091], device='cuda:0'), out_proj_covar=tensor([1.0917e-04, 1.2465e-04, 1.0020e-04, 7.0258e-05, 1.0476e-04, 8.5894e-05, 9.9815e-05, 8.0720e-05], device='cuda:0') 2023-04-27 13:10:50,414 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.23 vs. limit=5.0 2023-04-27 13:10:55,308 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2685.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:11:09,975 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=2696.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:11:15,043 INFO [train.py:904] (0/8) Epoch 1, batch 2700, loss[loss=0.3976, simple_loss=0.4305, pruned_loss=0.1824, over 16741.00 frames. ], tot_loss[loss=0.3822, simple_loss=0.4186, pruned_loss=0.1729, over 3322647.25 frames. ], batch size: 83, lr: 4.69e-02, grad_scale: 16.0 2023-04-27 13:11:29,646 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.771e+02 4.333e+02 5.264e+02 6.238e+02 1.242e+03, threshold=1.053e+03, percent-clipped=1.0 2023-04-27 13:12:13,560 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2746.0, num_to_drop=2, layers_to_drop={0, 3} 2023-04-27 13:12:20,245 INFO [train.py:904] (0/8) Epoch 1, batch 2750, loss[loss=0.3947, simple_loss=0.4485, pruned_loss=0.1704, over 17257.00 frames. ], tot_loss[loss=0.3794, simple_loss=0.4169, pruned_loss=0.1709, over 3314432.37 frames. ], batch size: 52, lr: 4.68e-02, grad_scale: 16.0 2023-04-27 13:12:24,644 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=2.81 vs. limit=2.0 2023-04-27 13:12:48,174 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-27 13:13:23,782 INFO [train.py:904] (0/8) Epoch 1, batch 2800, loss[loss=0.3616, simple_loss=0.3942, pruned_loss=0.1645, over 16820.00 frames. ], tot_loss[loss=0.3755, simple_loss=0.4142, pruned_loss=0.1684, over 3314945.20 frames. ], batch size: 102, lr: 4.67e-02, grad_scale: 16.0 2023-04-27 13:13:26,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2082, 4.1234, 3.9498, 4.4333, 4.2936, 4.4836, 4.3644, 4.2431], device='cuda:0'), covar=tensor([0.0307, 0.0330, 0.0922, 0.0338, 0.0437, 0.0276, 0.0358, 0.0361], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0119, 0.0193, 0.0145, 0.0123, 0.0136, 0.0116, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 13:13:35,363 INFO [optim.py:368] (0/8) 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 13:13:54,292 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2824.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:14:03,047 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.84 vs. limit=5.0 2023-04-27 13:14:26,008 INFO [train.py:904] (0/8) Epoch 1, batch 2850, loss[loss=0.295, simple_loss=0.3561, pruned_loss=0.117, over 16852.00 frames. ], tot_loss[loss=0.3714, simple_loss=0.411, pruned_loss=0.1658, over 3323706.81 frames. ], batch size: 42, lr: 4.66e-02, grad_scale: 16.0 2023-04-27 13:14:57,948 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8307, 4.6562, 4.0748, 4.2536, 4.7107, 4.7427, 4.2779, 4.6823], device='cuda:0'), covar=tensor([0.0142, 0.0146, 0.0197, 0.0257, 0.0094, 0.0143, 0.0153, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0051, 0.0047, 0.0067, 0.0065, 0.0047, 0.0054, 0.0062, 0.0058], device='cuda:0'), out_proj_covar=tensor([5.8812e-05, 5.3030e-05, 9.1082e-05, 7.8202e-05, 4.8529e-05, 5.8105e-05, 7.7988e-05, 7.2213e-05], device='cuda:0') 2023-04-27 13:15:09,663 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2885.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 13:15:27,289 INFO [train.py:904] (0/8) Epoch 1, batch 2900, loss[loss=0.3588, simple_loss=0.3862, pruned_loss=0.1658, over 16933.00 frames. ], tot_loss[loss=0.3696, simple_loss=0.4083, pruned_loss=0.1654, over 3333676.07 frames. ], batch size: 96, lr: 4.65e-02, grad_scale: 16.0 2023-04-27 13:15:40,741 INFO [optim.py:368] (0/8) 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 13:16:13,959 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=2937.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:16:19,064 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 13:16:32,278 INFO [train.py:904] (0/8) Epoch 1, batch 2950, loss[loss=0.3979, simple_loss=0.4293, pruned_loss=0.1832, over 16497.00 frames. ], tot_loss[loss=0.3682, simple_loss=0.4066, pruned_loss=0.1649, over 3329568.09 frames. ], batch size: 68, lr: 4.64e-02, grad_scale: 16.0 2023-04-27 13:16:47,447 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-27 13:16:51,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9003, 3.9859, 3.9554, 4.2489, 3.4762, 4.1429, 3.4283, 4.2946], device='cuda:0'), covar=tensor([0.2095, 0.0116, 0.0261, 0.0099, 0.0249, 0.0229, 0.0324, 0.0089], device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0042, 0.0045, 0.0029, 0.0026, 0.0039, 0.0050, 0.0038], device='cuda:0'), out_proj_covar=tensor([9.5897e-05, 3.9402e-05, 4.7007e-05, 3.1475e-05, 3.0789e-05, 4.0802e-05, 5.1296e-05, 3.5990e-05], device='cuda:0') 2023-04-27 13:17:29,841 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=2996.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:17:32,222 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=2998.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 13:17:33,543 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-27 13:17:35,328 INFO [train.py:904] (0/8) Epoch 1, batch 3000, loss[loss=0.3915, simple_loss=0.4111, pruned_loss=0.186, over 16317.00 frames. ], tot_loss[loss=0.3666, simple_loss=0.4051, pruned_loss=0.164, over 3330958.59 frames. ], batch size: 165, lr: 4.63e-02, grad_scale: 16.0 2023-04-27 13:17:35,329 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 13:17:45,053 INFO [train.py:938] (0/8) Epoch 1, validation: loss=0.2847, simple_loss=0.3895, pruned_loss=0.08992, over 944034.00 frames. 2023-04-27 13:17:45,054 INFO [train.py:939] (0/8) Maximum memory allocated so far is 16021MB 2023-04-27 13:17:59,844 INFO [optim.py:368] (0/8) 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 13:18:05,340 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-27 13:18:08,308 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0282, 4.0658, 3.8545, 3.6085, 3.7318, 3.6920, 3.9657, 3.9639], device='cuda:0'), covar=tensor([0.0219, 0.0185, 0.0217, 0.0196, 0.0270, 0.0216, 0.0305, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0045, 0.0037, 0.0037, 0.0043, 0.0040, 0.0043, 0.0046, 0.0041], device='cuda:0'), out_proj_covar=tensor([5.1934e-05, 4.5952e-05, 4.4373e-05, 4.7910e-05, 4.7046e-05, 5.7200e-05, 5.3643e-05, 4.7523e-05], device='cuda:0') 2023-04-27 13:18:09,272 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3019.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:18:36,925 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:18:40,966 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3044.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:18:50,096 INFO [train.py:904] (0/8) Epoch 1, batch 3050, loss[loss=0.3837, simple_loss=0.4059, pruned_loss=0.1808, over 16772.00 frames. ], tot_loss[loss=0.3642, simple_loss=0.4037, pruned_loss=0.1624, over 3325740.66 frames. ], batch size: 116, lr: 4.62e-02, grad_scale: 16.0 2023-04-27 13:19:27,596 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3080.0, num_to_drop=2, layers_to_drop={0, 2} 2023-04-27 13:19:53,987 INFO [train.py:904] (0/8) Epoch 1, batch 3100, loss[loss=0.2972, simple_loss=0.351, pruned_loss=0.1217, over 16860.00 frames. ], tot_loss[loss=0.3604, simple_loss=0.4008, pruned_loss=0.16, over 3331291.08 frames. ], batch size: 42, lr: 4.61e-02, grad_scale: 16.0 2023-04-27 13:20:07,655 INFO [optim.py:368] (0/8) 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 13:20:09,840 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0690, 3.8309, 3.4972, 4.0154, 3.4886, 3.9465, 3.3932, 3.9942], device='cuda:0'), covar=tensor([0.2044, 0.0112, 0.0366, 0.0125, 0.0239, 0.0242, 0.0310, 0.0082], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0045, 0.0051, 0.0031, 0.0029, 0.0043, 0.0054, 0.0040], device='cuda:0'), out_proj_covar=tensor([1.0482e-04, 4.3474e-05, 5.5018e-05, 3.4464e-05, 3.4132e-05, 4.6749e-05, 5.6307e-05, 3.8743e-05], device='cuda:0') 2023-04-27 13:20:44,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8879, 3.9713, 3.4451, 2.4396, 3.4290, 3.9030, 3.7235, 3.8180], device='cuda:0'), covar=tensor([0.0136, 0.0109, 0.0208, 0.1064, 0.0290, 0.0112, 0.0138, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0037, 0.0037, 0.0046, 0.0063, 0.0039, 0.0038, 0.0038, 0.0040], device='cuda:0'), out_proj_covar=tensor([3.9295e-05, 4.0496e-05, 4.9842e-05, 6.5789e-05, 4.7623e-05, 3.9638e-05, 4.6777e-05, 4.2081e-05], device='cuda:0') 2023-04-27 13:21:00,204 INFO [train.py:904] (0/8) Epoch 1, batch 3150, loss[loss=0.3573, simple_loss=0.3979, pruned_loss=0.1584, over 17170.00 frames. ], tot_loss[loss=0.3578, simple_loss=0.3987, pruned_loss=0.1584, over 3330092.54 frames. ], batch size: 46, lr: 4.60e-02, grad_scale: 16.0 2023-04-27 13:21:38,598 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3180.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:22:05,305 INFO [train.py:904] (0/8) Epoch 1, batch 3200, loss[loss=0.3155, simple_loss=0.3702, pruned_loss=0.1304, over 17223.00 frames. ], tot_loss[loss=0.3564, simple_loss=0.3977, pruned_loss=0.1575, over 3325548.70 frames. ], batch size: 44, lr: 4.59e-02, grad_scale: 16.0 2023-04-27 13:22:17,348 INFO [optim.py:368] (0/8) 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 13:23:07,328 INFO [train.py:904] (0/8) Epoch 1, batch 3250, loss[loss=0.3374, simple_loss=0.3853, pruned_loss=0.1447, over 17075.00 frames. ], tot_loss[loss=0.3548, simple_loss=0.3972, pruned_loss=0.1562, over 3323998.04 frames. ], batch size: 53, lr: 4.58e-02, grad_scale: 16.0 2023-04-27 13:23:56,306 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3288.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:24:01,847 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3293.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:24:12,262 INFO [train.py:904] (0/8) Epoch 1, batch 3300, loss[loss=0.331, simple_loss=0.3778, pruned_loss=0.1421, over 16904.00 frames. ], tot_loss[loss=0.3546, simple_loss=0.3975, pruned_loss=0.1559, over 3319569.88 frames. ], batch size: 96, lr: 4.57e-02, grad_scale: 16.0 2023-04-27 13:24:25,126 INFO [optim.py:368] (0/8) 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 13:24:44,179 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2977, 2.1743, 2.3455, 2.2452, 2.2691, 2.3741, 1.9577, 2.3179], device='cuda:0'), covar=tensor([0.0167, 0.0170, 0.0146, 0.0173, 0.0167, 0.0085, 0.0241, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0027, 0.0033, 0.0030, 0.0033, 0.0026, 0.0029, 0.0032, 0.0033], device='cuda:0'), out_proj_covar=tensor([2.5066e-05, 2.8151e-05, 2.7624e-05, 2.9406e-05, 2.2607e-05, 2.3308e-05, 2.7161e-05, 2.7758e-05], device='cuda:0') 2023-04-27 13:25:03,556 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3341.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:25:16,158 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3349.0, num_to_drop=2, layers_to_drop={0, 1} 2023-04-27 13:25:17,767 INFO [train.py:904] (0/8) Epoch 1, batch 3350, loss[loss=0.4188, simple_loss=0.4358, pruned_loss=0.2009, over 12166.00 frames. ], tot_loss[loss=0.3527, simple_loss=0.3969, pruned_loss=0.1542, over 3318305.69 frames. ], batch size: 246, lr: 4.56e-02, grad_scale: 16.0 2023-04-27 13:25:49,836 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3375.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:26:09,264 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:26:24,821 INFO [train.py:904] (0/8) Epoch 1, batch 3400, loss[loss=0.2969, simple_loss=0.3451, pruned_loss=0.1243, over 16776.00 frames. ], tot_loss[loss=0.3515, simple_loss=0.3961, pruned_loss=0.1534, over 3315212.85 frames. ], batch size: 39, lr: 4.55e-02, grad_scale: 16.0 2023-04-27 13:26:39,225 INFO [optim.py:368] (0/8) 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 13:27:21,055 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2347, 4.9431, 5.0164, 5.1280, 4.3818, 5.0867, 5.0643, 4.6169], device='cuda:0'), covar=tensor([0.0271, 0.0156, 0.0223, 0.0127, 0.1028, 0.0234, 0.0165, 0.0262], device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0076, 0.0126, 0.0091, 0.0155, 0.0102, 0.0092, 0.0099], device='cuda:0'), out_proj_covar=tensor([1.3987e-04, 9.9941e-05, 1.7299e-04, 1.1784e-04, 2.0902e-04, 1.4735e-04, 1.3103e-04, 1.4768e-04], device='cuda:0') 2023-04-27 13:27:30,726 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7639, 4.9217, 4.6156, 4.7123, 4.7462, 5.0587, 5.1050, 4.6375], device='cuda:0'), covar=tensor([0.0568, 0.0799, 0.0837, 0.1283, 0.1534, 0.0717, 0.0639, 0.1429], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0186, 0.0152, 0.0163, 0.0195, 0.0140, 0.0131, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 13:27:32,309 INFO [train.py:904] (0/8) Epoch 1, batch 3450, loss[loss=0.3094, simple_loss=0.3512, pruned_loss=0.1338, over 16772.00 frames. ], tot_loss[loss=0.3471, simple_loss=0.3925, pruned_loss=0.1508, over 3317818.14 frames. ], batch size: 124, lr: 4.54e-02, grad_scale: 16.0 2023-04-27 13:28:11,904 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3480.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:28:39,720 INFO [train.py:904] (0/8) Epoch 1, batch 3500, loss[loss=0.2957, simple_loss=0.3525, pruned_loss=0.1194, over 17015.00 frames. ], tot_loss[loss=0.3454, simple_loss=0.3909, pruned_loss=0.15, over 3299880.27 frames. ], batch size: 41, lr: 4.53e-02, grad_scale: 16.0 2023-04-27 13:28:53,818 INFO [optim.py:368] (0/8) 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 13:28:54,395 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4270, 4.7543, 4.3754, 5.0413, 4.9144, 4.6731, 3.7790, 4.7134], device='cuda:0'), covar=tensor([0.1257, 0.0071, 0.0253, 0.0045, 0.0046, 0.0148, 0.0280, 0.0078], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0047, 0.0067, 0.0034, 0.0032, 0.0046, 0.0061, 0.0044], device='cuda:0'), out_proj_covar=tensor([1.1413e-04, 4.6848e-05, 7.5055e-05, 4.0500e-05, 3.9814e-05, 5.4938e-05, 6.5842e-05, 4.5417e-05], device='cuda:0') 2023-04-27 13:29:16,289 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:29:16,456 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3069, 3.6830, 3.4462, 3.3805, 3.3684, 3.8161, 3.7872, 3.6899], device='cuda:0'), covar=tensor([0.0283, 0.0181, 0.0203, 0.0179, 0.0249, 0.0139, 0.0185, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0047, 0.0034, 0.0034, 0.0044, 0.0036, 0.0040, 0.0045, 0.0040], device='cuda:0'), out_proj_covar=tensor([6.3046e-05, 4.7780e-05, 4.6862e-05, 5.6791e-05, 4.9678e-05, 6.3945e-05, 5.9568e-05, 5.2666e-05], device='cuda:0') 2023-04-27 13:29:40,058 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3546.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:29:46,557 INFO [train.py:904] (0/8) Epoch 1, batch 3550, loss[loss=0.3806, simple_loss=0.4302, pruned_loss=0.1654, over 16601.00 frames. ], tot_loss[loss=0.3442, simple_loss=0.3896, pruned_loss=0.1494, over 3306920.01 frames. ], batch size: 62, lr: 4.51e-02, grad_scale: 16.0 2023-04-27 13:30:44,954 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:30:54,499 INFO [train.py:904] (0/8) Epoch 1, batch 3600, loss[loss=0.3156, simple_loss=0.383, pruned_loss=0.1241, over 17304.00 frames. ], tot_loss[loss=0.3415, simple_loss=0.3869, pruned_loss=0.148, over 3298906.96 frames. ], batch size: 52, lr: 4.50e-02, grad_scale: 16.0 2023-04-27 13:31:04,180 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3607.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:31:08,706 INFO [optim.py:368] (0/8) 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 13:31:37,279 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.49 vs. limit=2.0 2023-04-27 13:31:51,886 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3641.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:31:56,481 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3644.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:32:06,065 INFO [train.py:904] (0/8) Epoch 1, batch 3650, loss[loss=0.348, simple_loss=0.3796, pruned_loss=0.1582, over 11631.00 frames. ], tot_loss[loss=0.3391, simple_loss=0.3845, pruned_loss=0.1469, over 3291663.25 frames. ], batch size: 246, lr: 4.49e-02, grad_scale: 16.0 2023-04-27 13:32:41,804 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3675.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:32:43,501 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3676.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:33:02,936 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-27 13:33:08,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2441, 4.2273, 4.4977, 4.5466, 4.7040, 4.2611, 4.3302, 4.4922], device='cuda:0'), covar=tensor([0.0282, 0.0232, 0.0425, 0.0379, 0.0296, 0.0286, 0.0417, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0097, 0.0125, 0.0125, 0.0131, 0.0107, 0.0120, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:0') 2023-04-27 13:33:20,388 INFO [train.py:904] (0/8) Epoch 1, batch 3700, loss[loss=0.393, simple_loss=0.418, pruned_loss=0.184, over 11402.00 frames. ], tot_loss[loss=0.337, simple_loss=0.3808, pruned_loss=0.1466, over 3269037.01 frames. ], batch size: 248, lr: 4.48e-02, grad_scale: 16.0 2023-04-27 13:33:22,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0170, 3.2426, 2.8954, 1.8507, 2.8092, 3.1654, 2.9457, 3.1444], device='cuda:0'), covar=tensor([0.0256, 0.0155, 0.0310, 0.1540, 0.0421, 0.0180, 0.0203, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0046, 0.0047, 0.0051, 0.0092, 0.0045, 0.0047, 0.0043, 0.0053], device='cuda:0'), out_proj_covar=tensor([5.4019e-05, 5.6088e-05, 6.3284e-05, 1.1051e-04, 6.2110e-05, 5.8468e-05, 6.0856e-05, 5.9938e-05], device='cuda:0') 2023-04-27 13:33:35,078 INFO [optim.py:368] (0/8) 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 13:33:54,066 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3723.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:34:15,152 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3737.0, num_to_drop=2, layers_to_drop={2, 3} 2023-04-27 13:34:33,713 INFO [train.py:904] (0/8) Epoch 1, batch 3750, loss[loss=0.4433, simple_loss=0.4492, pruned_loss=0.2187, over 11816.00 frames. ], tot_loss[loss=0.3359, simple_loss=0.3786, pruned_loss=0.1466, over 3250681.96 frames. ], batch size: 248, lr: 4.47e-02, grad_scale: 16.0 2023-04-27 13:34:42,205 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 13:34:54,727 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0476, 2.3614, 2.6694, 2.4043, 2.3712, 2.3060, 2.4929, 2.3862], device='cuda:0'), covar=tensor([0.0357, 0.0286, 0.0243, 0.0256, 0.0178, 0.0244, 0.0244, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0024, 0.0035, 0.0027, 0.0029, 0.0023, 0.0028, 0.0027, 0.0031], device='cuda:0'), out_proj_covar=tensor([2.5522e-05, 3.3762e-05, 2.7059e-05, 2.8267e-05, 2.1302e-05, 2.3271e-05, 2.4342e-05, 2.7481e-05], device='cuda:0') 2023-04-27 13:35:45,605 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=2.03 vs. limit=2.0 2023-04-27 13:35:46,294 INFO [train.py:904] (0/8) Epoch 1, batch 3800, loss[loss=0.3666, simple_loss=0.3951, pruned_loss=0.1691, over 16332.00 frames. ], tot_loss[loss=0.3365, simple_loss=0.3784, pruned_loss=0.1473, over 3263474.57 frames. ], batch size: 165, lr: 4.46e-02, grad_scale: 16.0 2023-04-27 13:36:00,505 INFO [optim.py:368] (0/8) 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 13:36:57,577 INFO [train.py:904] (0/8) Epoch 1, batch 3850, loss[loss=0.3034, simple_loss=0.3463, pruned_loss=0.1303, over 16844.00 frames. ], tot_loss[loss=0.3351, simple_loss=0.3771, pruned_loss=0.1465, over 3248959.72 frames. ], batch size: 96, lr: 4.45e-02, grad_scale: 16.0 2023-04-27 13:36:59,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8491, 5.1074, 4.8113, 5.0512, 4.5521, 4.9668, 4.7576, 5.1989], device='cuda:0'), covar=tensor([0.0382, 0.0512, 0.0462, 0.0301, 0.0575, 0.0324, 0.0400, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0172, 0.0143, 0.0111, 0.0145, 0.0126, 0.0150, 0.0104], device='cuda:0'), out_proj_covar=tensor([1.3931e-04, 1.6595e-04, 1.2764e-04, 9.7974e-05, 1.3093e-04, 1.1348e-04, 1.4501e-04, 1.0248e-04], device='cuda:0') 2023-04-27 13:38:09,807 INFO [train.py:904] (0/8) Epoch 1, batch 3900, loss[loss=0.3264, simple_loss=0.3602, pruned_loss=0.1463, over 16877.00 frames. ], tot_loss[loss=0.3306, simple_loss=0.3734, pruned_loss=0.144, over 3249943.28 frames. ], batch size: 116, lr: 4.44e-02, grad_scale: 16.0 2023-04-27 13:38:11,820 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=3902.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:38:12,010 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0930, 2.5740, 2.6193, 3.3391, 3.3124, 3.1439, 2.4064, 3.0346], device='cuda:0'), covar=tensor([0.1728, 0.0245, 0.0732, 0.0083, 0.0134, 0.0218, 0.0406, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0060, 0.0097, 0.0040, 0.0039, 0.0054, 0.0079, 0.0056], device='cuda:0'), out_proj_covar=tensor([1.4948e-04, 6.3826e-05, 1.1145e-04, 4.9819e-05, 5.3620e-05, 7.0259e-05, 8.8592e-05, 6.1595e-05], device='cuda:0') 2023-04-27 13:38:24,739 INFO [optim.py:368] (0/8) 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 13:38:46,476 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=3926.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:39:11,940 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=3944.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:39:21,380 INFO [train.py:904] (0/8) Epoch 1, batch 3950, loss[loss=0.3169, simple_loss=0.3514, pruned_loss=0.1412, over 16856.00 frames. ], tot_loss[loss=0.3293, simple_loss=0.3712, pruned_loss=0.1437, over 3252312.08 frames. ], batch size: 116, lr: 4.43e-02, grad_scale: 16.0 2023-04-27 13:39:33,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3036, 4.0592, 4.2073, 4.3667, 3.7351, 4.2463, 4.0760, 3.8438], device='cuda:0'), covar=tensor([0.0240, 0.0213, 0.0216, 0.0121, 0.0825, 0.0220, 0.0298, 0.0239], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0069, 0.0114, 0.0085, 0.0140, 0.0094, 0.0083, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001], device='cuda:0') 2023-04-27 13:40:13,460 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=3987.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:40:20,469 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=3992.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:40:33,403 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-4000.pt 2023-04-27 13:40:37,512 INFO [train.py:904] (0/8) Epoch 1, batch 4000, loss[loss=0.3325, simple_loss=0.3772, pruned_loss=0.1438, over 15585.00 frames. ], tot_loss[loss=0.3279, simple_loss=0.3696, pruned_loss=0.1431, over 3254973.18 frames. ], batch size: 191, lr: 4.42e-02, grad_scale: 16.0 2023-04-27 13:40:52,177 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.081e+02 4.378e+02 5.733e+02 7.371e+02 1.613e+03, threshold=1.147e+03, percent-clipped=6.0 2023-04-27 13:41:24,008 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4032.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:41:50,491 INFO [train.py:904] (0/8) Epoch 1, batch 4050, loss[loss=0.26, simple_loss=0.3242, pruned_loss=0.09788, over 16724.00 frames. ], tot_loss[loss=0.3194, simple_loss=0.3645, pruned_loss=0.1371, over 3256137.33 frames. ], batch size: 124, lr: 4.41e-02, grad_scale: 16.0 2023-04-27 13:41:54,759 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 13:43:04,582 INFO [train.py:904] (0/8) Epoch 1, batch 4100, loss[loss=0.3176, simple_loss=0.3705, pruned_loss=0.1324, over 16679.00 frames. ], tot_loss[loss=0.314, simple_loss=0.3622, pruned_loss=0.1328, over 3256499.46 frames. ], batch size: 134, lr: 4.40e-02, grad_scale: 32.0 2023-04-27 13:43:18,966 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.292e+02 4.086e+02 5.427e+02 7.425e+02 1.337e+03, threshold=1.085e+03, percent-clipped=4.0 2023-04-27 13:44:04,961 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4141.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:44:20,307 INFO [train.py:904] (0/8) Epoch 1, batch 4150, loss[loss=0.2994, simple_loss=0.3634, pruned_loss=0.1177, over 16836.00 frames. ], tot_loss[loss=0.3256, simple_loss=0.3729, pruned_loss=0.1391, over 3225368.90 frames. ], batch size: 90, lr: 4.39e-02, grad_scale: 32.0 2023-04-27 13:45:37,087 INFO [train.py:904] (0/8) Epoch 1, batch 4200, loss[loss=0.3974, simple_loss=0.4263, pruned_loss=0.1842, over 11566.00 frames. ], tot_loss[loss=0.3342, simple_loss=0.3824, pruned_loss=0.143, over 3192396.69 frames. ], batch size: 248, lr: 4.38e-02, grad_scale: 16.0 2023-04-27 13:45:39,564 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4202.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:45:39,720 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:45:50,663 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5096, 2.0059, 2.2378, 2.4130, 2.8121, 2.5083, 2.7291, 2.7089], device='cuda:0'), covar=tensor([0.0079, 0.0522, 0.0210, 0.0135, 0.0079, 0.0167, 0.0097, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0027, 0.0049, 0.0037, 0.0033, 0.0030, 0.0033, 0.0035, 0.0031], device='cuda:0'), out_proj_covar=tensor([3.8270e-05, 7.7899e-05, 5.7463e-05, 4.2016e-05, 4.1067e-05, 4.6987e-05, 4.5006e-05, 4.0901e-05], device='cuda:0') 2023-04-27 13:45:52,581 INFO [optim.py:368] (0/8) 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 13:46:13,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4290, 3.5000, 3.1466, 2.6889, 3.3076, 3.1092, 3.2396, 3.1517], device='cuda:0'), covar=tensor([0.0499, 0.0076, 0.0132, 0.0179, 0.0136, 0.0129, 0.0169, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0053, 0.0028, 0.0029, 0.0040, 0.0031, 0.0031, 0.0037, 0.0035], device='cuda:0'), out_proj_covar=tensor([8.1221e-05, 4.6392e-05, 4.5310e-05, 5.8887e-05, 4.8651e-05, 5.3496e-05, 5.6270e-05, 5.4437e-05], device='cuda:0') 2023-04-27 13:46:13,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3151, 1.7645, 1.6728, 2.3300, 2.0081, 2.1552, 2.4856, 1.8294], device='cuda:0'), covar=tensor([0.0061, 0.0300, 0.0211, 0.0129, 0.0076, 0.0084, 0.0090, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0022, 0.0038, 0.0028, 0.0027, 0.0023, 0.0026, 0.0027, 0.0030], device='cuda:0'), out_proj_covar=tensor([2.3713e-05, 4.1483e-05, 2.9198e-05, 2.8500e-05, 2.0616e-05, 2.2997e-05, 2.6240e-05, 2.7746e-05], device='cuda:0') 2023-04-27 13:46:49,602 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4250.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 13:46:50,872 INFO [train.py:904] (0/8) Epoch 1, batch 4250, loss[loss=0.3486, simple_loss=0.4036, pruned_loss=0.1468, over 16414.00 frames. ], tot_loss[loss=0.3345, simple_loss=0.384, pruned_loss=0.1425, over 3171215.99 frames. ], batch size: 146, lr: 4.36e-02, grad_scale: 16.0 2023-04-27 13:47:19,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4886, 4.5527, 4.7088, 4.7779, 5.0919, 4.6074, 4.6623, 4.7515], device='cuda:0'), covar=tensor([0.0231, 0.0185, 0.0399, 0.0387, 0.0230, 0.0200, 0.0363, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0090, 0.0112, 0.0110, 0.0120, 0.0101, 0.0114, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:0') 2023-04-27 13:47:36,657 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4282.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:45,255 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=4288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:47:52,583 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.97 vs. limit=2.0 2023-04-27 13:47:59,086 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7835, 4.5604, 4.2066, 4.0507, 4.6589, 4.4680, 4.2277, 4.5537], device='cuda:0'), covar=tensor([0.0095, 0.0133, 0.0103, 0.0324, 0.0070, 0.0136, 0.0102, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0046, 0.0038, 0.0055, 0.0065, 0.0040, 0.0049, 0.0053, 0.0051], device='cuda:0'), out_proj_covar=tensor([7.9518e-05, 6.4211e-05, 1.0233e-04, 1.0859e-04, 6.0782e-05, 8.2056e-05, 9.6826e-05, 9.5816e-05], device='cuda:0') 2023-04-27 13:48:04,737 INFO [train.py:904] (0/8) Epoch 1, batch 4300, loss[loss=0.3309, simple_loss=0.3944, pruned_loss=0.1337, over 17027.00 frames. ], tot_loss[loss=0.3325, simple_loss=0.3843, pruned_loss=0.1404, over 3165064.24 frames. ], batch size: 50, lr: 4.35e-02, grad_scale: 16.0 2023-04-27 13:48:21,413 INFO [optim.py:368] (0/8) 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 13:48:22,500 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 13:48:51,352 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.08 vs. limit=5.0 2023-04-27 13:48:52,593 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4332.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 13:49:03,994 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-04-27 13:49:17,766 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=4349.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:49:20,287 INFO [train.py:904] (0/8) Epoch 1, batch 4350, loss[loss=0.3123, simple_loss=0.3824, pruned_loss=0.1211, over 16551.00 frames. ], tot_loss[loss=0.3352, simple_loss=0.3875, pruned_loss=0.1415, over 3159530.82 frames. ], batch size: 75, lr: 4.34e-02, grad_scale: 16.0 2023-04-27 13:49:35,162 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4772, 3.8899, 3.8660, 2.8667, 3.4832, 4.0064, 3.9397, 3.5643], device='cuda:0'), covar=tensor([0.0555, 0.0087, 0.0075, 0.0185, 0.0163, 0.0081, 0.0114, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0055, 0.0027, 0.0028, 0.0040, 0.0029, 0.0030, 0.0036, 0.0035], device='cuda:0'), out_proj_covar=tensor([8.5682e-05, 4.4550e-05, 4.4717e-05, 6.0844e-05, 4.8676e-05, 5.2508e-05, 5.5365e-05, 5.4587e-05], device='cuda:0') 2023-04-27 13:49:42,114 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9992, 4.3619, 4.1014, 4.6444, 3.6604, 3.9363, 3.2218, 3.5664], device='cuda:0'), covar=tensor([0.0320, 0.0304, 0.0300, 0.0212, 0.1175, 0.0295, 0.0652, 0.0962], device='cuda:0'), in_proj_covar=tensor([0.0054, 0.0061, 0.0051, 0.0050, 0.0103, 0.0057, 0.0074, 0.0060], device='cuda:0'), out_proj_covar=tensor([5.8224e-05, 6.3213e-05, 5.2604e-05, 5.7823e-05, 1.1139e-04, 6.1361e-05, 7.4056e-05, 7.2341e-05], device='cuda:0') 2023-04-27 13:49:42,356 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.99 vs. limit=2.0 2023-04-27 13:49:45,277 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8014, 4.7267, 4.2261, 3.9776, 4.8162, 4.5922, 4.3061, 4.5654], device='cuda:0'), covar=tensor([0.0093, 0.0090, 0.0077, 0.0317, 0.0040, 0.0099, 0.0070, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0045, 0.0037, 0.0054, 0.0066, 0.0039, 0.0049, 0.0053, 0.0050], device='cuda:0'), out_proj_covar=tensor([7.8493e-05, 6.3611e-05, 1.0166e-04, 1.1302e-04, 5.9456e-05, 8.3853e-05, 9.5271e-05, 9.5457e-05], device='cuda:0') 2023-04-27 13:50:04,951 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4380.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 13:50:36,415 INFO [train.py:904] (0/8) Epoch 1, batch 4400, loss[loss=0.3099, simple_loss=0.3814, pruned_loss=0.1192, over 16867.00 frames. ], tot_loss[loss=0.3363, simple_loss=0.389, pruned_loss=0.1418, over 3166278.59 frames. ], batch size: 42, lr: 4.33e-02, grad_scale: 16.0 2023-04-27 13:50:51,964 INFO [optim.py:368] (0/8) 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 13:51:15,205 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-27 13:51:42,906 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7933, 3.9273, 1.9753, 4.0150, 3.6089, 3.5730, 2.1499, 3.2906], device='cuda:0'), covar=tensor([0.0081, 0.0104, 0.1005, 0.0071, 0.0194, 0.0098, 0.0859, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0036, 0.0034, 0.0066, 0.0037, 0.0053, 0.0033, 0.0071, 0.0041], device='cuda:0'), out_proj_covar=tensor([4.3491e-05, 4.5090e-05, 9.0856e-05, 4.3098e-05, 6.4491e-05, 4.6520e-05, 9.0439e-05, 5.4258e-05], device='cuda:0') 2023-04-27 13:51:48,410 INFO [train.py:904] (0/8) Epoch 1, batch 4450, loss[loss=0.3004, simple_loss=0.3738, pruned_loss=0.1135, over 16522.00 frames. ], tot_loss[loss=0.3356, simple_loss=0.3903, pruned_loss=0.1404, over 3179955.21 frames. ], batch size: 68, lr: 4.32e-02, grad_scale: 16.0 2023-04-27 13:52:32,822 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7127, 4.7430, 4.2932, 4.0319, 4.7202, 4.5925, 4.3476, 4.4325], device='cuda:0'), covar=tensor([0.0079, 0.0039, 0.0067, 0.0242, 0.0031, 0.0078, 0.0057, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0045, 0.0036, 0.0054, 0.0066, 0.0039, 0.0049, 0.0052, 0.0050], device='cuda:0'), out_proj_covar=tensor([8.0892e-05, 6.2470e-05, 1.0504e-04, 1.1430e-04, 6.1751e-05, 8.6260e-05, 9.6017e-05, 9.7932e-05], device='cuda:0') 2023-04-27 13:52:57,215 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4497.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:53:02,594 INFO [train.py:904] (0/8) Epoch 1, batch 4500, loss[loss=0.3251, simple_loss=0.3773, pruned_loss=0.1365, over 16233.00 frames. ], tot_loss[loss=0.3319, simple_loss=0.3881, pruned_loss=0.1379, over 3200525.91 frames. ], batch size: 165, lr: 4.31e-02, grad_scale: 8.0 2023-04-27 13:53:16,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8652, 4.7366, 4.1560, 3.8713, 4.7445, 4.5007, 4.2293, 4.4245], device='cuda:0'), covar=tensor([0.0067, 0.0052, 0.0104, 0.0397, 0.0036, 0.0125, 0.0082, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0046, 0.0037, 0.0056, 0.0069, 0.0040, 0.0051, 0.0054, 0.0051], device='cuda:0'), out_proj_covar=tensor([8.3249e-05, 6.3884e-05, 1.0829e-04, 1.2070e-04, 6.3137e-05, 9.0540e-05, 1.0008e-04, 1.0124e-04], device='cuda:0') 2023-04-27 13:53:20,029 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.477e+02 4.034e+02 4.961e+02 6.637e+02 1.457e+03, threshold=9.923e+02, percent-clipped=1.0 2023-04-27 13:53:30,039 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 13:54:14,102 INFO [train.py:904] (0/8) Epoch 1, batch 4550, loss[loss=0.3617, simple_loss=0.4146, pruned_loss=0.1544, over 16542.00 frames. ], tot_loss[loss=0.3295, simple_loss=0.3868, pruned_loss=0.1361, over 3218170.81 frames. ], batch size: 75, lr: 4.30e-02, grad_scale: 8.0 2023-04-27 13:54:16,339 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6921, 5.5515, 5.3624, 5.3773, 5.4974, 5.8912, 5.6372, 5.1480], device='cuda:0'), covar=tensor([0.0545, 0.0659, 0.0540, 0.0937, 0.1312, 0.0371, 0.0524, 0.1227], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0180, 0.0141, 0.0152, 0.0188, 0.0132, 0.0139, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 13:54:29,069 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 13:54:58,071 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4582.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:55:07,400 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 13:55:25,716 INFO [train.py:904] (0/8) Epoch 1, batch 4600, loss[loss=0.2913, simple_loss=0.3596, pruned_loss=0.1115, over 16377.00 frames. ], tot_loss[loss=0.3285, simple_loss=0.3867, pruned_loss=0.1352, over 3213943.11 frames. ], batch size: 35, lr: 4.29e-02, grad_scale: 8.0 2023-04-27 13:55:43,317 INFO [optim.py:368] (0/8) 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 13:56:07,431 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:27,604 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=4644.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 13:56:36,574 INFO [train.py:904] (0/8) Epoch 1, batch 4650, loss[loss=0.3208, simple_loss=0.3802, pruned_loss=0.1307, over 15383.00 frames. ], tot_loss[loss=0.3245, simple_loss=0.3834, pruned_loss=0.1328, over 3217568.90 frames. ], batch size: 190, lr: 4.28e-02, grad_scale: 8.0 2023-04-27 13:57:50,186 INFO [train.py:904] (0/8) Epoch 1, batch 4700, loss[loss=0.3071, simple_loss=0.3702, pruned_loss=0.122, over 16286.00 frames. ], tot_loss[loss=0.3213, simple_loss=0.3802, pruned_loss=0.1312, over 3210548.92 frames. ], batch size: 35, lr: 4.27e-02, grad_scale: 8.0 2023-04-27 13:58:07,987 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.460e+02 4.210e+02 5.632e+02 6.697e+02 1.082e+03, threshold=1.126e+03, percent-clipped=2.0 2023-04-27 13:59:02,508 INFO [train.py:904] (0/8) Epoch 1, batch 4750, loss[loss=0.3036, simple_loss=0.3572, pruned_loss=0.125, over 16618.00 frames. ], tot_loss[loss=0.3182, simple_loss=0.3768, pruned_loss=0.1298, over 3205916.29 frames. ], batch size: 62, lr: 4.26e-02, grad_scale: 8.0 2023-04-27 13:59:19,967 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7551, 3.8762, 3.7511, 1.7442, 3.4831, 3.7470, 3.4906, 3.4341], device='cuda:0'), covar=tensor([0.0103, 0.0118, 0.0162, 0.2072, 0.0194, 0.0202, 0.0130, 0.0261], device='cuda:0'), in_proj_covar=tensor([0.0056, 0.0060, 0.0057, 0.0126, 0.0054, 0.0056, 0.0053, 0.0069], device='cuda:0'), out_proj_covar=tensor([7.2635e-05, 7.9667e-05, 8.1622e-05, 1.6925e-04, 7.9684e-05, 7.6430e-05, 8.0902e-05, 8.9797e-05], device='cuda:0') 2023-04-27 14:00:11,419 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4797.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:00:16,572 INFO [train.py:904] (0/8) Epoch 1, batch 4800, loss[loss=0.3214, simple_loss=0.3845, pruned_loss=0.1292, over 16673.00 frames. ], tot_loss[loss=0.3141, simple_loss=0.3732, pruned_loss=0.1275, over 3196279.41 frames. ], batch size: 134, lr: 4.25e-02, grad_scale: 8.0 2023-04-27 14:00:34,036 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.251e+02 4.322e+02 5.200e+02 6.651e+02 1.076e+03, threshold=1.040e+03, percent-clipped=0.0 2023-04-27 14:01:22,952 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:01:31,772 INFO [train.py:904] (0/8) Epoch 1, batch 4850, loss[loss=0.3298, simple_loss=0.3922, pruned_loss=0.1337, over 15431.00 frames. ], tot_loss[loss=0.3158, simple_loss=0.3755, pruned_loss=0.128, over 3190831.02 frames. ], batch size: 190, lr: 4.24e-02, grad_scale: 8.0 2023-04-27 14:02:21,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9138, 4.0640, 3.8780, 3.9763, 3.6406, 3.9987, 3.9102, 4.0502], device='cuda:0'), covar=tensor([0.0350, 0.0587, 0.0504, 0.0345, 0.0533, 0.0377, 0.0385, 0.0410], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0164, 0.0146, 0.0106, 0.0134, 0.0116, 0.0144, 0.0099], device='cuda:0'), out_proj_covar=tensor([1.3780e-04, 1.6399e-04, 1.3145e-04, 9.8607e-05, 1.2760e-04, 1.1031e-04, 1.4621e-04, 1.0272e-04], device='cuda:0') 2023-04-27 14:02:49,102 INFO [train.py:904] (0/8) Epoch 1, batch 4900, loss[loss=0.3311, simple_loss=0.3866, pruned_loss=0.1378, over 15478.00 frames. ], tot_loss[loss=0.3145, simple_loss=0.375, pruned_loss=0.127, over 3167999.15 frames. ], batch size: 190, lr: 4.23e-02, grad_scale: 8.0 2023-04-27 14:03:07,684 INFO [optim.py:368] (0/8) 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 14:03:54,933 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=4944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:04:04,717 INFO [train.py:904] (0/8) Epoch 1, batch 4950, loss[loss=0.2982, simple_loss=0.3735, pruned_loss=0.1115, over 16914.00 frames. ], tot_loss[loss=0.3156, simple_loss=0.3757, pruned_loss=0.1277, over 3171847.00 frames. ], batch size: 96, lr: 4.21e-02, grad_scale: 8.0 2023-04-27 14:04:38,863 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9402, 4.3912, 5.0110, 5.0961, 4.1832, 4.7829, 4.6955, 4.6671], device='cuda:0'), covar=tensor([0.0306, 0.0215, 0.0152, 0.0091, 0.0828, 0.0270, 0.0116, 0.0161], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0060, 0.0108, 0.0081, 0.0135, 0.0087, 0.0074, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 14:05:04,827 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=4992.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:05:17,369 INFO [train.py:904] (0/8) Epoch 1, batch 5000, loss[loss=0.316, simple_loss=0.3853, pruned_loss=0.1233, over 16705.00 frames. ], tot_loss[loss=0.3153, simple_loss=0.3766, pruned_loss=0.127, over 3184635.69 frames. ], batch size: 134, lr: 4.20e-02, grad_scale: 8.0 2023-04-27 14:05:35,392 INFO [optim.py:368] (0/8) 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 14:06:31,107 INFO [train.py:904] (0/8) Epoch 1, batch 5050, loss[loss=0.2938, simple_loss=0.355, pruned_loss=0.1163, over 16383.00 frames. ], tot_loss[loss=0.3145, simple_loss=0.3761, pruned_loss=0.1265, over 3188268.02 frames. ], batch size: 35, lr: 4.19e-02, grad_scale: 8.0 2023-04-27 14:07:42,605 INFO [train.py:904] (0/8) Epoch 1, batch 5100, loss[loss=0.3039, simple_loss=0.3629, pruned_loss=0.1225, over 15493.00 frames. ], tot_loss[loss=0.311, simple_loss=0.3732, pruned_loss=0.1244, over 3199808.84 frames. ], batch size: 190, lr: 4.18e-02, grad_scale: 8.0 2023-04-27 14:07:57,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9260, 3.8156, 3.7939, 4.2424, 4.1901, 4.2353, 4.1632, 4.1253], device='cuda:0'), covar=tensor([0.0284, 0.0249, 0.0846, 0.0261, 0.0270, 0.0203, 0.0233, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0137, 0.0231, 0.0161, 0.0135, 0.0136, 0.0122, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 14:07:59,830 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.617e+02 4.246e+02 5.535e+02 6.678e+02 1.197e+03, threshold=1.107e+03, percent-clipped=1.0 2023-04-27 14:08:58,114 INFO [train.py:904] (0/8) Epoch 1, batch 5150, loss[loss=0.2718, simple_loss=0.3579, pruned_loss=0.09282, over 16900.00 frames. ], tot_loss[loss=0.3101, simple_loss=0.3734, pruned_loss=0.1234, over 3197591.07 frames. ], batch size: 96, lr: 4.17e-02, grad_scale: 8.0 2023-04-27 14:09:22,623 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-27 14:10:12,917 INFO [train.py:904] (0/8) Epoch 1, batch 5200, loss[loss=0.3264, simple_loss=0.378, pruned_loss=0.1374, over 16360.00 frames. ], tot_loss[loss=0.3073, simple_loss=0.3706, pruned_loss=0.122, over 3215213.83 frames. ], batch size: 165, lr: 4.16e-02, grad_scale: 8.0 2023-04-27 14:10:30,246 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.582e+02 4.001e+02 4.794e+02 6.093e+02 1.086e+03, threshold=9.588e+02, percent-clipped=0.0 2023-04-27 14:11:26,089 INFO [train.py:904] (0/8) Epoch 1, batch 5250, loss[loss=0.2635, simple_loss=0.3274, pruned_loss=0.09982, over 17126.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3663, pruned_loss=0.1203, over 3222653.79 frames. ], batch size: 48, lr: 4.15e-02, grad_scale: 8.0 2023-04-27 14:11:34,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0182, 5.2980, 4.9359, 5.1479, 4.6644, 4.8131, 4.7638, 5.3368], device='cuda:0'), covar=tensor([0.0347, 0.0468, 0.0587, 0.0333, 0.0508, 0.0302, 0.0477, 0.0314], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0175, 0.0164, 0.0119, 0.0144, 0.0115, 0.0156, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-27 14:11:43,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6042, 4.7817, 4.3790, 2.1930, 3.5755, 2.3740, 3.9852, 4.8834], device='cuda:0'), covar=tensor([0.0203, 0.0209, 0.0233, 0.1983, 0.0814, 0.1378, 0.0764, 0.0119], device='cuda:0'), in_proj_covar=tensor([0.0075, 0.0056, 0.0091, 0.0136, 0.0126, 0.0122, 0.0109, 0.0052], device='cuda:0'), out_proj_covar=tensor([1.2136e-04, 9.5474e-05, 1.2528e-04, 1.7107e-04, 1.6916e-04, 1.5744e-04, 1.6259e-04, 8.6782e-05], device='cuda:0') 2023-04-27 14:12:37,170 INFO [train.py:904] (0/8) Epoch 1, batch 5300, loss[loss=0.2581, simple_loss=0.3228, pruned_loss=0.09664, over 16562.00 frames. ], tot_loss[loss=0.2987, simple_loss=0.3614, pruned_loss=0.118, over 3223246.38 frames. ], batch size: 57, lr: 4.14e-02, grad_scale: 8.0 2023-04-27 14:12:38,988 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 14:12:49,140 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8440, 3.8442, 3.5874, 3.3241, 2.7815, 2.3792, 4.0776, 4.3336], device='cuda:0'), covar=tensor([0.1790, 0.0529, 0.0748, 0.0467, 0.1622, 0.1257, 0.0246, 0.0052], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0120, 0.0153, 0.0090, 0.0142, 0.0125, 0.0090, 0.0050], device='cuda:0'), out_proj_covar=tensor([1.9267e-04, 1.4663e-04, 1.6300e-04, 1.0120e-04, 1.7361e-04, 1.4072e-04, 1.0598e-04, 6.1616e-05], device='cuda:0') 2023-04-27 14:12:54,714 INFO [optim.py:368] (0/8) 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 14:13:40,816 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-27 14:13:49,540 INFO [train.py:904] (0/8) Epoch 1, batch 5350, loss[loss=0.361, simple_loss=0.3983, pruned_loss=0.1619, over 11853.00 frames. ], tot_loss[loss=0.2965, simple_loss=0.3595, pruned_loss=0.1167, over 3222982.80 frames. ], batch size: 247, lr: 4.13e-02, grad_scale: 8.0 2023-04-27 14:14:20,553 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.38 vs. limit=5.0 2023-04-27 14:14:55,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6356, 3.5982, 3.8931, 3.9515, 4.1080, 3.6532, 3.7078, 4.0034], device='cuda:0'), covar=tensor([0.0277, 0.0314, 0.0562, 0.0397, 0.0346, 0.0308, 0.0541, 0.0229], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0105, 0.0136, 0.0129, 0.0141, 0.0113, 0.0137, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-27 14:15:00,997 INFO [train.py:904] (0/8) Epoch 1, batch 5400, loss[loss=0.3075, simple_loss=0.3711, pruned_loss=0.122, over 16514.00 frames. ], tot_loss[loss=0.3022, simple_loss=0.3644, pruned_loss=0.12, over 3200033.43 frames. ], batch size: 75, lr: 4.12e-02, grad_scale: 8.0 2023-04-27 14:15:09,295 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.50 vs. limit=5.0 2023-04-27 14:15:18,320 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.226e+02 4.427e+02 5.350e+02 6.399e+02 9.942e+02, threshold=1.070e+03, percent-clipped=0.0 2023-04-27 14:16:19,559 INFO [train.py:904] (0/8) Epoch 1, batch 5450, loss[loss=0.3785, simple_loss=0.4291, pruned_loss=0.1639, over 16862.00 frames. ], tot_loss[loss=0.3091, simple_loss=0.3696, pruned_loss=0.1243, over 3178047.60 frames. ], batch size: 102, lr: 4.11e-02, grad_scale: 8.0 2023-04-27 14:16:33,844 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6852, 3.8468, 3.4602, 2.7038, 3.6368, 3.7350, 3.6883, 3.0087], device='cuda:0'), covar=tensor([0.1162, 0.0106, 0.0146, 0.0275, 0.0129, 0.0144, 0.0202, 0.0227], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0035, 0.0038, 0.0051, 0.0036, 0.0038, 0.0043, 0.0048], device='cuda:0'), out_proj_covar=tensor([1.4791e-04, 6.5992e-05, 7.1769e-05, 8.8976e-05, 6.5980e-05, 7.6357e-05, 8.0790e-05, 8.8030e-05], device='cuda:0') 2023-04-27 14:16:59,628 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1482, 3.1919, 3.0253, 3.3958, 3.3261, 3.3865, 3.2580, 3.2754], device='cuda:0'), covar=tensor([0.0407, 0.0338, 0.1184, 0.0418, 0.0526, 0.0397, 0.0539, 0.0373], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0140, 0.0232, 0.0160, 0.0136, 0.0141, 0.0127, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 14:17:37,171 INFO [train.py:904] (0/8) Epoch 1, batch 5500, loss[loss=0.3763, simple_loss=0.4179, pruned_loss=0.1673, over 16429.00 frames. ], tot_loss[loss=0.3253, simple_loss=0.3812, pruned_loss=0.1347, over 3143465.56 frames. ], batch size: 68, lr: 4.10e-02, grad_scale: 8.0 2023-04-27 14:17:56,201 INFO [optim.py:368] (0/8) 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 14:18:57,395 INFO [train.py:904] (0/8) Epoch 1, batch 5550, loss[loss=0.5028, simple_loss=0.4829, pruned_loss=0.2614, over 11302.00 frames. ], tot_loss[loss=0.3411, simple_loss=0.3923, pruned_loss=0.145, over 3119175.29 frames. ], batch size: 248, lr: 4.09e-02, grad_scale: 8.0 2023-04-27 14:19:11,959 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-27 14:20:17,852 INFO [train.py:904] (0/8) Epoch 1, batch 5600, loss[loss=0.4466, simple_loss=0.445, pruned_loss=0.2241, over 11242.00 frames. ], tot_loss[loss=0.3513, simple_loss=0.3991, pruned_loss=0.1517, over 3102326.56 frames. ], batch size: 247, lr: 4.08e-02, grad_scale: 8.0 2023-04-27 14:20:37,630 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.538e+02 5.637e+02 6.821e+02 8.546e+02 1.933e+03, threshold=1.364e+03, percent-clipped=3.0 2023-04-27 14:21:39,922 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5650.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:21:40,606 INFO [train.py:904] (0/8) Epoch 1, batch 5650, loss[loss=0.391, simple_loss=0.4277, pruned_loss=0.1772, over 16375.00 frames. ], tot_loss[loss=0.3627, simple_loss=0.4069, pruned_loss=0.1592, over 3081786.93 frames. ], batch size: 146, lr: 4.07e-02, grad_scale: 8.0 2023-04-27 14:22:29,786 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3936, 3.1162, 3.0744, 3.3226, 2.9596, 3.1874, 3.1809, 3.0585], device='cuda:0'), covar=tensor([0.0201, 0.0178, 0.0206, 0.0146, 0.0678, 0.0172, 0.0536, 0.0230], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0062, 0.0110, 0.0085, 0.0136, 0.0086, 0.0076, 0.0089], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 14:22:59,486 INFO [train.py:904] (0/8) Epoch 1, batch 5700, loss[loss=0.3708, simple_loss=0.425, pruned_loss=0.1583, over 16874.00 frames. ], tot_loss[loss=0.3656, simple_loss=0.4091, pruned_loss=0.161, over 3088500.02 frames. ], batch size: 116, lr: 4.06e-02, grad_scale: 8.0 2023-04-27 14:23:15,497 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5711.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:23:17,968 INFO [optim.py:368] (0/8) 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 14:24:06,825 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1056, 3.7039, 4.0311, 4.1273, 3.5398, 4.0283, 3.9208, 3.7333], device='cuda:0'), covar=tensor([0.0244, 0.0183, 0.0147, 0.0100, 0.0755, 0.0163, 0.0270, 0.0194], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0060, 0.0106, 0.0082, 0.0133, 0.0082, 0.0074, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 14:24:21,195 INFO [train.py:904] (0/8) Epoch 1, batch 5750, loss[loss=0.359, simple_loss=0.4196, pruned_loss=0.1492, over 16456.00 frames. ], tot_loss[loss=0.3704, simple_loss=0.4132, pruned_loss=0.1638, over 3058079.89 frames. ], batch size: 75, lr: 4.05e-02, grad_scale: 8.0 2023-04-27 14:24:28,395 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5755.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:25:42,941 INFO [train.py:904] (0/8) Epoch 1, batch 5800, loss[loss=0.3139, simple_loss=0.3684, pruned_loss=0.1297, over 16633.00 frames. ], tot_loss[loss=0.3674, simple_loss=0.4122, pruned_loss=0.1613, over 3066143.24 frames. ], batch size: 57, lr: 4.04e-02, grad_scale: 8.0 2023-04-27 14:26:01,839 INFO [optim.py:368] (0/8) 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 14:26:07,002 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:26:36,019 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2333, 1.6519, 1.5014, 1.7629, 2.1499, 1.8618, 2.1854, 2.0260], device='cuda:0'), covar=tensor([0.0076, 0.0415, 0.0179, 0.0176, 0.0095, 0.0207, 0.0127, 0.0255], device='cuda:0'), in_proj_covar=tensor([0.0030, 0.0054, 0.0036, 0.0032, 0.0032, 0.0037, 0.0029, 0.0032], device='cuda:0'), out_proj_covar=tensor([3.3822e-05, 7.5288e-05, 4.2826e-05, 4.1167e-05, 3.6749e-05, 4.2478e-05, 3.6717e-05, 3.9339e-05], device='cuda:0') 2023-04-27 14:27:02,344 INFO [train.py:904] (0/8) Epoch 1, batch 5850, loss[loss=0.4229, simple_loss=0.4333, pruned_loss=0.2062, over 11624.00 frames. ], tot_loss[loss=0.3646, simple_loss=0.4101, pruned_loss=0.1595, over 3041900.11 frames. ], batch size: 247, lr: 4.03e-02, grad_scale: 8.0 2023-04-27 14:27:09,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0558, 4.2943, 4.1998, 4.1569, 4.1972, 4.6367, 4.5313, 4.2067], device='cuda:0'), covar=tensor([0.1136, 0.1100, 0.0884, 0.1640, 0.2137, 0.0648, 0.0771, 0.1816], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0187, 0.0154, 0.0165, 0.0204, 0.0150, 0.0149, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 14:27:14,313 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5858.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:28:25,820 INFO [train.py:904] (0/8) Epoch 1, batch 5900, loss[loss=0.3121, simple_loss=0.3781, pruned_loss=0.123, over 17034.00 frames. ], tot_loss[loss=0.3596, simple_loss=0.4072, pruned_loss=0.156, over 3064483.93 frames. ], batch size: 53, lr: 4.02e-02, grad_scale: 8.0 2023-04-27 14:28:48,066 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.490e+02 5.425e+02 6.564e+02 8.517e+02 1.690e+03, threshold=1.313e+03, percent-clipped=2.0 2023-04-27 14:28:58,977 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=5919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:29:24,704 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.95 vs. limit=2.0 2023-04-27 14:29:43,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8550, 3.4500, 1.9569, 3.8706, 3.9739, 3.9464, 2.5883, 3.5818], device='cuda:0'), covar=tensor([0.2803, 0.0288, 0.2271, 0.0104, 0.0102, 0.0225, 0.0938, 0.0290], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0080, 0.0150, 0.0051, 0.0050, 0.0064, 0.0112, 0.0088], device='cuda:0'), out_proj_covar=tensor([2.0012e-04, 1.1004e-04, 1.8792e-04, 7.9646e-05, 8.1633e-05, 1.1274e-04, 1.5219e-04, 1.2211e-04], device='cuda:0') 2023-04-27 14:29:49,236 INFO [train.py:904] (0/8) Epoch 1, batch 5950, loss[loss=0.41, simple_loss=0.4339, pruned_loss=0.193, over 11607.00 frames. ], tot_loss[loss=0.3574, simple_loss=0.4078, pruned_loss=0.1535, over 3074636.62 frames. ], batch size: 246, lr: 4.01e-02, grad_scale: 8.0 2023-04-27 14:30:42,688 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 14:30:52,043 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=5990.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:31:09,871 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-6000.pt 2023-04-27 14:31:14,075 INFO [train.py:904] (0/8) Epoch 1, batch 6000, loss[loss=0.3141, simple_loss=0.3661, pruned_loss=0.1311, over 16513.00 frames. ], tot_loss[loss=0.3568, simple_loss=0.4071, pruned_loss=0.1532, over 3092411.77 frames. ], batch size: 68, lr: 4.00e-02, grad_scale: 8.0 2023-04-27 14:31:14,076 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 14:31:23,946 INFO [train.py:938] (0/8) Epoch 1, validation: loss=0.2762, simple_loss=0.3752, pruned_loss=0.08863, over 944034.00 frames. 2023-04-27 14:31:23,947 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17656MB 2023-04-27 14:31:31,567 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6006.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:31:41,458 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.294e+02 5.611e+02 7.114e+02 9.006e+02 1.900e+03, threshold=1.423e+03, percent-clipped=2.0 2023-04-27 14:32:29,959 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6043.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:32:42,921 INFO [train.py:904] (0/8) Epoch 1, batch 6050, loss[loss=0.3454, simple_loss=0.4089, pruned_loss=0.141, over 16490.00 frames. ], tot_loss[loss=0.3541, simple_loss=0.405, pruned_loss=0.1516, over 3110597.92 frames. ], batch size: 68, lr: 3.99e-02, grad_scale: 8.0 2023-04-27 14:32:44,292 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6051.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:33:45,937 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7087, 3.3359, 2.1344, 3.8021, 3.9431, 3.8099, 2.2264, 3.4411], device='cuda:0'), covar=tensor([0.2458, 0.0246, 0.1809, 0.0102, 0.0101, 0.0225, 0.0879, 0.0311], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0082, 0.0151, 0.0051, 0.0052, 0.0066, 0.0115, 0.0091], device='cuda:0'), out_proj_covar=tensor([2.0456e-04, 1.1457e-04, 1.9167e-04, 8.2313e-05, 8.5536e-05, 1.1712e-04, 1.5756e-04, 1.2808e-04], device='cuda:0') 2023-04-27 14:34:03,533 INFO [train.py:904] (0/8) Epoch 1, batch 6100, loss[loss=0.3921, simple_loss=0.4093, pruned_loss=0.1874, over 11589.00 frames. ], tot_loss[loss=0.351, simple_loss=0.403, pruned_loss=0.1495, over 3105377.70 frames. ], batch size: 248, lr: 3.98e-02, grad_scale: 8.0 2023-04-27 14:34:08,711 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6104.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:21,380 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:34:24,022 INFO [optim.py:368] (0/8) 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 14:34:56,493 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6469, 5.4543, 5.2783, 5.3011, 5.3387, 5.8387, 5.6067, 5.2586], device='cuda:0'), covar=tensor([0.0612, 0.0873, 0.0717, 0.0996, 0.1383, 0.0422, 0.0583, 0.1247], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0196, 0.0159, 0.0167, 0.0212, 0.0154, 0.0154, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 14:34:59,860 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6135.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:35:26,100 INFO [train.py:904] (0/8) Epoch 1, batch 6150, loss[loss=0.2995, simple_loss=0.3685, pruned_loss=0.1153, over 16713.00 frames. ], tot_loss[loss=0.347, simple_loss=0.3991, pruned_loss=0.1474, over 3112909.46 frames. ], batch size: 89, lr: 3.97e-02, grad_scale: 8.0 2023-04-27 14:35:57,133 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6170.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:39,058 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6196.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:36:46,265 INFO [train.py:904] (0/8) Epoch 1, batch 6200, loss[loss=0.3489, simple_loss=0.3938, pruned_loss=0.152, over 16775.00 frames. ], tot_loss[loss=0.3463, simple_loss=0.3978, pruned_loss=0.1475, over 3111871.99 frames. ], batch size: 124, lr: 3.96e-02, grad_scale: 8.0 2023-04-27 14:36:49,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0458, 3.6630, 3.9135, 4.1035, 3.4796, 3.9141, 3.7844, 3.4794], device='cuda:0'), covar=tensor([0.0258, 0.0211, 0.0181, 0.0112, 0.0774, 0.0193, 0.0332, 0.0237], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0063, 0.0114, 0.0090, 0.0141, 0.0091, 0.0078, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 14:36:49,314 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:03,238 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6211.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:05,931 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.990e+02 5.160e+02 6.786e+02 8.603e+02 1.824e+03, threshold=1.357e+03, percent-clipped=7.0 2023-04-27 14:37:08,090 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6214.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:37:27,976 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-27 14:37:33,337 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6231.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:02,372 INFO [train.py:904] (0/8) Epoch 1, batch 6250, loss[loss=0.3573, simple_loss=0.4188, pruned_loss=0.1479, over 16780.00 frames. ], tot_loss[loss=0.3465, simple_loss=0.3974, pruned_loss=0.1478, over 3092321.57 frames. ], batch size: 76, lr: 3.95e-02, grad_scale: 8.0 2023-04-27 14:38:20,996 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6264.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:24,220 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6266.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:32,058 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6272.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:38:38,957 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6854, 3.1655, 2.9396, 3.9684, 2.7657, 3.8330, 2.9834, 3.0566], device='cuda:0'), covar=tensor([0.0232, 0.0348, 0.0333, 0.0202, 0.1187, 0.0163, 0.0518, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0094, 0.0078, 0.0095, 0.0166, 0.0090, 0.0114, 0.0110], device='cuda:0'), out_proj_covar=tensor([1.1715e-04, 1.1612e-04, 9.6596e-05, 1.2682e-04, 2.1152e-04, 1.1242e-04, 1.2941e-04, 1.4768e-04], device='cuda:0') 2023-04-27 14:38:48,363 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 14:39:17,032 INFO [train.py:904] (0/8) Epoch 1, batch 6300, loss[loss=0.3306, simple_loss=0.3808, pruned_loss=0.1402, over 16678.00 frames. ], tot_loss[loss=0.3449, simple_loss=0.3966, pruned_loss=0.1466, over 3096360.52 frames. ], batch size: 57, lr: 3.94e-02, grad_scale: 8.0 2023-04-27 14:39:25,261 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6306.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:39:36,725 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.400e+02 5.487e+02 6.666e+02 8.343e+02 1.856e+03, threshold=1.333e+03, percent-clipped=2.0 2023-04-27 14:39:59,073 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6327.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:28,670 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6346.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:40:36,343 INFO [train.py:904] (0/8) Epoch 1, batch 6350, loss[loss=0.3177, simple_loss=0.3784, pruned_loss=0.1285, over 16882.00 frames. ], tot_loss[loss=0.3479, simple_loss=0.3982, pruned_loss=0.1488, over 3103312.73 frames. ], batch size: 96, lr: 3.93e-02, grad_scale: 8.0 2023-04-27 14:40:40,275 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6354.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:41:49,933 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:41:53,496 INFO [train.py:904] (0/8) Epoch 1, batch 6400, loss[loss=0.4509, simple_loss=0.464, pruned_loss=0.2189, over 11001.00 frames. ], tot_loss[loss=0.3496, simple_loss=0.3989, pruned_loss=0.1502, over 3092833.28 frames. ], batch size: 247, lr: 3.92e-02, grad_scale: 8.0 2023-04-27 14:42:08,602 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6411.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:42:10,412 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 4.039e+02 5.999e+02 7.400e+02 9.043e+02 1.587e+03, threshold=1.480e+03, percent-clipped=2.0 2023-04-27 14:42:29,715 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6502, 4.8247, 4.5187, 4.6955, 4.2355, 4.6036, 4.3751, 4.8895], device='cuda:0'), covar=tensor([0.0344, 0.0592, 0.0718, 0.0301, 0.0582, 0.0349, 0.0482, 0.0380], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0199, 0.0190, 0.0125, 0.0162, 0.0132, 0.0172, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-27 14:43:09,640 INFO [train.py:904] (0/8) Epoch 1, batch 6450, loss[loss=0.3227, simple_loss=0.36, pruned_loss=0.1426, over 11674.00 frames. ], tot_loss[loss=0.3472, simple_loss=0.3978, pruned_loss=0.1483, over 3080209.00 frames. ], batch size: 248, lr: 3.91e-02, grad_scale: 8.0 2023-04-27 14:43:22,172 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:26,521 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6462.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:43:33,741 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:12,253 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6491.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:26,977 INFO [train.py:904] (0/8) Epoch 1, batch 6500, loss[loss=0.2788, simple_loss=0.3437, pruned_loss=0.107, over 17219.00 frames. ], tot_loss[loss=0.3432, simple_loss=0.3939, pruned_loss=0.1463, over 3101605.55 frames. ], batch size: 44, lr: 3.90e-02, grad_scale: 16.0 2023-04-27 14:44:45,189 INFO [optim.py:368] (0/8) 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 14:44:46,879 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6514.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:44:59,742 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6523.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 14:45:04,521 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6526.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:07,580 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6528.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:45:42,982 INFO [train.py:904] (0/8) Epoch 1, batch 6550, loss[loss=0.343, simple_loss=0.4145, pruned_loss=0.1357, over 16277.00 frames. ], tot_loss[loss=0.346, simple_loss=0.3969, pruned_loss=0.1476, over 3095690.63 frames. ], batch size: 165, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:45:55,592 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6559.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:00,521 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6562.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:08,772 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6567.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:46:59,704 INFO [train.py:904] (0/8) Epoch 1, batch 6600, loss[loss=0.3528, simple_loss=0.4031, pruned_loss=0.1513, over 16495.00 frames. ], tot_loss[loss=0.3489, simple_loss=0.3998, pruned_loss=0.149, over 3089181.61 frames. ], batch size: 68, lr: 3.89e-02, grad_scale: 16.0 2023-04-27 14:47:18,180 INFO [optim.py:368] (0/8) 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 14:47:32,828 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6622.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:47:43,252 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6629.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:10,703 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6646.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:48:14,677 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1600, 3.1443, 3.0877, 3.4084, 3.3730, 3.3069, 3.3065, 3.3369], device='cuda:0'), covar=tensor([0.0361, 0.0315, 0.0970, 0.0300, 0.0371, 0.0477, 0.0438, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0156, 0.0248, 0.0171, 0.0145, 0.0157, 0.0130, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 14:48:18,090 INFO [train.py:904] (0/8) Epoch 1, batch 6650, loss[loss=0.3854, simple_loss=0.4195, pruned_loss=0.1757, over 15477.00 frames. ], tot_loss[loss=0.3482, simple_loss=0.3991, pruned_loss=0.1487, over 3092584.85 frames. ], batch size: 190, lr: 3.88e-02, grad_scale: 16.0 2023-04-27 14:49:18,743 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:23,555 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6694.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:32,158 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:49:34,324 INFO [train.py:904] (0/8) Epoch 1, batch 6700, loss[loss=0.3305, simple_loss=0.3832, pruned_loss=0.1389, over 16901.00 frames. ], tot_loss[loss=0.3487, simple_loss=0.3984, pruned_loss=0.1495, over 3079085.16 frames. ], batch size: 109, lr: 3.87e-02, grad_scale: 16.0 2023-04-27 14:49:52,617 INFO [optim.py:368] (0/8) 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 14:49:58,920 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2814, 3.9719, 3.6662, 3.3612, 4.1209, 3.0715, 3.7138, 3.9608], device='cuda:0'), covar=tensor([0.0086, 0.0105, 0.0114, 0.0422, 0.0059, 0.0501, 0.0104, 0.0118], device='cuda:0'), in_proj_covar=tensor([0.0048, 0.0037, 0.0053, 0.0075, 0.0040, 0.0072, 0.0052, 0.0052], device='cuda:0'), out_proj_covar=tensor([1.2022e-04, 9.4388e-05, 1.3480e-04, 1.7031e-04, 9.4616e-05, 1.6731e-04, 1.3380e-04, 1.4352e-04], device='cuda:0') 2023-04-27 14:50:28,474 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 14:50:37,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9135, 4.0537, 3.7925, 3.9777, 3.5388, 3.9179, 3.8514, 3.9922], device='cuda:0'), covar=tensor([0.0425, 0.0637, 0.0766, 0.0322, 0.0596, 0.0491, 0.0487, 0.0525], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0192, 0.0184, 0.0124, 0.0155, 0.0128, 0.0173, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-27 14:50:45,164 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6747.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:50:50,996 INFO [train.py:904] (0/8) Epoch 1, batch 6750, loss[loss=0.3317, simple_loss=0.3846, pruned_loss=0.1394, over 16397.00 frames. ], tot_loss[loss=0.3468, simple_loss=0.3964, pruned_loss=0.1486, over 3083534.89 frames. ], batch size: 146, lr: 3.86e-02, grad_scale: 16.0 2023-04-27 14:51:23,693 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0135, 4.1088, 4.0419, 4.0451, 4.2357, 4.5818, 4.4833, 4.0643], device='cuda:0'), covar=tensor([0.1131, 0.1279, 0.0962, 0.1718, 0.1896, 0.0680, 0.0736, 0.1977], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0205, 0.0172, 0.0181, 0.0223, 0.0171, 0.0165, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 14:51:50,847 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:03,863 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9664, 2.5899, 2.4074, 3.1500, 2.4758, 2.9553, 2.7193, 2.2684], device='cuda:0'), covar=tensor([0.0279, 0.0285, 0.0241, 0.0262, 0.0749, 0.0214, 0.0439, 0.0853], device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0101, 0.0084, 0.0108, 0.0178, 0.0097, 0.0123, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 14:52:05,980 INFO [train.py:904] (0/8) Epoch 1, batch 6800, loss[loss=0.3208, simple_loss=0.3928, pruned_loss=0.1244, over 16791.00 frames. ], tot_loss[loss=0.346, simple_loss=0.396, pruned_loss=0.148, over 3083651.98 frames. ], batch size: 102, lr: 3.85e-02, grad_scale: 16.0 2023-04-27 14:52:24,956 INFO [optim.py:368] (0/8) 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 14:52:33,201 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6823.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:52:46,859 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:02,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2561, 4.2465, 2.3848, 4.7896, 4.8382, 4.6716, 2.8180, 4.5905], device='cuda:0'), covar=tensor([0.2011, 0.0199, 0.1747, 0.0074, 0.0098, 0.0219, 0.0795, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0084, 0.0154, 0.0052, 0.0058, 0.0071, 0.0120, 0.0100], device='cuda:0'), out_proj_covar=tensor([2.1438e-04, 1.2325e-04, 2.0321e-04, 8.8478e-05, 9.8114e-05, 1.2806e-04, 1.7221e-04, 1.4632e-04], device='cuda:0') 2023-04-27 14:53:05,418 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6839.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:23,176 INFO [train.py:904] (0/8) Epoch 1, batch 6850, loss[loss=0.2832, simple_loss=0.3679, pruned_loss=0.09927, over 17115.00 frames. ], tot_loss[loss=0.3502, simple_loss=0.3991, pruned_loss=0.1507, over 3053909.35 frames. ], batch size: 49, lr: 3.84e-02, grad_scale: 16.0 2023-04-27 14:53:35,282 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6859.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:42,859 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9289, 4.0597, 1.6936, 4.1075, 2.4574, 3.9565, 1.8069, 2.8390], device='cuda:0'), covar=tensor([0.0067, 0.0114, 0.1717, 0.0049, 0.0772, 0.0172, 0.1472, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0056, 0.0058, 0.0125, 0.0054, 0.0106, 0.0061, 0.0134, 0.0097], device='cuda:0'), out_proj_covar=tensor([9.3158e-05, 1.0248e-04, 1.8894e-04, 8.4571e-05, 1.6433e-04, 1.1480e-04, 2.0293e-04, 1.6163e-04], device='cuda:0') 2023-04-27 14:53:45,273 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2116, 3.6185, 4.0823, 4.2850, 3.1816, 4.0543, 4.0824, 3.7158], device='cuda:0'), covar=tensor([0.0320, 0.0305, 0.0282, 0.0137, 0.1097, 0.0252, 0.0463, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0064, 0.0116, 0.0088, 0.0140, 0.0091, 0.0081, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 14:53:46,335 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:53:57,191 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6874.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:03,524 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6878.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 14:54:33,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5506, 3.2421, 2.7112, 3.1692, 2.5950, 1.9971, 3.3530, 3.8713], device='cuda:0'), covar=tensor([0.1525, 0.0531, 0.1009, 0.0306, 0.1550, 0.1311, 0.0291, 0.0065], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0150, 0.0179, 0.0105, 0.0191, 0.0143, 0.0116, 0.0063], device='cuda:0'), out_proj_covar=tensor([2.3284e-04, 1.8335e-04, 1.9691e-04, 1.2338e-04, 2.3217e-04, 1.6848e-04, 1.4038e-04, 8.1823e-05], device='cuda:0') 2023-04-27 14:54:37,402 INFO [train.py:904] (0/8) Epoch 1, batch 6900, loss[loss=0.4465, simple_loss=0.4478, pruned_loss=0.2226, over 11353.00 frames. ], tot_loss[loss=0.3504, simple_loss=0.4008, pruned_loss=0.1499, over 3043322.86 frames. ], batch size: 246, lr: 3.83e-02, grad_scale: 16.0 2023-04-27 14:54:47,055 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6907.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:49,869 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=6909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:54:55,482 INFO [optim.py:368] (0/8) 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 14:54:57,035 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3171, 3.2962, 3.0649, 2.8936, 3.3255, 2.6416, 3.0220, 3.0307], device='cuda:0'), covar=tensor([0.0111, 0.0066, 0.0105, 0.0267, 0.0063, 0.0450, 0.0092, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0047, 0.0036, 0.0052, 0.0073, 0.0039, 0.0071, 0.0051, 0.0051], device='cuda:0'), out_proj_covar=tensor([1.2203e-04, 9.2398e-05, 1.3512e-04, 1.6726e-04, 9.4866e-05, 1.6713e-04, 1.3270e-04, 1.4598e-04], device='cuda:0') 2023-04-27 14:54:59,428 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6915.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:07,127 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-04-27 14:55:09,435 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=6922.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:55:36,081 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6939.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 14:55:54,246 INFO [train.py:904] (0/8) Epoch 1, batch 6950, loss[loss=0.3233, simple_loss=0.3847, pruned_loss=0.131, over 16887.00 frames. ], tot_loss[loss=0.3542, simple_loss=0.4035, pruned_loss=0.1525, over 3052409.74 frames. ], batch size: 116, lr: 3.82e-02, grad_scale: 16.0 2023-04-27 14:55:56,339 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=6.22 vs. limit=5.0 2023-04-27 14:56:25,071 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:25,257 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=6970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:56:25,431 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 14:56:47,695 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=6985.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 14:57:12,243 INFO [train.py:904] (0/8) Epoch 1, batch 7000, loss[loss=0.3181, simple_loss=0.3956, pruned_loss=0.1203, over 17114.00 frames. ], tot_loss[loss=0.3541, simple_loss=0.4038, pruned_loss=0.1522, over 3042560.26 frames. ], batch size: 47, lr: 3.81e-02, grad_scale: 16.0 2023-04-27 14:57:13,881 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8769, 3.6492, 3.4943, 3.1007, 3.7778, 2.7580, 3.4477, 3.6424], device='cuda:0'), covar=tensor([0.0077, 0.0083, 0.0089, 0.0349, 0.0058, 0.0512, 0.0091, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0048, 0.0038, 0.0053, 0.0075, 0.0040, 0.0075, 0.0052, 0.0053], device='cuda:0'), out_proj_covar=tensor([1.2472e-04, 9.7793e-05, 1.3791e-04, 1.7277e-04, 9.7969e-05, 1.7622e-04, 1.3429e-04, 1.5279e-04], device='cuda:0') 2023-04-27 14:57:14,455 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 14:57:30,868 INFO [optim.py:368] (0/8) 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 14:57:52,905 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2017, 4.3377, 3.7337, 1.7944, 3.0354, 2.2105, 3.3309, 4.2576], device='cuda:0'), covar=tensor([0.0252, 0.0198, 0.0276, 0.1922, 0.0836, 0.1230, 0.0870, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0093, 0.0069, 0.0103, 0.0147, 0.0140, 0.0134, 0.0132, 0.0066], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-27 14:58:27,049 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7048.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 14:58:31,415 INFO [train.py:904] (0/8) Epoch 1, batch 7050, loss[loss=0.2891, simple_loss=0.3591, pruned_loss=0.1096, over 16324.00 frames. ], tot_loss[loss=0.3551, simple_loss=0.4051, pruned_loss=0.1526, over 3037389.46 frames. ], batch size: 68, lr: 3.80e-02, grad_scale: 16.0 2023-04-27 14:59:26,158 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7536, 3.4468, 2.9093, 4.1550, 3.0597, 4.1689, 3.3138, 2.9788], device='cuda:0'), covar=tensor([0.0330, 0.0287, 0.0307, 0.0212, 0.1025, 0.0163, 0.0450, 0.0914], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0101, 0.0085, 0.0112, 0.0181, 0.0098, 0.0124, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 14:59:51,872 INFO [train.py:904] (0/8) Epoch 1, batch 7100, loss[loss=0.317, simple_loss=0.3799, pruned_loss=0.127, over 16934.00 frames. ], tot_loss[loss=0.3518, simple_loss=0.4022, pruned_loss=0.1507, over 3031725.18 frames. ], batch size: 90, lr: 3.79e-02, grad_scale: 16.0 2023-04-27 15:00:05,737 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7109.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:00:11,247 INFO [optim.py:368] (0/8) 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 15:00:19,912 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7118.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:00:27,710 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7123.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:11,103 INFO [train.py:904] (0/8) Epoch 1, batch 7150, loss[loss=0.3959, simple_loss=0.4214, pruned_loss=0.1852, over 11424.00 frames. ], tot_loss[loss=0.348, simple_loss=0.3988, pruned_loss=0.1486, over 3033102.29 frames. ], batch size: 248, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:01:30,107 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:34,184 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7166.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:37,201 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8052, 3.6874, 3.5077, 1.5597, 3.7301, 3.7312, 3.1617, 3.2948], device='cuda:0'), covar=tensor([0.0368, 0.0107, 0.0145, 0.2071, 0.0096, 0.0076, 0.0194, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0065, 0.0062, 0.0148, 0.0064, 0.0057, 0.0068, 0.0082], device='cuda:0'), out_proj_covar=tensor([1.3081e-04, 1.0230e-04, 1.0481e-04, 2.2392e-04, 1.0803e-04, 9.4809e-05, 1.2356e-04, 1.2983e-04], device='cuda:0') 2023-04-27 15:01:41,724 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7171.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:01:53,886 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.41 vs. limit=5.0 2023-04-27 15:02:27,368 INFO [train.py:904] (0/8) Epoch 1, batch 7200, loss[loss=0.2698, simple_loss=0.3427, pruned_loss=0.09849, over 16535.00 frames. ], tot_loss[loss=0.3432, simple_loss=0.3954, pruned_loss=0.1456, over 3042037.49 frames. ], batch size: 75, lr: 3.78e-02, grad_scale: 8.0 2023-04-27 15:02:46,731 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.108e+02 4.516e+02 5.501e+02 7.177e+02 1.508e+03, threshold=1.100e+03, percent-clipped=3.0 2023-04-27 15:02:56,334 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4106, 4.0893, 3.9224, 3.5580, 4.1729, 3.1225, 3.9050, 4.0781], device='cuda:0'), covar=tensor([0.0103, 0.0098, 0.0093, 0.0369, 0.0075, 0.0522, 0.0082, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0048, 0.0037, 0.0053, 0.0076, 0.0040, 0.0075, 0.0051, 0.0052], device='cuda:0'), out_proj_covar=tensor([1.2977e-04, 9.9509e-05, 1.4089e-04, 1.7878e-04, 9.9906e-05, 1.7745e-04, 1.3239e-04, 1.5122e-04], device='cuda:0') 2023-04-27 15:03:02,471 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7224.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 15:03:19,632 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7234.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 15:03:46,520 INFO [train.py:904] (0/8) Epoch 1, batch 7250, loss[loss=0.3255, simple_loss=0.3757, pruned_loss=0.1377, over 15447.00 frames. ], tot_loss[loss=0.3399, simple_loss=0.3928, pruned_loss=0.1435, over 3045642.27 frames. ], batch size: 191, lr: 3.77e-02, grad_scale: 8.0 2023-04-27 15:04:06,717 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:10,970 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 15:04:20,609 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7275.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:30,651 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 15:04:35,188 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7285.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:04:58,733 INFO [train.py:904] (0/8) Epoch 1, batch 7300, loss[loss=0.3304, simple_loss=0.3829, pruned_loss=0.139, over 16585.00 frames. ], tot_loss[loss=0.3381, simple_loss=0.3911, pruned_loss=0.1425, over 3040580.23 frames. ], batch size: 62, lr: 3.76e-02, grad_scale: 8.0 2023-04-27 15:05:00,119 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2556, 3.0176, 2.7472, 3.6334, 2.5598, 3.4263, 2.8075, 2.4945], device='cuda:0'), covar=tensor([0.0329, 0.0341, 0.0336, 0.0241, 0.1161, 0.0212, 0.0541, 0.1171], device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0106, 0.0089, 0.0116, 0.0188, 0.0101, 0.0127, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 15:05:01,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3145, 3.1969, 3.0428, 2.8733, 3.2533, 2.4734, 3.0276, 3.0450], device='cuda:0'), covar=tensor([0.0084, 0.0067, 0.0106, 0.0302, 0.0071, 0.0590, 0.0090, 0.0118], device='cuda:0'), in_proj_covar=tensor([0.0049, 0.0039, 0.0054, 0.0079, 0.0042, 0.0079, 0.0053, 0.0054], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 15:05:19,388 INFO [optim.py:368] (0/8) 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 15:05:28,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-27 15:05:30,654 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9472, 4.0048, 1.7577, 3.8931, 2.2439, 3.9870, 1.9116, 2.5312], device='cuda:0'), covar=tensor([0.0037, 0.0057, 0.1432, 0.0035, 0.0754, 0.0101, 0.1331, 0.0473], device='cuda:0'), in_proj_covar=tensor([0.0060, 0.0064, 0.0136, 0.0061, 0.0116, 0.0068, 0.0144, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:05:47,756 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1689, 3.1544, 3.1340, 3.4465, 3.3112, 3.2524, 3.3037, 3.2917], device='cuda:0'), covar=tensor([0.0391, 0.0344, 0.0845, 0.0269, 0.0460, 0.0581, 0.0375, 0.0314], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0154, 0.0237, 0.0167, 0.0145, 0.0156, 0.0127, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:05:49,056 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7333.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:05:53,101 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7336.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:06:15,331 INFO [train.py:904] (0/8) Epoch 1, batch 7350, loss[loss=0.2972, simple_loss=0.3562, pruned_loss=0.1191, over 17108.00 frames. ], tot_loss[loss=0.3351, simple_loss=0.3889, pruned_loss=0.1406, over 3033498.86 frames. ], batch size: 47, lr: 3.75e-02, grad_scale: 8.0 2023-04-27 15:06:22,428 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5947, 3.5146, 2.7591, 3.2555, 2.6684, 2.0608, 3.7861, 4.1858], device='cuda:0'), covar=tensor([0.1894, 0.0545, 0.1135, 0.0402, 0.1983, 0.1489, 0.0201, 0.0062], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0160, 0.0191, 0.0116, 0.0208, 0.0150, 0.0124, 0.0068], device='cuda:0'), out_proj_covar=tensor([2.4774e-04, 1.9288e-04, 2.1084e-04, 1.3709e-04, 2.5106e-04, 1.7933e-04, 1.5149e-04, 8.5305e-05], device='cuda:0') 2023-04-27 15:07:03,537 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8631, 2.6112, 2.4908, 3.1813, 2.3497, 2.8721, 2.5383, 2.2483], device='cuda:0'), covar=tensor([0.0275, 0.0264, 0.0221, 0.0223, 0.0757, 0.0196, 0.0458, 0.0986], device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0108, 0.0090, 0.0121, 0.0190, 0.0104, 0.0130, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 15:07:28,578 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0539, 3.9551, 3.9737, 4.3854, 4.2700, 4.1987, 4.3444, 4.1709], device='cuda:0'), covar=tensor([0.0393, 0.0315, 0.0942, 0.0294, 0.0384, 0.0346, 0.0237, 0.0307], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0160, 0.0247, 0.0177, 0.0153, 0.0159, 0.0132, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:07:30,953 INFO [train.py:904] (0/8) Epoch 1, batch 7400, loss[loss=0.3018, simple_loss=0.3694, pruned_loss=0.1171, over 16572.00 frames. ], tot_loss[loss=0.3358, simple_loss=0.3896, pruned_loss=0.1409, over 3051324.88 frames. ], batch size: 68, lr: 3.74e-02, grad_scale: 8.0 2023-04-27 15:07:35,574 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7404.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:07:40,473 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 15:08:52,566 INFO [train.py:904] (0/8) Epoch 1, batch 7450, loss[loss=0.315, simple_loss=0.3717, pruned_loss=0.1291, over 17034.00 frames. ], tot_loss[loss=0.3365, simple_loss=0.3904, pruned_loss=0.1413, over 3078315.44 frames. ], batch size: 50, lr: 3.73e-02, grad_scale: 8.0 2023-04-27 15:09:16,555 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-27 15:09:22,074 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:10:15,204 INFO [train.py:904] (0/8) Epoch 1, batch 7500, loss[loss=0.3579, simple_loss=0.4114, pruned_loss=0.1522, over 16384.00 frames. ], tot_loss[loss=0.3389, simple_loss=0.3927, pruned_loss=0.1425, over 3064872.19 frames. ], batch size: 146, lr: 3.72e-02, grad_scale: 8.0 2023-04-27 15:10:35,043 INFO [optim.py:368] (0/8) 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 15:10:43,194 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7519.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:10:48,803 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-27 15:11:06,273 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7534.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:11:31,649 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.89 vs. limit=2.0 2023-04-27 15:11:31,899 INFO [train.py:904] (0/8) Epoch 1, batch 7550, loss[loss=0.3669, simple_loss=0.411, pruned_loss=0.1614, over 15455.00 frames. ], tot_loss[loss=0.3395, simple_loss=0.3922, pruned_loss=0.1434, over 3052505.43 frames. ], batch size: 191, lr: 3.72e-02, grad_scale: 4.0 2023-04-27 15:11:42,468 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 15:11:54,179 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7565.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:12:21,505 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7582.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:12:38,192 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1114, 3.8194, 3.9349, 4.1772, 3.4871, 3.9667, 3.8362, 3.6187], device='cuda:0'), covar=tensor([0.0277, 0.0157, 0.0210, 0.0111, 0.0823, 0.0221, 0.0331, 0.0233], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0063, 0.0118, 0.0093, 0.0144, 0.0095, 0.0085, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:12:50,420 INFO [train.py:904] (0/8) Epoch 1, batch 7600, loss[loss=0.3851, simple_loss=0.4157, pruned_loss=0.1773, over 15267.00 frames. ], tot_loss[loss=0.339, simple_loss=0.3914, pruned_loss=0.1433, over 3072473.55 frames. ], batch size: 190, lr: 3.71e-02, grad_scale: 8.0 2023-04-27 15:13:07,469 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7087, 5.4952, 5.3099, 5.3021, 5.3029, 5.7490, 5.5572, 5.3579], device='cuda:0'), covar=tensor([0.0523, 0.0787, 0.0704, 0.1126, 0.1701, 0.0601, 0.0534, 0.1517], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0203, 0.0179, 0.0180, 0.0221, 0.0172, 0.0158, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:13:10,045 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 15:13:26,951 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 15:13:29,708 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:32,919 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8950, 3.9366, 3.5088, 2.7682, 3.7028, 3.5059, 3.9123, 2.0447], device='cuda:0'), covar=tensor([0.1212, 0.0054, 0.0124, 0.0382, 0.0105, 0.0119, 0.0074, 0.0722], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0042, 0.0044, 0.0077, 0.0041, 0.0043, 0.0048, 0.0082], device='cuda:0'), out_proj_covar=tensor([2.2459e-04, 9.1039e-05, 1.0093e-04, 1.5991e-04, 8.9805e-05, 1.0016e-04, 9.9222e-05, 1.7012e-04], device='cuda:0') 2023-04-27 15:13:39,808 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:13:54,075 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2484, 3.2243, 1.5761, 3.3266, 2.0098, 3.1786, 1.6330, 2.4214], device='cuda:0'), covar=tensor([0.0067, 0.0131, 0.1417, 0.0058, 0.0851, 0.0223, 0.1343, 0.0573], device='cuda:0'), in_proj_covar=tensor([0.0060, 0.0064, 0.0137, 0.0061, 0.0119, 0.0071, 0.0146, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:14:13,744 INFO [train.py:904] (0/8) Epoch 1, batch 7650, loss[loss=0.3403, simple_loss=0.3965, pruned_loss=0.142, over 15563.00 frames. ], tot_loss[loss=0.3381, simple_loss=0.391, pruned_loss=0.1426, over 3096752.15 frames. ], batch size: 191, lr: 3.70e-02, grad_scale: 8.0 2023-04-27 15:15:11,409 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1197, 4.4969, 4.1973, 1.8224, 4.5801, 4.4981, 3.8144, 3.8180], device='cuda:0'), covar=tensor([0.0381, 0.0151, 0.0177, 0.1934, 0.0098, 0.0117, 0.0201, 0.0197], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0064, 0.0065, 0.0146, 0.0065, 0.0055, 0.0072, 0.0082], device='cuda:0'), out_proj_covar=tensor([1.3965e-04, 1.0823e-04, 1.1196e-04, 2.2613e-04, 1.1121e-04, 9.6976e-05, 1.3152e-04, 1.3429e-04], device='cuda:0') 2023-04-27 15:15:14,468 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7685.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:15:38,627 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1222, 3.0578, 1.5410, 3.1069, 1.9443, 2.9677, 1.6940, 2.3170], device='cuda:0'), covar=tensor([0.0096, 0.0154, 0.1670, 0.0066, 0.0931, 0.0338, 0.1504, 0.0723], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0067, 0.0142, 0.0063, 0.0123, 0.0074, 0.0152, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:15:39,939 INFO [train.py:904] (0/8) Epoch 1, batch 7700, loss[loss=0.3334, simple_loss=0.3868, pruned_loss=0.14, over 15358.00 frames. ], tot_loss[loss=0.3416, simple_loss=0.3928, pruned_loss=0.1452, over 3069941.23 frames. ], batch size: 190, lr: 3.69e-02, grad_scale: 8.0 2023-04-27 15:15:45,476 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7704.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:16:00,938 INFO [optim.py:368] (0/8) 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 15:16:01,664 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4523, 3.5697, 3.2268, 1.6967, 2.5911, 2.1699, 3.0188, 3.6023], device='cuda:0'), covar=tensor([0.0376, 0.0332, 0.0309, 0.1950, 0.0986, 0.1174, 0.0903, 0.0259], device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0077, 0.0115, 0.0152, 0.0147, 0.0141, 0.0137, 0.0068], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-27 15:16:45,750 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:16:57,057 INFO [train.py:904] (0/8) Epoch 1, batch 7750, loss[loss=0.3687, simple_loss=0.4103, pruned_loss=0.1636, over 15359.00 frames. ], tot_loss[loss=0.3405, simple_loss=0.3926, pruned_loss=0.1442, over 3090968.74 frames. ], batch size: 190, lr: 3.68e-02, grad_scale: 8.0 2023-04-27 15:16:59,255 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7752.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 15:17:10,825 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-04-27 15:17:16,061 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:17:21,402 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 15:18:13,947 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 15:18:14,297 INFO [train.py:904] (0/8) Epoch 1, batch 7800, loss[loss=0.312, simple_loss=0.3823, pruned_loss=0.1208, over 16984.00 frames. ], tot_loss[loss=0.3423, simple_loss=0.3939, pruned_loss=0.1454, over 3088077.73 frames. ], batch size: 41, lr: 3.67e-02, grad_scale: 8.0 2023-04-27 15:18:16,736 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=7802.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:19,404 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=7804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:18:36,281 INFO [optim.py:368] (0/8) 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 15:18:43,851 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7819.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 15:19:29,401 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5618, 3.2459, 2.7627, 3.0269, 2.4943, 2.0267, 3.4492, 3.7305], device='cuda:0'), covar=tensor([0.1420, 0.0623, 0.1028, 0.0306, 0.1733, 0.1251, 0.0238, 0.0079], device='cuda:0'), in_proj_covar=tensor([0.0211, 0.0167, 0.0195, 0.0118, 0.0206, 0.0150, 0.0131, 0.0070], device='cuda:0'), out_proj_covar=tensor([2.5201e-04, 2.0012e-04, 2.1395e-04, 1.3907e-04, 2.4879e-04, 1.7970e-04, 1.5936e-04, 8.8579e-05], device='cuda:0') 2023-04-27 15:19:31,871 INFO [train.py:904] (0/8) Epoch 1, batch 7850, loss[loss=0.4121, simple_loss=0.4346, pruned_loss=0.1948, over 11413.00 frames. ], tot_loss[loss=0.3423, simple_loss=0.3947, pruned_loss=0.145, over 3088252.83 frames. ], batch size: 248, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:19:51,310 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7867.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:19:59,520 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7135, 3.5141, 1.8951, 3.6855, 2.2576, 3.4383, 1.7752, 2.6742], device='cuda:0'), covar=tensor([0.0048, 0.0086, 0.1331, 0.0057, 0.0748, 0.0190, 0.1379, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0065, 0.0139, 0.0063, 0.0117, 0.0072, 0.0149, 0.0111], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:20:39,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-27 15:20:49,093 INFO [train.py:904] (0/8) Epoch 1, batch 7900, loss[loss=0.3229, simple_loss=0.3869, pruned_loss=0.1295, over 16374.00 frames. ], tot_loss[loss=0.3399, simple_loss=0.393, pruned_loss=0.1435, over 3089239.15 frames. ], batch size: 146, lr: 3.66e-02, grad_scale: 8.0 2023-04-27 15:21:11,998 INFO [optim.py:368] (0/8) 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 15:21:27,431 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1067, 3.9779, 4.0855, 4.4636, 4.4646, 4.2125, 4.4181, 4.2803], device='cuda:0'), covar=tensor([0.0379, 0.0366, 0.1003, 0.0313, 0.0319, 0.0290, 0.0246, 0.0280], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0168, 0.0253, 0.0176, 0.0151, 0.0155, 0.0141, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:21:39,243 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=7931.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:08,695 INFO [train.py:904] (0/8) Epoch 1, batch 7950, loss[loss=0.2874, simple_loss=0.3455, pruned_loss=0.1146, over 16447.00 frames. ], tot_loss[loss=0.3405, simple_loss=0.3929, pruned_loss=0.144, over 3083293.10 frames. ], batch size: 68, lr: 3.65e-02, grad_scale: 8.0 2023-04-27 15:22:42,617 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2359, 4.5790, 5.0290, 5.3190, 4.4755, 4.9212, 4.9140, 4.8169], device='cuda:0'), covar=tensor([0.0280, 0.0275, 0.0184, 0.0089, 0.0702, 0.0203, 0.0174, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0067, 0.0122, 0.0097, 0.0145, 0.0094, 0.0088, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:22:45,116 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3338, 4.6401, 4.5758, 4.6247, 4.6020, 5.0635, 4.8860, 4.5167], device='cuda:0'), covar=tensor([0.0812, 0.0961, 0.0897, 0.1197, 0.1716, 0.0660, 0.0717, 0.1809], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0208, 0.0184, 0.0187, 0.0228, 0.0185, 0.0165, 0.0243], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:22:54,184 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=7979.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:22:56,129 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=7980.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:23:26,662 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-8000.pt 2023-04-27 15:23:30,893 INFO [train.py:904] (0/8) Epoch 1, batch 8000, loss[loss=0.3755, simple_loss=0.4081, pruned_loss=0.1715, over 11242.00 frames. ], tot_loss[loss=0.3421, simple_loss=0.3934, pruned_loss=0.1454, over 3046776.38 frames. ], batch size: 247, lr: 3.64e-02, grad_scale: 8.0 2023-04-27 15:23:45,275 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1327, 3.1060, 3.0203, 3.4192, 3.3714, 3.2214, 3.2929, 3.2835], device='cuda:0'), covar=tensor([0.0408, 0.0357, 0.0992, 0.0284, 0.0418, 0.0612, 0.0415, 0.0341], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0169, 0.0254, 0.0176, 0.0152, 0.0157, 0.0142, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:23:51,334 INFO [optim.py:368] (0/8) 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 15:24:45,928 INFO [train.py:904] (0/8) Epoch 1, batch 8050, loss[loss=0.3162, simple_loss=0.3748, pruned_loss=0.1288, over 16911.00 frames. ], tot_loss[loss=0.341, simple_loss=0.3933, pruned_loss=0.1444, over 3065546.41 frames. ], batch size: 109, lr: 3.63e-02, grad_scale: 8.0 2023-04-27 15:24:53,450 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-27 15:25:04,664 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8063.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:25:38,730 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 15:25:57,898 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:00,734 INFO [train.py:904] (0/8) Epoch 1, batch 8100, loss[loss=0.3077, simple_loss=0.37, pruned_loss=0.1228, over 16517.00 frames. ], tot_loss[loss=0.3385, simple_loss=0.3917, pruned_loss=0.1427, over 3071523.21 frames. ], batch size: 68, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:26:01,235 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6965, 4.6839, 4.3034, 3.4700, 4.7041, 2.3416, 4.2249, 4.6958], device='cuda:0'), covar=tensor([0.0137, 0.0074, 0.0077, 0.0452, 0.0059, 0.1012, 0.0097, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0048, 0.0039, 0.0056, 0.0081, 0.0041, 0.0085, 0.0055, 0.0055], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 15:26:15,159 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8111.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:26:22,784 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.844e+02 5.411e+02 6.350e+02 7.571e+02 1.310e+03, threshold=1.270e+03, percent-clipped=1.0 2023-04-27 15:27:16,442 INFO [train.py:904] (0/8) Epoch 1, batch 8150, loss[loss=0.3362, simple_loss=0.3972, pruned_loss=0.1376, over 15347.00 frames. ], tot_loss[loss=0.3355, simple_loss=0.3888, pruned_loss=0.1411, over 3087963.91 frames. ], batch size: 190, lr: 3.62e-02, grad_scale: 4.0 2023-04-27 15:27:27,365 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:27:32,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7390, 3.1294, 3.0088, 2.4436, 3.1394, 3.1052, 3.2551, 1.7436], device='cuda:0'), covar=tensor([0.1217, 0.0140, 0.0109, 0.0410, 0.0107, 0.0111, 0.0118, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0043, 0.0048, 0.0079, 0.0044, 0.0045, 0.0050, 0.0086], device='cuda:0'), out_proj_covar=tensor([2.3034e-04, 9.7388e-05, 1.1021e-04, 1.7045e-04, 9.8763e-05, 1.0824e-04, 1.0841e-04, 1.7970e-04], device='cuda:0') 2023-04-27 15:27:33,738 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4377, 1.4665, 1.8073, 1.3714, 2.0260, 2.0365, 2.3420, 2.3775], device='cuda:0'), covar=tensor([0.0040, 0.0390, 0.0154, 0.0249, 0.0119, 0.0174, 0.0086, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0032, 0.0070, 0.0052, 0.0055, 0.0047, 0.0055, 0.0033, 0.0039], device='cuda:0'), out_proj_covar=tensor([4.2303e-05, 1.0332e-04, 7.5695e-05, 8.1559e-05, 7.0507e-05, 7.8749e-05, 5.4224e-05, 6.1625e-05], device='cuda:0') 2023-04-27 15:28:25,024 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 15:28:33,144 INFO [train.py:904] (0/8) Epoch 1, batch 8200, loss[loss=0.3557, simple_loss=0.4067, pruned_loss=0.1523, over 15394.00 frames. ], tot_loss[loss=0.3318, simple_loss=0.385, pruned_loss=0.1394, over 3093512.05 frames. ], batch size: 190, lr: 3.61e-02, grad_scale: 4.0 2023-04-27 15:28:56,627 INFO [optim.py:368] (0/8) 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 15:29:07,408 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-27 15:29:31,426 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 15:29:53,210 INFO [train.py:904] (0/8) Epoch 1, batch 8250, loss[loss=0.3013, simple_loss=0.3568, pruned_loss=0.1229, over 12059.00 frames. ], tot_loss[loss=0.3294, simple_loss=0.3842, pruned_loss=0.1373, over 3073406.85 frames. ], batch size: 246, lr: 3.60e-02, grad_scale: 4.0 2023-04-27 15:30:15,965 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:30:41,260 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8280.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:14,614 INFO [train.py:904] (0/8) Epoch 1, batch 8300, loss[loss=0.2966, simple_loss=0.3506, pruned_loss=0.1213, over 12155.00 frames. ], tot_loss[loss=0.3199, simple_loss=0.3783, pruned_loss=0.1308, over 3079695.66 frames. ], batch size: 248, lr: 3.59e-02, grad_scale: 4.0 2023-04-27 15:31:34,405 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.88 vs. limit=2.0 2023-04-27 15:31:40,151 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.807e+02 4.177e+02 5.020e+02 6.033e+02 1.438e+03, threshold=1.004e+03, percent-clipped=1.0 2023-04-27 15:31:56,000 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:31:59,950 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:32:03,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6973, 3.4107, 3.2332, 2.6337, 3.1893, 2.9209, 3.3428, 1.8503], device='cuda:0'), covar=tensor([0.1092, 0.0078, 0.0133, 0.0361, 0.0096, 0.0162, 0.0118, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0043, 0.0049, 0.0079, 0.0045, 0.0048, 0.0048, 0.0089], device='cuda:0'), out_proj_covar=tensor([2.3445e-04, 9.9983e-05, 1.1230e-04, 1.7109e-04, 1.0306e-04, 1.1592e-04, 1.0460e-04, 1.8648e-04], device='cuda:0') 2023-04-27 15:32:36,047 INFO [train.py:904] (0/8) Epoch 1, batch 8350, loss[loss=0.3145, simple_loss=0.3843, pruned_loss=0.1224, over 15315.00 frames. ], tot_loss[loss=0.3143, simple_loss=0.3753, pruned_loss=0.1266, over 3063778.50 frames. ], batch size: 191, lr: 3.58e-02, grad_scale: 4.0 2023-04-27 15:33:40,076 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-27 15:33:55,093 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:33:57,935 INFO [train.py:904] (0/8) Epoch 1, batch 8400, loss[loss=0.3072, simple_loss=0.3546, pruned_loss=0.1299, over 12269.00 frames. ], tot_loss[loss=0.3077, simple_loss=0.3706, pruned_loss=0.1224, over 3059348.94 frames. ], batch size: 247, lr: 3.58e-02, grad_scale: 8.0 2023-04-27 15:34:21,545 INFO [optim.py:368] (0/8) 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 15:35:12,365 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:18,239 INFO [train.py:904] (0/8) Epoch 1, batch 8450, loss[loss=0.2594, simple_loss=0.3379, pruned_loss=0.0905, over 15402.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.367, pruned_loss=0.1188, over 3067628.00 frames. ], batch size: 191, lr: 3.57e-02, grad_scale: 8.0 2023-04-27 15:35:30,930 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:35:53,854 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7400, 1.4345, 1.5170, 1.8334, 1.8107, 1.7421, 1.5680, 1.7643], device='cuda:0'), covar=tensor([0.0074, 0.0438, 0.0224, 0.0149, 0.0085, 0.0134, 0.0206, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0044, 0.0091, 0.0073, 0.0054, 0.0042, 0.0046, 0.0064, 0.0041], device='cuda:0'), out_proj_covar=tensor([8.2127e-05, 1.6614e-04, 1.4095e-04, 1.0197e-04, 7.5075e-05, 8.6939e-05, 1.1038e-04, 7.4984e-05], device='cuda:0') 2023-04-27 15:36:39,495 INFO [train.py:904] (0/8) Epoch 1, batch 8500, loss[loss=0.2747, simple_loss=0.3449, pruned_loss=0.1022, over 15116.00 frames. ], tot_loss[loss=0.2938, simple_loss=0.3607, pruned_loss=0.1134, over 3074132.72 frames. ], batch size: 190, lr: 3.56e-02, grad_scale: 8.0 2023-04-27 15:36:49,164 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:37:04,609 INFO [optim.py:368] (0/8) 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 15:37:15,069 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0134, 2.6248, 2.1687, 3.2341, 2.4433, 3.0416, 2.4351, 2.1304], device='cuda:0'), covar=tensor([0.0286, 0.0290, 0.0285, 0.0235, 0.0989, 0.0209, 0.0511, 0.1147], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0116, 0.0094, 0.0124, 0.0194, 0.0114, 0.0133, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:38:03,885 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 15:38:04,000 INFO [train.py:904] (0/8) Epoch 1, batch 8550, loss[loss=0.2799, simple_loss=0.3316, pruned_loss=0.1141, over 11743.00 frames. ], tot_loss[loss=0.2901, simple_loss=0.3571, pruned_loss=0.1116, over 3025789.41 frames. ], batch size: 248, lr: 3.55e-02, grad_scale: 8.0 2023-04-27 15:38:55,163 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1766, 4.1978, 4.4280, 4.4811, 4.6301, 4.2618, 4.2377, 4.3848], device='cuda:0'), covar=tensor([0.0260, 0.0242, 0.0461, 0.0368, 0.0397, 0.0250, 0.0644, 0.0276], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0111, 0.0128, 0.0129, 0.0138, 0.0118, 0.0165, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:39:44,572 INFO [train.py:904] (0/8) Epoch 1, batch 8600, loss[loss=0.2935, simple_loss=0.3614, pruned_loss=0.1128, over 16545.00 frames. ], tot_loss[loss=0.2895, simple_loss=0.3579, pruned_loss=0.1105, over 3028866.52 frames. ], batch size: 68, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:40:17,880 INFO [optim.py:368] (0/8) 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 15:40:27,933 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=8621.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:40:31,428 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 15:41:26,192 INFO [train.py:904] (0/8) Epoch 1, batch 8650, loss[loss=0.2651, simple_loss=0.3357, pruned_loss=0.09726, over 12159.00 frames. ], tot_loss[loss=0.2841, simple_loss=0.3541, pruned_loss=0.107, over 3028266.10 frames. ], batch size: 247, lr: 3.54e-02, grad_scale: 8.0 2023-04-27 15:43:13,575 INFO [train.py:904] (0/8) Epoch 1, batch 8700, loss[loss=0.2809, simple_loss=0.3566, pruned_loss=0.1026, over 15281.00 frames. ], tot_loss[loss=0.2804, simple_loss=0.3506, pruned_loss=0.1051, over 3030710.24 frames. ], batch size: 191, lr: 3.53e-02, grad_scale: 8.0 2023-04-27 15:43:16,671 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5854, 4.8214, 4.8568, 4.9402, 5.0532, 5.4143, 5.3555, 4.9067], device='cuda:0'), covar=tensor([0.0562, 0.1169, 0.0798, 0.1425, 0.1757, 0.0728, 0.0798, 0.1770], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0207, 0.0168, 0.0174, 0.0211, 0.0174, 0.0150, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 15:43:41,226 INFO [optim.py:368] (0/8) 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 15:44:49,929 INFO [train.py:904] (0/8) Epoch 1, batch 8750, loss[loss=0.2431, simple_loss=0.3205, pruned_loss=0.08279, over 12178.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3494, pruned_loss=0.1035, over 3033849.53 frames. ], batch size: 246, lr: 3.52e-02, grad_scale: 8.0 2023-04-27 15:45:14,119 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:46:09,068 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3861, 3.0774, 2.6808, 2.7459, 2.2772, 1.8454, 2.9556, 3.2809], device='cuda:0'), covar=tensor([0.1379, 0.0545, 0.0805, 0.0338, 0.1495, 0.1333, 0.0301, 0.0105], device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0174, 0.0196, 0.0120, 0.0169, 0.0156, 0.0131, 0.0071], device='cuda:0'), out_proj_covar=tensor([2.4753e-04, 2.0344e-04, 2.1607e-04, 1.4150e-04, 2.0582e-04, 1.8927e-04, 1.5919e-04, 9.0911e-05], device='cuda:0') 2023-04-27 15:46:31,266 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5130, 3.4417, 3.0863, 3.2058, 2.6144, 2.1591, 3.5936, 4.0874], device='cuda:0'), covar=tensor([0.1796, 0.0644, 0.0916, 0.0352, 0.1432, 0.1283, 0.0272, 0.0072], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0176, 0.0198, 0.0121, 0.0170, 0.0157, 0.0132, 0.0071], device='cuda:0'), out_proj_covar=tensor([2.4883e-04, 2.0628e-04, 2.1758e-04, 1.4264e-04, 2.0723e-04, 1.9020e-04, 1.6030e-04, 9.1841e-05], device='cuda:0') 2023-04-27 15:46:42,410 INFO [train.py:904] (0/8) Epoch 1, batch 8800, loss[loss=0.2844, simple_loss=0.3456, pruned_loss=0.1115, over 12619.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3463, pruned_loss=0.1014, over 3037173.03 frames. ], batch size: 247, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:47:05,300 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3522, 3.8402, 4.1328, 4.2162, 3.7264, 4.1149, 4.0526, 3.8610], device='cuda:0'), covar=tensor([0.0174, 0.0195, 0.0146, 0.0113, 0.0567, 0.0175, 0.0238, 0.0198], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0061, 0.0114, 0.0088, 0.0136, 0.0089, 0.0083, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:47:13,494 INFO [optim.py:368] (0/8) 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 15:47:25,078 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8821.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:48:27,345 INFO [train.py:904] (0/8) Epoch 1, batch 8850, loss[loss=0.2367, simple_loss=0.3125, pruned_loss=0.08044, over 12355.00 frames. ], tot_loss[loss=0.2744, simple_loss=0.3482, pruned_loss=0.1003, over 3039919.40 frames. ], batch size: 248, lr: 3.51e-02, grad_scale: 8.0 2023-04-27 15:49:34,638 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4838, 3.4917, 3.8107, 3.9094, 3.9966, 3.6331, 3.6788, 3.7577], device='cuda:0'), covar=tensor([0.0263, 0.0296, 0.0409, 0.0333, 0.0305, 0.0268, 0.0581, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0110, 0.0128, 0.0124, 0.0136, 0.0115, 0.0161, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:49:37,803 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 2023-04-27 15:50:13,551 INFO [train.py:904] (0/8) Epoch 1, batch 8900, loss[loss=0.3082, simple_loss=0.3769, pruned_loss=0.1198, over 16264.00 frames. ], tot_loss[loss=0.2725, simple_loss=0.3474, pruned_loss=0.09878, over 3038234.17 frames. ], batch size: 165, lr: 3.50e-02, grad_scale: 8.0 2023-04-27 15:50:42,919 INFO [optim.py:368] (0/8) 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 15:50:47,989 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6168, 3.5006, 3.4533, 1.3472, 3.6644, 3.6602, 3.1432, 3.2632], device='cuda:0'), covar=tensor([0.0441, 0.0075, 0.0127, 0.1964, 0.0061, 0.0064, 0.0193, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0094, 0.0065, 0.0062, 0.0145, 0.0061, 0.0058, 0.0075, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0001], device='cuda:0') 2023-04-27 15:50:53,384 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-27 15:50:56,768 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=8921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:51:13,364 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8927.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:52:04,594 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6913, 3.5845, 3.8985, 4.0308, 4.1126, 3.7279, 3.7496, 3.9115], device='cuda:0'), covar=tensor([0.0275, 0.0333, 0.0503, 0.0494, 0.0381, 0.0278, 0.0590, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0109, 0.0126, 0.0123, 0.0137, 0.0114, 0.0162, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:52:19,954 INFO [train.py:904] (0/8) Epoch 1, batch 8950, loss[loss=0.2486, simple_loss=0.3312, pruned_loss=0.08296, over 16262.00 frames. ], tot_loss[loss=0.2735, simple_loss=0.3475, pruned_loss=0.09974, over 3026227.19 frames. ], batch size: 165, lr: 3.49e-02, grad_scale: 8.0 2023-04-27 15:53:00,573 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=8969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:05,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9619, 3.6782, 2.7870, 4.3804, 2.9179, 4.2522, 3.3127, 2.6741], device='cuda:0'), covar=tensor([0.0251, 0.0221, 0.0284, 0.0182, 0.1070, 0.0151, 0.0433, 0.1070], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0121, 0.0098, 0.0133, 0.0196, 0.0118, 0.0138, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 15:53:38,178 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=8986.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:53:42,192 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=8988.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:08,289 INFO [train.py:904] (0/8) Epoch 1, batch 9000, loss[loss=0.2802, simple_loss=0.3394, pruned_loss=0.1105, over 12509.00 frames. ], tot_loss[loss=0.2687, simple_loss=0.343, pruned_loss=0.09714, over 3034199.62 frames. ], batch size: 250, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:54:08,290 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 15:54:19,193 INFO [train.py:938] (0/8) Epoch 1, validation: loss=0.2299, simple_loss=0.3267, pruned_loss=0.06658, over 944034.00 frames. 2023-04-27 15:54:19,194 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 15:54:46,200 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9013.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:54:51,599 INFO [optim.py:368] (0/8) 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 15:55:56,685 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9047.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:56:02,943 INFO [train.py:904] (0/8) Epoch 1, batch 9050, loss[loss=0.2583, simple_loss=0.3267, pruned_loss=0.09496, over 16270.00 frames. ], tot_loss[loss=0.2719, simple_loss=0.3456, pruned_loss=0.09912, over 3049936.43 frames. ], batch size: 165, lr: 3.48e-02, grad_scale: 8.0 2023-04-27 15:56:48,448 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:57:45,084 INFO [train.py:904] (0/8) Epoch 1, batch 9100, loss[loss=0.297, simple_loss=0.3756, pruned_loss=0.1092, over 16647.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3454, pruned_loss=0.09988, over 3051336.90 frames. ], batch size: 62, lr: 3.47e-02, grad_scale: 8.0 2023-04-27 15:58:14,295 INFO [optim.py:368] (0/8) 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 15:58:14,879 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9116.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 15:59:15,991 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.81 vs. limit=2.0 2023-04-27 15:59:42,647 INFO [train.py:904] (0/8) Epoch 1, batch 9150, loss[loss=0.262, simple_loss=0.3356, pruned_loss=0.0942, over 15413.00 frames. ], tot_loss[loss=0.2718, simple_loss=0.3458, pruned_loss=0.09895, over 3053519.18 frames. ], batch size: 191, lr: 3.46e-02, grad_scale: 8.0 2023-04-27 16:01:27,879 INFO [train.py:904] (0/8) Epoch 1, batch 9200, loss[loss=0.2343, simple_loss=0.3045, pruned_loss=0.08203, over 12448.00 frames. ], tot_loss[loss=0.2686, simple_loss=0.3414, pruned_loss=0.09787, over 3060092.71 frames. ], batch size: 248, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:01:54,928 INFO [optim.py:368] (0/8) 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 16:02:18,787 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4869, 4.8418, 4.7673, 4.8864, 4.7294, 5.2103, 5.1210, 4.7737], device='cuda:0'), covar=tensor([0.0567, 0.0907, 0.0876, 0.1148, 0.1614, 0.0772, 0.0633, 0.1530], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0209, 0.0178, 0.0177, 0.0218, 0.0179, 0.0149, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-27 16:02:29,566 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2172, 3.2187, 3.2442, 3.4961, 3.4068, 3.2159, 3.3814, 3.4120], device='cuda:0'), covar=tensor([0.0358, 0.0310, 0.0845, 0.0294, 0.0411, 0.0823, 0.0424, 0.0270], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0161, 0.0241, 0.0167, 0.0143, 0.0148, 0.0133, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:03:04,242 INFO [train.py:904] (0/8) Epoch 1, batch 9250, loss[loss=0.2647, simple_loss=0.3417, pruned_loss=0.09387, over 16523.00 frames. ], tot_loss[loss=0.2688, simple_loss=0.3413, pruned_loss=0.09816, over 3061349.16 frames. ], batch size: 62, lr: 3.45e-02, grad_scale: 8.0 2023-04-27 16:03:45,757 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4830, 1.6379, 1.7320, 1.4680, 2.1516, 2.0658, 2.2660, 2.3412], device='cuda:0'), covar=tensor([0.0032, 0.0304, 0.0161, 0.0240, 0.0094, 0.0168, 0.0072, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0032, 0.0072, 0.0056, 0.0063, 0.0050, 0.0061, 0.0034, 0.0043], device='cuda:0'), out_proj_covar=tensor([4.4451e-05, 1.1044e-04, 8.4905e-05, 9.6607e-05, 7.7394e-05, 9.1934e-05, 5.2046e-05, 7.0755e-05], device='cuda:0') 2023-04-27 16:04:16,342 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9283.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:04:19,003 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-27 16:04:58,126 INFO [train.py:904] (0/8) Epoch 1, batch 9300, loss[loss=0.2367, simple_loss=0.3154, pruned_loss=0.07897, over 16571.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3385, pruned_loss=0.09628, over 3061404.46 frames. ], batch size: 68, lr: 3.44e-02, grad_scale: 8.0 2023-04-27 16:05:33,419 INFO [optim.py:368] (0/8) 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 16:06:25,977 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-27 16:06:27,286 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9342.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:06:44,340 INFO [train.py:904] (0/8) Epoch 1, batch 9350, loss[loss=0.2732, simple_loss=0.3456, pruned_loss=0.1004, over 16314.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3384, pruned_loss=0.09602, over 3069709.16 frames. ], batch size: 146, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:07:24,453 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9369.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:07:24,818 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 16:08:26,602 INFO [train.py:904] (0/8) Epoch 1, batch 9400, loss[loss=0.2389, simple_loss=0.3071, pruned_loss=0.08531, over 12527.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.3377, pruned_loss=0.09508, over 3062787.40 frames. ], batch size: 250, lr: 3.43e-02, grad_scale: 8.0 2023-04-27 16:08:47,060 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8869, 3.7979, 3.9806, 4.2708, 4.2219, 3.9941, 4.2438, 4.0907], device='cuda:0'), covar=tensor([0.0399, 0.0286, 0.0813, 0.0264, 0.0316, 0.0352, 0.0244, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0157, 0.0235, 0.0164, 0.0137, 0.0143, 0.0131, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:08:54,853 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:08:56,915 INFO [optim.py:368] (0/8) 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 16:08:57,389 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9416.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:08,170 INFO [train.py:904] (0/8) Epoch 1, batch 9450, loss[loss=0.2537, simple_loss=0.322, pruned_loss=0.09274, over 12318.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3394, pruned_loss=0.09526, over 3061368.26 frames. ], batch size: 249, lr: 3.42e-02, grad_scale: 8.0 2023-04-27 16:10:33,966 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:10:58,573 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:11:35,311 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4042, 3.1117, 3.1093, 1.6757, 3.1873, 3.1957, 2.8081, 2.9488], device='cuda:0'), covar=tensor([0.0555, 0.0110, 0.0198, 0.1717, 0.0101, 0.0099, 0.0348, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0104, 0.0073, 0.0071, 0.0151, 0.0065, 0.0068, 0.0081, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:11:47,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0121, 2.6578, 2.2110, 3.2467, 2.3006, 3.0782, 2.4696, 2.1038], device='cuda:0'), covar=tensor([0.0320, 0.0287, 0.0268, 0.0291, 0.1054, 0.0189, 0.0513, 0.1182], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0127, 0.0103, 0.0144, 0.0201, 0.0121, 0.0143, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:11:50,038 INFO [train.py:904] (0/8) Epoch 1, batch 9500, loss[loss=0.2713, simple_loss=0.3498, pruned_loss=0.09639, over 16226.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3382, pruned_loss=0.09413, over 3062335.06 frames. ], batch size: 165, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:12:01,167 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-04-27 16:12:21,415 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.089e+02 4.155e+02 5.314e+02 6.779e+02 1.064e+03, threshold=1.063e+03, percent-clipped=4.0 2023-04-27 16:13:37,018 INFO [train.py:904] (0/8) Epoch 1, batch 9550, loss[loss=0.2696, simple_loss=0.3345, pruned_loss=0.1024, over 12173.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.337, pruned_loss=0.09429, over 3041591.10 frames. ], batch size: 247, lr: 3.41e-02, grad_scale: 8.0 2023-04-27 16:14:41,653 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=2.00 vs. limit=2.0 2023-04-27 16:14:46,564 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9583.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:14:48,090 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9584.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:15:18,635 INFO [train.py:904] (0/8) Epoch 1, batch 9600, loss[loss=0.2546, simple_loss=0.3225, pruned_loss=0.09334, over 11996.00 frames. ], tot_loss[loss=0.2649, simple_loss=0.3386, pruned_loss=0.0956, over 3035188.58 frames. ], batch size: 248, lr: 3.40e-02, grad_scale: 8.0 2023-04-27 16:15:32,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5218, 4.2464, 1.6376, 4.2536, 2.4267, 4.4082, 1.7019, 3.0338], device='cuda:0'), covar=tensor([0.0029, 0.0074, 0.1642, 0.0053, 0.0882, 0.0108, 0.1582, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0066, 0.0076, 0.0158, 0.0071, 0.0137, 0.0087, 0.0165, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-27 16:15:48,622 INFO [optim.py:368] (0/8) 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 16:16:20,299 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9631.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:45,960 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9642.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:16:48,388 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.4520, 3.5305, 2.2833, 4.1495, 4.2256, 4.1476, 1.8972, 3.1276], device='cuda:0'), covar=tensor([0.2571, 0.0324, 0.1722, 0.0110, 0.0110, 0.0298, 0.1452, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0096, 0.0159, 0.0060, 0.0071, 0.0084, 0.0138, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:16:53,240 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9645.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:08,573 INFO [train.py:904] (0/8) Epoch 1, batch 9650, loss[loss=0.2563, simple_loss=0.3316, pruned_loss=0.09048, over 12089.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3412, pruned_loss=0.09582, over 3048295.07 frames. ], batch size: 247, lr: 3.39e-02, grad_scale: 8.0 2023-04-27 16:17:26,899 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9658.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:17:52,715 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=9669.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:35,173 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9690.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:18:57,055 INFO [train.py:904] (0/8) Epoch 1, batch 9700, loss[loss=0.2924, simple_loss=0.3546, pruned_loss=0.1152, over 12600.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3396, pruned_loss=0.09496, over 3050021.83 frames. ], batch size: 248, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:19:24,080 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4447, 3.3819, 3.2143, 3.3955, 3.0374, 3.2918, 3.2665, 3.1097], device='cuda:0'), covar=tensor([0.0240, 0.0134, 0.0192, 0.0136, 0.0558, 0.0152, 0.0674, 0.0213], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0066, 0.0114, 0.0091, 0.0139, 0.0090, 0.0082, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:19:24,741 INFO [optim.py:368] (0/8) 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 16:19:28,116 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=9717.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:19:33,541 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9719.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:20:27,167 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=9743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:20:40,576 INFO [train.py:904] (0/8) Epoch 1, batch 9750, loss[loss=0.2781, simple_loss=0.3397, pruned_loss=0.1083, over 12649.00 frames. ], tot_loss[loss=0.2642, simple_loss=0.3382, pruned_loss=0.09511, over 3047368.28 frames. ], batch size: 248, lr: 3.38e-02, grad_scale: 8.0 2023-04-27 16:21:16,663 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:19,302 INFO [train.py:904] (0/8) Epoch 1, batch 9800, loss[loss=0.2611, simple_loss=0.346, pruned_loss=0.0881, over 16678.00 frames. ], tot_loss[loss=0.2621, simple_loss=0.3378, pruned_loss=0.09321, over 3057299.54 frames. ], batch size: 134, lr: 3.37e-02, grad_scale: 8.0 2023-04-27 16:22:21,786 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2681, 3.2523, 2.8399, 2.8683, 2.3505, 1.9823, 3.2216, 3.7748], device='cuda:0'), covar=tensor([0.1719, 0.0632, 0.1023, 0.0400, 0.1368, 0.1324, 0.0341, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0193, 0.0210, 0.0131, 0.0175, 0.0166, 0.0147, 0.0080], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-27 16:22:26,159 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=9804.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:22:47,909 INFO [optim.py:368] (0/8) 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 16:24:05,893 INFO [train.py:904] (0/8) Epoch 1, batch 9850, loss[loss=0.2642, simple_loss=0.337, pruned_loss=0.0957, over 12356.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3391, pruned_loss=0.09319, over 3043492.64 frames. ], batch size: 249, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:25:58,737 INFO [train.py:904] (0/8) Epoch 1, batch 9900, loss[loss=0.2644, simple_loss=0.3488, pruned_loss=0.09002, over 16840.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3393, pruned_loss=0.09286, over 3043254.37 frames. ], batch size: 124, lr: 3.36e-02, grad_scale: 8.0 2023-04-27 16:26:29,443 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 16:26:31,898 INFO [optim.py:368] (0/8) 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 16:27:32,226 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=9940.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:27:57,263 INFO [train.py:904] (0/8) Epoch 1, batch 9950, loss[loss=0.2819, simple_loss=0.3512, pruned_loss=0.1063, over 16994.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3415, pruned_loss=0.09369, over 3049752.68 frames. ], batch size: 109, lr: 3.35e-02, grad_scale: 8.0 2023-04-27 16:28:01,930 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0219, 3.9968, 3.8567, 1.6007, 3.1544, 2.3476, 3.5177, 4.2805], device='cuda:0'), covar=tensor([0.0249, 0.0471, 0.0298, 0.2257, 0.0762, 0.1236, 0.0713, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0084, 0.0127, 0.0157, 0.0147, 0.0140, 0.0141, 0.0080], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 16:29:25,713 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.09 vs. limit=2.0 2023-04-27 16:29:57,912 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-10000.pt 2023-04-27 16:30:02,401 INFO [train.py:904] (0/8) Epoch 1, batch 10000, loss[loss=0.2347, simple_loss=0.33, pruned_loss=0.06966, over 16376.00 frames. ], tot_loss[loss=0.2627, simple_loss=0.3395, pruned_loss=0.09299, over 3055948.28 frames. ], batch size: 146, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:30:30,116 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10014.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:30:32,721 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.207e+02 4.345e+02 5.531e+02 6.703e+02 1.367e+03, threshold=1.106e+03, percent-clipped=3.0 2023-04-27 16:30:33,368 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9545, 2.7103, 2.5852, 1.6607, 2.6915, 2.8351, 2.5201, 2.4665], device='cuda:0'), covar=tensor([0.0567, 0.0108, 0.0197, 0.1536, 0.0103, 0.0090, 0.0259, 0.0255], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0067, 0.0067, 0.0147, 0.0062, 0.0063, 0.0078, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:31:21,940 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10039.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:31:43,685 INFO [train.py:904] (0/8) Epoch 1, batch 10050, loss[loss=0.2705, simple_loss=0.3473, pruned_loss=0.09686, over 16899.00 frames. ], tot_loss[loss=0.2603, simple_loss=0.3377, pruned_loss=0.09148, over 3047112.82 frames. ], batch size: 96, lr: 3.34e-02, grad_scale: 8.0 2023-04-27 16:32:06,097 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10062.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:32:23,045 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:16,689 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10099.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:18,978 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10100.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 16:33:19,592 INFO [train.py:904] (0/8) Epoch 1, batch 10100, loss[loss=0.2665, simple_loss=0.3292, pruned_loss=0.1019, over 12939.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3393, pruned_loss=0.09284, over 3053248.09 frames. ], batch size: 247, lr: 3.33e-02, grad_scale: 16.0 2023-04-27 16:33:20,099 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2849, 4.4365, 4.4417, 4.4687, 4.4700, 4.9092, 4.8628, 4.4423], device='cuda:0'), covar=tensor([0.0859, 0.1123, 0.0785, 0.1223, 0.2030, 0.0666, 0.0666, 0.1604], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0212, 0.0181, 0.0182, 0.0226, 0.0182, 0.0157, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:33:49,435 INFO [optim.py:368] (0/8) 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 16:33:53,061 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:33:58,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6061, 3.2198, 2.0819, 3.7431, 3.7962, 3.6860, 2.1465, 3.0442], device='cuda:0'), covar=tensor([0.2304, 0.0332, 0.1727, 0.0103, 0.0113, 0.0322, 0.1221, 0.0555], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0099, 0.0157, 0.0059, 0.0071, 0.0084, 0.0140, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:34:06,072 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10123.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:34:36,273 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-27 16:34:38,595 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-1.pt 2023-04-27 16:35:05,149 INFO [train.py:904] (0/8) Epoch 2, batch 0, loss[loss=0.497, simple_loss=0.4741, pruned_loss=0.2599, over 16687.00 frames. ], tot_loss[loss=0.497, simple_loss=0.4741, pruned_loss=0.2599, over 16687.00 frames. ], batch size: 134, lr: 3.26e-02, grad_scale: 8.0 2023-04-27 16:35:05,150 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 16:35:12,741 INFO [train.py:938] (0/8) Epoch 2, validation: loss=0.2235, simple_loss=0.3221, pruned_loss=0.06242, over 944034.00 frames. 2023-04-27 16:35:12,741 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 16:35:26,371 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3561, 5.7839, 5.4375, 5.7359, 4.8996, 4.7172, 5.3034, 5.9151], device='cuda:0'), covar=tensor([0.0449, 0.0668, 0.0791, 0.0286, 0.0551, 0.0440, 0.0427, 0.0457], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0218, 0.0195, 0.0133, 0.0160, 0.0128, 0.0179, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:35:48,111 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1998, 4.7581, 4.5829, 4.0855, 4.9453, 1.9625, 4.4854, 5.0800], device='cuda:0'), covar=tensor([0.0069, 0.0094, 0.0080, 0.0297, 0.0058, 0.1397, 0.0090, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0047, 0.0042, 0.0060, 0.0071, 0.0043, 0.0099, 0.0057, 0.0056], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:35:56,914 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8878, 4.6763, 4.8949, 5.3155, 5.4110, 4.8433, 5.5057, 5.1879], device='cuda:0'), covar=tensor([0.0377, 0.0371, 0.0933, 0.0308, 0.0379, 0.0288, 0.0177, 0.0280], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0187, 0.0277, 0.0197, 0.0161, 0.0164, 0.0151, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:36:12,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7060, 4.4740, 4.2081, 4.9276, 5.0510, 4.6417, 5.1426, 4.9664], device='cuda:0'), covar=tensor([0.0372, 0.0428, 0.1325, 0.0465, 0.0414, 0.0334, 0.0366, 0.0329], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0192, 0.0285, 0.0202, 0.0163, 0.0165, 0.0154, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:36:22,854 INFO [train.py:904] (0/8) Epoch 2, batch 50, loss[loss=0.3672, simple_loss=0.3848, pruned_loss=0.1748, over 16713.00 frames. ], tot_loss[loss=0.3301, simple_loss=0.3741, pruned_loss=0.1431, over 749796.08 frames. ], batch size: 134, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:36:45,681 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.307e+02 5.068e+02 6.475e+02 7.712e+02 1.660e+03, threshold=1.295e+03, percent-clipped=4.0 2023-04-27 16:37:16,283 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10240.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:37:30,898 INFO [train.py:904] (0/8) Epoch 2, batch 100, loss[loss=0.3019, simple_loss=0.3512, pruned_loss=0.1263, over 16506.00 frames. ], tot_loss[loss=0.3138, simple_loss=0.3633, pruned_loss=0.1322, over 1320731.08 frames. ], batch size: 146, lr: 3.25e-02, grad_scale: 4.0 2023-04-27 16:37:33,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5894, 4.4584, 4.1159, 3.7846, 4.3139, 2.2141, 4.0516, 4.3724], device='cuda:0'), covar=tensor([0.0148, 0.0092, 0.0109, 0.0354, 0.0074, 0.1239, 0.0099, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0049, 0.0043, 0.0062, 0.0076, 0.0044, 0.0099, 0.0058, 0.0059], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:38:13,350 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.80 vs. limit=2.0 2023-04-27 16:38:22,580 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10288.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:38:38,641 INFO [train.py:904] (0/8) Epoch 2, batch 150, loss[loss=0.3306, simple_loss=0.3731, pruned_loss=0.1441, over 16733.00 frames. ], tot_loss[loss=0.3053, simple_loss=0.3593, pruned_loss=0.1256, over 1762982.24 frames. ], batch size: 124, lr: 3.24e-02, grad_scale: 4.0 2023-04-27 16:38:56,019 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10314.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:39:01,500 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.010e+02 4.543e+02 5.695e+02 6.778e+02 1.541e+03, threshold=1.139e+03, percent-clipped=2.0 2023-04-27 16:39:47,458 INFO [train.py:904] (0/8) Epoch 2, batch 200, loss[loss=0.2275, simple_loss=0.3004, pruned_loss=0.07727, over 17030.00 frames. ], tot_loss[loss=0.3001, simple_loss=0.3565, pruned_loss=0.1218, over 2106359.07 frames. ], batch size: 41, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:39:57,735 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-27 16:40:02,885 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:47,540 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9281, 4.6199, 4.6436, 5.1825, 5.1738, 4.7622, 5.2763, 5.0538], device='cuda:0'), covar=tensor([0.0402, 0.0404, 0.1148, 0.0337, 0.0432, 0.0369, 0.0278, 0.0332], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0210, 0.0318, 0.0217, 0.0182, 0.0179, 0.0165, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:40:50,474 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10395.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:40:55,084 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10399.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:40:57,617 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-27 16:40:58,067 INFO [train.py:904] (0/8) Epoch 2, batch 250, loss[loss=0.3156, simple_loss=0.3539, pruned_loss=0.1387, over 16805.00 frames. ], tot_loss[loss=0.2969, simple_loss=0.3531, pruned_loss=0.1204, over 2375181.20 frames. ], batch size: 102, lr: 3.23e-02, grad_scale: 4.0 2023-04-27 16:41:10,388 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10409.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:41:22,655 INFO [optim.py:368] (0/8) 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 16:41:22,940 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10418.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:42:03,221 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10447.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:09,639 INFO [train.py:904] (0/8) Epoch 2, batch 300, loss[loss=0.2617, simple_loss=0.3206, pruned_loss=0.1014, over 16648.00 frames. ], tot_loss[loss=0.2915, simple_loss=0.3484, pruned_loss=0.1173, over 2590003.62 frames. ], batch size: 76, lr: 3.22e-02, grad_scale: 4.0 2023-04-27 16:42:14,818 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10454.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:27,478 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2336, 5.0109, 5.0630, 5.1509, 4.3684, 5.1275, 5.0806, 4.8215], device='cuda:0'), covar=tensor([0.0266, 0.0128, 0.0185, 0.0133, 0.0896, 0.0154, 0.0142, 0.0177], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0081, 0.0149, 0.0116, 0.0183, 0.0115, 0.0104, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 16:42:36,324 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10470.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:42:37,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9968, 3.4617, 3.7449, 2.8334, 3.8348, 3.6517, 3.8939, 2.1334], device='cuda:0'), covar=tensor([0.1018, 0.0153, 0.0071, 0.0427, 0.0052, 0.0092, 0.0055, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0051, 0.0057, 0.0093, 0.0052, 0.0054, 0.0057, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 16:42:39,493 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.86 vs. limit=2.0 2023-04-27 16:43:19,871 INFO [train.py:904] (0/8) Epoch 2, batch 350, loss[loss=0.2837, simple_loss=0.3423, pruned_loss=0.1126, over 17147.00 frames. ], tot_loss[loss=0.286, simple_loss=0.3439, pruned_loss=0.1141, over 2750769.86 frames. ], batch size: 47, lr: 3.21e-02, grad_scale: 4.0 2023-04-27 16:43:39,876 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10515.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:43:42,894 INFO [optim.py:368] (0/8) 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 16:44:13,996 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8407, 3.8050, 2.1708, 4.3604, 4.2513, 4.2127, 1.9666, 3.2033], device='cuda:0'), covar=tensor([0.1794, 0.0267, 0.1538, 0.0072, 0.0166, 0.0238, 0.1128, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0103, 0.0163, 0.0061, 0.0078, 0.0090, 0.0140, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:44:27,081 INFO [train.py:904] (0/8) Epoch 2, batch 400, loss[loss=0.2515, simple_loss=0.3209, pruned_loss=0.09101, over 16826.00 frames. ], tot_loss[loss=0.2826, simple_loss=0.3413, pruned_loss=0.112, over 2883761.64 frames. ], batch size: 42, lr: 3.21e-02, grad_scale: 8.0 2023-04-27 16:45:36,495 INFO [train.py:904] (0/8) Epoch 2, batch 450, loss[loss=0.2893, simple_loss=0.3422, pruned_loss=0.1182, over 12219.00 frames. ], tot_loss[loss=0.2812, simple_loss=0.3398, pruned_loss=0.1113, over 2958152.47 frames. ], batch size: 248, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:45:59,202 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-27 16:45:59,586 INFO [optim.py:368] (0/8) 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 16:46:44,107 INFO [train.py:904] (0/8) Epoch 2, batch 500, loss[loss=0.2689, simple_loss=0.3411, pruned_loss=0.09835, over 17116.00 frames. ], tot_loss[loss=0.278, simple_loss=0.3375, pruned_loss=0.1092, over 3042008.88 frames. ], batch size: 47, lr: 3.20e-02, grad_scale: 8.0 2023-04-27 16:47:39,715 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1061, 4.6235, 4.8358, 5.0374, 4.2463, 4.8300, 4.9121, 4.5167], device='cuda:0'), covar=tensor([0.0268, 0.0218, 0.0210, 0.0116, 0.1033, 0.0212, 0.0214, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0087, 0.0154, 0.0123, 0.0189, 0.0122, 0.0108, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 16:47:44,953 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10695.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:47:52,737 INFO [train.py:904] (0/8) Epoch 2, batch 550, loss[loss=0.255, simple_loss=0.3202, pruned_loss=0.09494, over 16834.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3361, pruned_loss=0.1081, over 3105965.18 frames. ], batch size: 42, lr: 3.19e-02, grad_scale: 8.0 2023-04-27 16:48:17,047 INFO [optim.py:368] (0/8) 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 16:48:17,345 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=10718.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:48:47,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6107, 4.3565, 4.1468, 1.5790, 4.4071, 4.3953, 3.6839, 3.7316], device='cuda:0'), covar=tensor([0.0894, 0.0113, 0.0261, 0.2234, 0.0148, 0.0145, 0.0311, 0.0292], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0075, 0.0074, 0.0152, 0.0067, 0.0065, 0.0087, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:48:51,488 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10743.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:52,719 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:48:57,877 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0721, 3.2374, 3.0677, 4.4178, 2.6233, 4.1357, 2.8187, 2.8298], device='cuda:0'), covar=tensor([0.0196, 0.0293, 0.0239, 0.0149, 0.1109, 0.0135, 0.0492, 0.0911], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0143, 0.0116, 0.0163, 0.0224, 0.0134, 0.0155, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:49:02,286 INFO [train.py:904] (0/8) Epoch 2, batch 600, loss[loss=0.2615, simple_loss=0.3102, pruned_loss=0.1064, over 16701.00 frames. ], tot_loss[loss=0.2762, simple_loss=0.3347, pruned_loss=0.1088, over 3153149.88 frames. ], batch size: 124, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:49:17,438 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1792, 4.1406, 3.9088, 3.5915, 4.1479, 1.7631, 3.9085, 3.9724], device='cuda:0'), covar=tensor([0.0083, 0.0062, 0.0092, 0.0311, 0.0062, 0.1252, 0.0078, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0056, 0.0048, 0.0068, 0.0089, 0.0050, 0.0100, 0.0065, 0.0067], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 16:49:21,922 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:49:23,783 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=10766.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:49:33,722 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:11,109 INFO [train.py:904] (0/8) Epoch 2, batch 650, loss[loss=0.2359, simple_loss=0.3102, pruned_loss=0.08082, over 17178.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3335, pruned_loss=0.1082, over 3186831.54 frames. ], batch size: 44, lr: 3.18e-02, grad_scale: 8.0 2023-04-27 16:50:16,387 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:50:23,986 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=10810.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:50:33,028 INFO [optim.py:368] (0/8) 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 16:50:57,055 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10835.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:51:19,285 INFO [train.py:904] (0/8) Epoch 2, batch 700, loss[loss=0.2386, simple_loss=0.3035, pruned_loss=0.08685, over 15909.00 frames. ], tot_loss[loss=0.2733, simple_loss=0.3326, pruned_loss=0.107, over 3205798.01 frames. ], batch size: 35, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:51:58,182 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-27 16:52:06,338 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=10885.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 16:52:26,676 INFO [train.py:904] (0/8) Epoch 2, batch 750, loss[loss=0.262, simple_loss=0.3377, pruned_loss=0.0932, over 17293.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3339, pruned_loss=0.1069, over 3223587.13 frames. ], batch size: 52, lr: 3.17e-02, grad_scale: 8.0 2023-04-27 16:52:50,210 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.205e+02 3.722e+02 4.836e+02 5.954e+02 8.013e+02, threshold=9.671e+02, percent-clipped=0.0 2023-04-27 16:53:27,438 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=10946.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 16:53:33,381 INFO [train.py:904] (0/8) Epoch 2, batch 800, loss[loss=0.2263, simple_loss=0.2967, pruned_loss=0.07799, over 17198.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3331, pruned_loss=0.1066, over 3249728.48 frames. ], batch size: 44, lr: 3.16e-02, grad_scale: 8.0 2023-04-27 16:54:05,433 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6162, 4.3186, 4.4335, 4.5473, 3.9348, 4.4272, 4.3030, 4.1785], device='cuda:0'), covar=tensor([0.0258, 0.0200, 0.0157, 0.0116, 0.0795, 0.0178, 0.0241, 0.0217], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0087, 0.0161, 0.0129, 0.0198, 0.0125, 0.0110, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 16:54:13,921 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.36 vs. limit=5.0 2023-04-27 16:54:20,199 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-04-27 16:54:43,065 INFO [train.py:904] (0/8) Epoch 2, batch 850, loss[loss=0.2586, simple_loss=0.3333, pruned_loss=0.0919, over 17257.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3322, pruned_loss=0.1054, over 3273397.59 frames. ], batch size: 52, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:55:06,017 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.379e+02 3.969e+02 4.868e+02 6.050e+02 1.097e+03, threshold=9.736e+02, percent-clipped=1.0 2023-04-27 16:55:49,295 INFO [train.py:904] (0/8) Epoch 2, batch 900, loss[loss=0.297, simple_loss=0.3396, pruned_loss=0.1272, over 16743.00 frames. ], tot_loss[loss=0.2694, simple_loss=0.3305, pruned_loss=0.1041, over 3289521.23 frames. ], batch size: 124, lr: 3.15e-02, grad_scale: 8.0 2023-04-27 16:56:07,927 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 16:56:10,037 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11065.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:21,393 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8520, 4.8537, 5.2638, 5.3011, 5.4093, 4.8316, 5.0446, 5.1882], device='cuda:0'), covar=tensor([0.0262, 0.0310, 0.0405, 0.0475, 0.0342, 0.0262, 0.0576, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0145, 0.0161, 0.0162, 0.0187, 0.0155, 0.0233, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 16:56:56,862 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11100.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:56:58,326 INFO [train.py:904] (0/8) Epoch 2, batch 950, loss[loss=0.2378, simple_loss=0.3111, pruned_loss=0.0822, over 17206.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3302, pruned_loss=0.1039, over 3294481.11 frames. ], batch size: 45, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:57:10,367 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11110.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 16:57:13,956 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:57:21,474 INFO [optim.py:368] (0/8) 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 16:57:38,007 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:06,962 INFO [train.py:904] (0/8) Epoch 2, batch 1000, loss[loss=0.2829, simple_loss=0.3251, pruned_loss=0.1203, over 16485.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3292, pruned_loss=0.1044, over 3289977.93 frames. ], batch size: 146, lr: 3.14e-02, grad_scale: 8.0 2023-04-27 16:58:16,904 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11158.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:58:37,541 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 16:59:13,988 INFO [train.py:904] (0/8) Epoch 2, batch 1050, loss[loss=0.2748, simple_loss=0.3275, pruned_loss=0.111, over 16702.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3293, pruned_loss=0.1039, over 3301830.17 frames. ], batch size: 134, lr: 3.13e-02, grad_scale: 8.0 2023-04-27 16:59:36,194 INFO [optim.py:368] (0/8) 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 17:00:00,130 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:00:03,969 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 17:00:08,445 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11241.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:00:21,903 INFO [train.py:904] (0/8) Epoch 2, batch 1100, loss[loss=0.2819, simple_loss=0.3497, pruned_loss=0.1071, over 17130.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.328, pruned_loss=0.1036, over 3300490.67 frames. ], batch size: 49, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:00:42,879 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-27 17:01:28,366 INFO [train.py:904] (0/8) Epoch 2, batch 1150, loss[loss=0.275, simple_loss=0.3283, pruned_loss=0.1109, over 16477.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3269, pruned_loss=0.102, over 3295818.07 frames. ], batch size: 75, lr: 3.12e-02, grad_scale: 8.0 2023-04-27 17:01:52,667 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.647e+02 4.017e+02 4.974e+02 5.619e+02 1.017e+03, threshold=9.949e+02, percent-clipped=1.0 2023-04-27 17:02:39,329 INFO [train.py:904] (0/8) Epoch 2, batch 1200, loss[loss=0.2508, simple_loss=0.3092, pruned_loss=0.09624, over 16845.00 frames. ], tot_loss[loss=0.2641, simple_loss=0.3263, pruned_loss=0.1009, over 3302561.62 frames. ], batch size: 96, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:03:24,750 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 17:03:31,011 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.81 vs. limit=5.0 2023-04-27 17:03:46,866 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11400.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:03:47,502 INFO [train.py:904] (0/8) Epoch 2, batch 1250, loss[loss=0.321, simple_loss=0.3831, pruned_loss=0.1294, over 17049.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3271, pruned_loss=0.1023, over 3298602.58 frames. ], batch size: 55, lr: 3.11e-02, grad_scale: 8.0 2023-04-27 17:03:49,548 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0404, 5.6756, 5.5454, 5.5580, 5.6369, 6.0255, 5.9597, 5.7022], device='cuda:0'), covar=tensor([0.0500, 0.0951, 0.0939, 0.1274, 0.1840, 0.0627, 0.0521, 0.1785], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0265, 0.0233, 0.0230, 0.0289, 0.0234, 0.0201, 0.0291], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:04:10,336 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.776e+02 4.393e+02 5.372e+02 6.605e+02 1.355e+03, threshold=1.074e+03, percent-clipped=5.0 2023-04-27 17:04:26,830 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11430.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:49,847 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11448.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:04:53,727 INFO [train.py:904] (0/8) Epoch 2, batch 1300, loss[loss=0.2571, simple_loss=0.3362, pruned_loss=0.08899, over 17125.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3267, pruned_loss=0.102, over 3293461.39 frames. ], batch size: 48, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:04:54,079 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6053, 4.6639, 5.1784, 5.2206, 5.2892, 4.8171, 4.8385, 5.0338], device='cuda:0'), covar=tensor([0.0296, 0.0317, 0.0357, 0.0354, 0.0349, 0.0293, 0.0636, 0.0226], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0143, 0.0164, 0.0158, 0.0190, 0.0153, 0.0231, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:05:13,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3987, 3.5363, 4.0526, 3.0956, 4.0141, 4.1001, 4.3647, 2.0599], device='cuda:0'), covar=tensor([0.0895, 0.0130, 0.0080, 0.0381, 0.0048, 0.0077, 0.0052, 0.0793], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0051, 0.0057, 0.0098, 0.0049, 0.0055, 0.0060, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002], device='cuda:0') 2023-04-27 17:05:30,716 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11478.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:05:30,900 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3148, 5.0686, 4.9937, 4.4634, 5.1187, 2.7223, 4.8795, 5.1767], device='cuda:0'), covar=tensor([0.0058, 0.0049, 0.0057, 0.0305, 0.0048, 0.0889, 0.0059, 0.0078], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0053, 0.0074, 0.0100, 0.0057, 0.0105, 0.0071, 0.0075], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 17:06:01,559 INFO [train.py:904] (0/8) Epoch 2, batch 1350, loss[loss=0.2527, simple_loss=0.3313, pruned_loss=0.08703, over 16697.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.3265, pruned_loss=0.1006, over 3305370.64 frames. ], batch size: 57, lr: 3.10e-02, grad_scale: 8.0 2023-04-27 17:06:04,645 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9454, 3.9055, 1.9867, 4.5742, 4.3839, 4.4044, 1.9264, 3.3138], device='cuda:0'), covar=tensor([0.1608, 0.0232, 0.1605, 0.0066, 0.0165, 0.0220, 0.1098, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0107, 0.0162, 0.0065, 0.0089, 0.0098, 0.0144, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 17:06:24,755 INFO [optim.py:368] (0/8) 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 17:06:34,982 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6860, 4.5090, 1.9950, 4.5995, 2.6416, 4.4383, 2.7229, 3.2463], device='cuda:0'), covar=tensor([0.0035, 0.0117, 0.1574, 0.0047, 0.0950, 0.0221, 0.1240, 0.0653], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0093, 0.0166, 0.0078, 0.0153, 0.0114, 0.0171, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 17:06:41,107 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:44,526 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11532.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:06:57,577 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11541.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:07:09,723 INFO [train.py:904] (0/8) Epoch 2, batch 1400, loss[loss=0.2332, simple_loss=0.3071, pruned_loss=0.07964, over 17125.00 frames. ], tot_loss[loss=0.2635, simple_loss=0.326, pruned_loss=0.1005, over 3304298.41 frames. ], batch size: 47, lr: 3.09e-02, grad_scale: 8.0 2023-04-27 17:07:19,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9603, 4.5587, 4.4539, 2.0466, 4.4881, 4.6157, 3.7677, 3.9498], device='cuda:0'), covar=tensor([0.0672, 0.0045, 0.0137, 0.1569, 0.0077, 0.0050, 0.0213, 0.0206], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0077, 0.0076, 0.0155, 0.0071, 0.0069, 0.0096, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0001, 0.0003, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 17:08:02,442 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 17:08:03,055 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11589.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:08:08,177 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:08:19,414 INFO [train.py:904] (0/8) Epoch 2, batch 1450, loss[loss=0.2284, simple_loss=0.2979, pruned_loss=0.0795, over 16966.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3263, pruned_loss=0.1006, over 3298768.49 frames. ], batch size: 41, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:08:43,797 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.475e+02 3.985e+02 4.876e+02 5.970e+02 8.943e+02, threshold=9.752e+02, percent-clipped=0.0 2023-04-27 17:09:05,832 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11635.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:09:26,116 INFO [train.py:904] (0/8) Epoch 2, batch 1500, loss[loss=0.3014, simple_loss=0.3431, pruned_loss=0.1299, over 16885.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3266, pruned_loss=0.1014, over 3299374.96 frames. ], batch size: 109, lr: 3.08e-02, grad_scale: 8.0 2023-04-27 17:10:11,975 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 17:10:29,358 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11696.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:33,687 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:10:36,152 INFO [train.py:904] (0/8) Epoch 2, batch 1550, loss[loss=0.2882, simple_loss=0.3568, pruned_loss=0.1098, over 17180.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3277, pruned_loss=0.103, over 3314282.03 frames. ], batch size: 46, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:10:58,965 INFO [optim.py:368] (0/8) 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 17:10:59,415 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8976, 4.6926, 4.6363, 4.1825, 4.7500, 2.1595, 4.3084, 4.7502], device='cuda:0'), covar=tensor([0.0097, 0.0079, 0.0081, 0.0342, 0.0063, 0.1200, 0.0099, 0.0119], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0054, 0.0075, 0.0100, 0.0058, 0.0104, 0.0072, 0.0076], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 17:11:44,748 INFO [train.py:904] (0/8) Epoch 2, batch 1600, loss[loss=0.2668, simple_loss=0.3374, pruned_loss=0.09805, over 17087.00 frames. ], tot_loss[loss=0.2696, simple_loss=0.3305, pruned_loss=0.1044, over 3317198.46 frames. ], batch size: 53, lr: 3.07e-02, grad_scale: 8.0 2023-04-27 17:11:57,762 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11760.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:12:53,459 INFO [train.py:904] (0/8) Epoch 2, batch 1650, loss[loss=0.278, simple_loss=0.3347, pruned_loss=0.1107, over 16504.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3317, pruned_loss=0.1043, over 3327109.06 frames. ], batch size: 68, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:12:56,211 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6185, 3.0157, 2.5967, 3.8949, 2.3752, 3.6608, 2.7216, 2.5059], device='cuda:0'), covar=tensor([0.0252, 0.0279, 0.0273, 0.0187, 0.1099, 0.0160, 0.0472, 0.0987], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0151, 0.0125, 0.0179, 0.0231, 0.0143, 0.0162, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:13:15,842 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:17,809 INFO [optim.py:368] (0/8) 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 17:13:35,509 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=11830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:13:56,917 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:03,725 INFO [train.py:904] (0/8) Epoch 2, batch 1700, loss[loss=0.2364, simple_loss=0.3019, pruned_loss=0.08546, over 16738.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.335, pruned_loss=0.1059, over 3323431.59 frames. ], batch size: 102, lr: 3.06e-02, grad_scale: 8.0 2023-04-27 17:14:40,633 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11877.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:41,666 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=11878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:14:50,177 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7494, 4.8015, 3.9554, 1.5105, 2.7866, 2.1539, 3.6311, 4.6739], device='cuda:0'), covar=tensor([0.0268, 0.0360, 0.0489, 0.2390, 0.1140, 0.1442, 0.0871, 0.0396], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0098, 0.0139, 0.0156, 0.0143, 0.0140, 0.0146, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:14:55,383 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11888.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:15:01,411 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 17:15:13,210 INFO [train.py:904] (0/8) Epoch 2, batch 1750, loss[loss=0.2924, simple_loss=0.3439, pruned_loss=0.1204, over 16691.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3353, pruned_loss=0.105, over 3324239.29 frames. ], batch size: 124, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:15:21,925 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 17:15:39,685 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11920.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:05,968 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2378, 4.0791, 3.6404, 1.7791, 2.8898, 2.2260, 3.7348, 4.1004], device='cuda:0'), covar=tensor([0.0188, 0.0279, 0.0309, 0.1590, 0.0658, 0.1033, 0.0477, 0.0311], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0095, 0.0135, 0.0150, 0.0138, 0.0135, 0.0142, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:16:17,836 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11948.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:16:20,553 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-27 17:16:21,490 INFO [train.py:904] (0/8) Epoch 2, batch 1800, loss[loss=0.2441, simple_loss=0.3203, pruned_loss=0.08398, over 17254.00 frames. ], tot_loss[loss=0.2717, simple_loss=0.3354, pruned_loss=0.104, over 3330868.50 frames. ], batch size: 45, lr: 3.05e-02, grad_scale: 8.0 2023-04-27 17:17:03,365 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=11981.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:03,762 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 17:17:04,410 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4658, 4.5791, 4.4557, 1.7537, 3.1165, 2.2660, 3.7819, 4.4718], device='cuda:0'), covar=tensor([0.0295, 0.0360, 0.0245, 0.1804, 0.0741, 0.1101, 0.0788, 0.0573], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0097, 0.0135, 0.0153, 0.0141, 0.0137, 0.0144, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:17:07,155 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=11984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:16,289 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=11991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:17:29,579 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-12000.pt 2023-04-27 17:17:33,776 INFO [train.py:904] (0/8) Epoch 2, batch 1850, loss[loss=0.2747, simple_loss=0.3399, pruned_loss=0.1047, over 16889.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3364, pruned_loss=0.1045, over 3330310.93 frames. ], batch size: 90, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:17:34,334 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1201, 2.0677, 2.0339, 1.7320, 2.7226, 2.5530, 3.4629, 3.2437], device='cuda:0'), covar=tensor([0.0024, 0.0166, 0.0153, 0.0218, 0.0078, 0.0135, 0.0045, 0.0049], device='cuda:0'), in_proj_covar=tensor([0.0044, 0.0085, 0.0078, 0.0086, 0.0075, 0.0084, 0.0049, 0.0059], device='cuda:0'), out_proj_covar=tensor([7.2220e-05, 1.3198e-04, 1.2087e-04, 1.3653e-04, 1.1987e-04, 1.3622e-04, 7.7806e-05, 1.0051e-04], device='cuda:0') 2023-04-27 17:17:44,923 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 17:18:00,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1791, 4.1520, 4.0628, 3.6869, 4.0503, 1.7819, 3.9408, 4.1125], device='cuda:0'), covar=tensor([0.0106, 0.0073, 0.0085, 0.0347, 0.0070, 0.1306, 0.0083, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0054, 0.0075, 0.0100, 0.0058, 0.0102, 0.0072, 0.0076], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:18:35,124 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12045.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:18:43,209 INFO [train.py:904] (0/8) Epoch 2, batch 1900, loss[loss=0.2909, simple_loss=0.3447, pruned_loss=0.1186, over 16698.00 frames. ], tot_loss[loss=0.2699, simple_loss=0.3343, pruned_loss=0.1028, over 3322699.03 frames. ], batch size: 134, lr: 3.04e-02, grad_scale: 8.0 2023-04-27 17:18:48,968 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12055.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:13,910 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6943, 4.5958, 4.3274, 1.7813, 3.2244, 2.7151, 4.1113, 4.7225], device='cuda:0'), covar=tensor([0.0293, 0.0469, 0.0351, 0.2016, 0.0887, 0.1136, 0.0739, 0.0388], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0101, 0.0138, 0.0156, 0.0144, 0.0139, 0.0149, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:19:17,254 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12076.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:19:45,402 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7632, 4.8068, 5.2807, 5.3317, 5.4007, 4.8940, 5.0011, 4.9945], device='cuda:0'), covar=tensor([0.0251, 0.0224, 0.0335, 0.0318, 0.0274, 0.0250, 0.0529, 0.0261], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0146, 0.0171, 0.0165, 0.0190, 0.0154, 0.0243, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:19:51,458 INFO [train.py:904] (0/8) Epoch 2, batch 1950, loss[loss=0.2313, simple_loss=0.2988, pruned_loss=0.08195, over 16961.00 frames. ], tot_loss[loss=0.2682, simple_loss=0.3333, pruned_loss=0.1015, over 3326480.60 frames. ], batch size: 41, lr: 3.03e-02, grad_scale: 8.0 2023-04-27 17:20:14,620 INFO [optim.py:368] (0/8) 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 17:20:18,123 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-04-27 17:20:38,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4199, 4.0323, 4.2373, 4.3981, 3.8700, 4.2401, 4.1903, 3.9948], device='cuda:0'), covar=tensor([0.0279, 0.0218, 0.0174, 0.0118, 0.0637, 0.0196, 0.0355, 0.0214], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0091, 0.0168, 0.0133, 0.0204, 0.0134, 0.0118, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:20:40,781 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12137.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:20:59,931 INFO [train.py:904] (0/8) Epoch 2, batch 2000, loss[loss=0.2719, simple_loss=0.3471, pruned_loss=0.09839, over 17221.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3334, pruned_loss=0.1016, over 3327652.22 frames. ], batch size: 46, lr: 3.02e-02, grad_scale: 8.0 2023-04-27 17:21:28,832 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12172.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:21:51,886 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12188.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:09,012 INFO [train.py:904] (0/8) Epoch 2, batch 2050, loss[loss=0.2873, simple_loss=0.3316, pruned_loss=0.1215, over 16800.00 frames. ], tot_loss[loss=0.2691, simple_loss=0.3341, pruned_loss=0.1021, over 3323566.51 frames. ], batch size: 124, lr: 3.02e-02, grad_scale: 16.0 2023-04-27 17:22:10,421 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12202.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:22:32,915 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.569e+02 4.228e+02 5.159e+02 6.169e+02 1.128e+03, threshold=1.032e+03, percent-clipped=1.0 2023-04-27 17:22:59,000 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:20,414 INFO [train.py:904] (0/8) Epoch 2, batch 2100, loss[loss=0.293, simple_loss=0.3503, pruned_loss=0.1178, over 16779.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3354, pruned_loss=0.1033, over 3330312.07 frames. ], batch size: 124, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:23:53,791 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:23:57,278 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-27 17:24:15,658 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12291.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:28,737 INFO [train.py:904] (0/8) Epoch 2, batch 2150, loss[loss=0.2847, simple_loss=0.3504, pruned_loss=0.1095, over 16500.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3371, pruned_loss=0.1045, over 3336692.02 frames. ], batch size: 68, lr: 3.01e-02, grad_scale: 16.0 2023-04-27 17:24:33,518 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12304.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:24:51,839 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.440e+02 4.527e+02 5.287e+02 6.148e+02 1.100e+03, threshold=1.057e+03, percent-clipped=4.0 2023-04-27 17:25:22,234 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12339.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:23,342 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12340.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:25:38,423 INFO [train.py:904] (0/8) Epoch 2, batch 2200, loss[loss=0.275, simple_loss=0.3248, pruned_loss=0.1126, over 16223.00 frames. ], tot_loss[loss=0.2739, simple_loss=0.3375, pruned_loss=0.1051, over 3327719.75 frames. ], batch size: 36, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:25:44,328 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12355.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:26:13,300 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 17:26:20,241 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9074, 1.8578, 1.4339, 1.6933, 2.3285, 2.3830, 2.3853, 2.4181], device='cuda:0'), covar=tensor([0.0053, 0.0169, 0.0144, 0.0142, 0.0074, 0.0106, 0.0067, 0.0081], device='cuda:0'), in_proj_covar=tensor([0.0044, 0.0087, 0.0080, 0.0084, 0.0076, 0.0088, 0.0049, 0.0061], device='cuda:0'), out_proj_covar=tensor([7.2413e-05, 1.3412e-04, 1.2366e-04, 1.3220e-04, 1.2124e-04, 1.4268e-04, 7.8612e-05, 1.0388e-04], device='cuda:0') 2023-04-27 17:26:48,779 INFO [train.py:904] (0/8) Epoch 2, batch 2250, loss[loss=0.3024, simple_loss=0.3576, pruned_loss=0.1236, over 16114.00 frames. ], tot_loss[loss=0.2741, simple_loss=0.3379, pruned_loss=0.1052, over 3329100.33 frames. ], batch size: 164, lr: 3.00e-02, grad_scale: 16.0 2023-04-27 17:26:51,420 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12403.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:12,083 INFO [optim.py:368] (0/8) 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 17:27:12,471 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4675, 4.4344, 4.1837, 3.8395, 4.3812, 2.0552, 4.2068, 4.4509], device='cuda:0'), covar=tensor([0.0092, 0.0063, 0.0085, 0.0306, 0.0059, 0.1128, 0.0071, 0.0078], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0056, 0.0079, 0.0104, 0.0060, 0.0105, 0.0075, 0.0080], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:27:32,296 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=12432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:50,203 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4840, 4.5448, 4.4977, 4.5084, 4.4669, 5.0414, 4.8808, 4.4877], device='cuda:0'), covar=tensor([0.0917, 0.1134, 0.0951, 0.1509, 0.2246, 0.0780, 0.0706, 0.1639], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0269, 0.0245, 0.0233, 0.0306, 0.0243, 0.0212, 0.0310], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:27:51,333 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:27:57,829 INFO [train.py:904] (0/8) Epoch 2, batch 2300, loss[loss=0.2788, simple_loss=0.333, pruned_loss=0.1123, over 16829.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3375, pruned_loss=0.1049, over 3314468.39 frames. ], batch size: 83, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:28:10,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7826, 3.3263, 2.1122, 4.0302, 3.9496, 3.8213, 1.5222, 2.9125], device='cuda:0'), covar=tensor([0.1642, 0.0347, 0.1551, 0.0082, 0.0175, 0.0328, 0.1285, 0.0595], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0109, 0.0166, 0.0069, 0.0096, 0.0102, 0.0147, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:28:26,568 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12472.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:06,517 INFO [train.py:904] (0/8) Epoch 2, batch 2350, loss[loss=0.2749, simple_loss=0.3577, pruned_loss=0.09599, over 17110.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3389, pruned_loss=0.1067, over 3302588.17 frames. ], batch size: 48, lr: 2.99e-02, grad_scale: 8.0 2023-04-27 17:29:07,835 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12502.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:15,009 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 17:29:33,876 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12520.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:29:58,595 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2554, 5.0673, 5.0236, 4.2934, 5.0800, 2.5904, 4.7075, 5.1712], device='cuda:0'), covar=tensor([0.0048, 0.0044, 0.0047, 0.0277, 0.0040, 0.0902, 0.0069, 0.0070], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0055, 0.0078, 0.0101, 0.0061, 0.0104, 0.0074, 0.0079], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:30:14,655 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12550.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:30:15,620 INFO [train.py:904] (0/8) Epoch 2, batch 2400, loss[loss=0.272, simple_loss=0.3482, pruned_loss=0.09794, over 15992.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.3405, pruned_loss=0.107, over 3304361.45 frames. ], batch size: 35, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:30:51,809 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12576.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:26,501 INFO [train.py:904] (0/8) Epoch 2, batch 2450, loss[loss=0.294, simple_loss=0.352, pruned_loss=0.118, over 16506.00 frames. ], tot_loss[loss=0.2751, simple_loss=0.3395, pruned_loss=0.1053, over 3319496.94 frames. ], batch size: 146, lr: 2.98e-02, grad_scale: 8.0 2023-04-27 17:31:31,746 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12604.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:31:35,562 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.93 vs. limit=2.0 2023-04-27 17:31:51,026 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.702e+02 4.143e+02 4.943e+02 6.147e+02 1.117e+03, threshold=9.887e+02, percent-clipped=0.0 2023-04-27 17:31:59,230 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12624.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:20,310 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12640.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:32:35,559 INFO [train.py:904] (0/8) Epoch 2, batch 2500, loss[loss=0.2326, simple_loss=0.3027, pruned_loss=0.08128, over 16809.00 frames. ], tot_loss[loss=0.2731, simple_loss=0.3377, pruned_loss=0.1042, over 3320601.41 frames. ], batch size: 39, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:32:36,843 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12652.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:26,476 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12688.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:33:43,145 INFO [train.py:904] (0/8) Epoch 2, batch 2550, loss[loss=0.2604, simple_loss=0.3419, pruned_loss=0.08947, over 17260.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.338, pruned_loss=0.1036, over 3322017.29 frames. ], batch size: 52, lr: 2.97e-02, grad_scale: 8.0 2023-04-27 17:34:03,953 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 17:34:08,164 INFO [optim.py:368] (0/8) 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 17:34:08,670 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1575, 3.3583, 3.2661, 1.5577, 3.4205, 3.4510, 3.0287, 2.8588], device='cuda:0'), covar=tensor([0.0777, 0.0086, 0.0167, 0.1590, 0.0105, 0.0067, 0.0262, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0078, 0.0081, 0.0150, 0.0073, 0.0069, 0.0097, 0.0109], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 17:34:24,310 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.72 vs. limit=2.0 2023-04-27 17:34:27,629 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=12732.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:34:53,766 INFO [train.py:904] (0/8) Epoch 2, batch 2600, loss[loss=0.2421, simple_loss=0.3135, pruned_loss=0.08538, over 16492.00 frames. ], tot_loss[loss=0.2714, simple_loss=0.3369, pruned_loss=0.103, over 3324934.77 frames. ], batch size: 75, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:35:08,871 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=12763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:20,189 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3092, 1.8056, 2.4152, 2.6990, 3.3307, 3.0118, 2.0174, 3.5068], device='cuda:0'), covar=tensor([0.0049, 0.0283, 0.0155, 0.0130, 0.0039, 0.0110, 0.0191, 0.0032], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0109, 0.0091, 0.0081, 0.0054, 0.0055, 0.0087, 0.0054], device='cuda:0'), out_proj_covar=tensor([1.3150e-04, 1.9735e-04, 1.7435e-04, 1.5584e-04, 9.7988e-05, 1.0559e-04, 1.5139e-04, 1.0219e-04], device='cuda:0') 2023-04-27 17:35:22,201 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-27 17:35:32,887 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=12780.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:35:40,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4659, 4.1902, 4.4113, 4.7814, 4.8430, 4.4861, 4.7855, 4.7482], device='cuda:0'), covar=tensor([0.0408, 0.0482, 0.0940, 0.0287, 0.0355, 0.0423, 0.0317, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0269, 0.0389, 0.0283, 0.0219, 0.0218, 0.0203, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:36:01,522 INFO [train.py:904] (0/8) Epoch 2, batch 2650, loss[loss=0.2867, simple_loss=0.3492, pruned_loss=0.1121, over 16794.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3383, pruned_loss=0.1035, over 3324568.46 frames. ], batch size: 83, lr: 2.96e-02, grad_scale: 8.0 2023-04-27 17:36:03,319 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.569e+02 4.221e+02 4.762e+02 5.639e+02 1.002e+03, threshold=9.524e+02, percent-clipped=0.0 2023-04-27 17:36:34,007 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=12824.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:36:35,078 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5217, 4.2869, 4.4684, 4.8563, 4.9114, 4.4363, 4.8600, 4.7961], device='cuda:0'), covar=tensor([0.0515, 0.0475, 0.1113, 0.0333, 0.0465, 0.0641, 0.0331, 0.0331], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0272, 0.0397, 0.0286, 0.0225, 0.0222, 0.0207, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:36:36,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2814, 4.2887, 2.8113, 5.3049, 5.2183, 4.6152, 3.0753, 3.9564], device='cuda:0'), covar=tensor([0.1409, 0.0295, 0.1316, 0.0098, 0.0183, 0.0281, 0.0795, 0.0409], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0110, 0.0165, 0.0069, 0.0100, 0.0107, 0.0147, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-27 17:37:09,368 INFO [train.py:904] (0/8) Epoch 2, batch 2700, loss[loss=0.2369, simple_loss=0.3051, pruned_loss=0.08432, over 16820.00 frames. ], tot_loss[loss=0.2713, simple_loss=0.3382, pruned_loss=0.1023, over 3328717.91 frames. ], batch size: 39, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:19,510 INFO [train.py:904] (0/8) Epoch 2, batch 2750, loss[loss=0.2557, simple_loss=0.3236, pruned_loss=0.09393, over 16801.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3373, pruned_loss=0.1007, over 3332817.45 frames. ], batch size: 83, lr: 2.95e-02, grad_scale: 8.0 2023-04-27 17:38:41,653 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 2, batch 2800, loss[loss=0.2599, simple_loss=0.3353, pruned_loss=0.0922, over 17250.00 frames. ], tot_loss[loss=0.2694, simple_loss=0.3375, pruned_loss=0.1007, over 3331168.55 frames. ], batch size: 52, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:39:48,227 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4169, 4.6030, 4.3956, 1.7612, 3.1566, 2.3190, 3.6034, 4.3666], device='cuda:0'), covar=tensor([0.0386, 0.0401, 0.0313, 0.2088, 0.0861, 0.1415, 0.0925, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0108, 0.0143, 0.0157, 0.0148, 0.0140, 0.0151, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 17:40:33,621 INFO [train.py:904] (0/8) Epoch 2, batch 2850, loss[loss=0.26, simple_loss=0.3321, pruned_loss=0.09393, over 17236.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3367, pruned_loss=0.1007, over 3331979.35 frames. ], batch size: 45, lr: 2.94e-02, grad_scale: 8.0 2023-04-27 17:40:57,337 INFO [optim.py:368] (0/8) 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 17:41:21,805 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-27 17:41:30,576 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 17:41:41,134 INFO [train.py:904] (0/8) Epoch 2, batch 2900, loss[loss=0.2841, simple_loss=0.3476, pruned_loss=0.1103, over 16730.00 frames. ], tot_loss[loss=0.2681, simple_loss=0.335, pruned_loss=0.1005, over 3329420.68 frames. ], batch size: 62, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:42:08,640 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-04-27 17:42:49,016 INFO [train.py:904] (0/8) Epoch 2, batch 2950, loss[loss=0.2738, simple_loss=0.3413, pruned_loss=0.1031, over 16584.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3346, pruned_loss=0.1016, over 3333241.94 frames. ], batch size: 68, lr: 2.93e-02, grad_scale: 8.0 2023-04-27 17:42:50,402 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:13,804 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.070e+02 4.352e+02 4.979e+02 5.894e+02 1.092e+03, threshold=9.958e+02, percent-clipped=0.0 2023-04-27 17:43:14,149 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=13119.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 17:43:33,916 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-27 17:43:53,664 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:43:54,570 INFO [train.py:904] (0/8) Epoch 2, batch 3000, loss[loss=0.2886, simple_loss=0.3484, pruned_loss=0.1144, over 16509.00 frames. ], tot_loss[loss=0.27, simple_loss=0.3352, pruned_loss=0.1023, over 3333238.92 frames. ], batch size: 68, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:43:54,570 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 17:44:03,916 INFO [train.py:938] (0/8) Epoch 2, validation: loss=0.1858, simple_loss=0.2917, pruned_loss=0.04, over 944034.00 frames. 2023-04-27 17:44:03,917 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 17:44:08,771 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 17:45:09,407 INFO [train.py:904] (0/8) Epoch 2, batch 3050, loss[loss=0.2673, simple_loss=0.3398, pruned_loss=0.09739, over 17022.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3351, pruned_loss=0.1026, over 3330338.92 frames. ], batch size: 50, lr: 2.92e-02, grad_scale: 8.0 2023-04-27 17:45:29,283 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4129, 2.1838, 2.0620, 1.9574, 2.5930, 2.7116, 3.0711, 2.9805], device='cuda:0'), covar=tensor([0.0067, 0.0126, 0.0114, 0.0149, 0.0067, 0.0103, 0.0051, 0.0047], device='cuda:0'), in_proj_covar=tensor([0.0045, 0.0088, 0.0081, 0.0085, 0.0076, 0.0087, 0.0054, 0.0064], device='cuda:0'), out_proj_covar=tensor([7.6979e-05, 1.3759e-04, 1.2556e-04, 1.3496e-04, 1.2537e-04, 1.4217e-04, 8.7288e-05, 1.0901e-04], device='cuda:0') 2023-04-27 17:45:33,142 INFO [optim.py:368] (0/8) 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 17:46:04,750 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7958, 5.6148, 5.4531, 5.5097, 5.5234, 6.0115, 5.7580, 5.4427], device='cuda:0'), covar=tensor([0.0623, 0.0961, 0.0936, 0.1242, 0.2009, 0.0630, 0.0793, 0.1921], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0277, 0.0248, 0.0238, 0.0314, 0.0250, 0.0213, 0.0319], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:46:15,668 INFO [train.py:904] (0/8) Epoch 2, batch 3100, loss[loss=0.2489, simple_loss=0.3073, pruned_loss=0.09528, over 16891.00 frames. ], tot_loss[loss=0.2701, simple_loss=0.3344, pruned_loss=0.1029, over 3316304.92 frames. ], batch size: 96, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:47:22,161 INFO [train.py:904] (0/8) Epoch 2, batch 3150, loss[loss=0.2926, simple_loss=0.3433, pruned_loss=0.1209, over 16886.00 frames. ], tot_loss[loss=0.2685, simple_loss=0.3335, pruned_loss=0.1018, over 3323452.91 frames. ], batch size: 96, lr: 2.91e-02, grad_scale: 8.0 2023-04-27 17:47:44,594 INFO [optim.py:368] (0/8) 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 17:47:47,777 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7638, 4.5081, 4.7212, 5.1249, 5.1650, 4.6491, 5.1254, 5.0870], device='cuda:0'), covar=tensor([0.0419, 0.0396, 0.0956, 0.0265, 0.0293, 0.0334, 0.0280, 0.0254], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0285, 0.0408, 0.0296, 0.0230, 0.0224, 0.0219, 0.0233], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0003, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:48:28,285 INFO [train.py:904] (0/8) Epoch 2, batch 3200, loss[loss=0.2677, simple_loss=0.3424, pruned_loss=0.09649, over 17141.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3333, pruned_loss=0.1022, over 3320259.80 frames. ], batch size: 49, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:03,950 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5844, 5.1681, 5.2394, 5.3731, 4.7386, 5.2989, 5.2247, 4.9224], device='cuda:0'), covar=tensor([0.0189, 0.0161, 0.0154, 0.0097, 0.0705, 0.0143, 0.0108, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0094, 0.0171, 0.0135, 0.0206, 0.0142, 0.0119, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 17:49:34,225 INFO [train.py:904] (0/8) Epoch 2, batch 3250, loss[loss=0.3098, simple_loss=0.3571, pruned_loss=0.1312, over 16878.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3335, pruned_loss=0.1028, over 3321303.82 frames. ], batch size: 116, lr: 2.90e-02, grad_scale: 8.0 2023-04-27 17:49:58,159 INFO [optim.py:368] (0/8) 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 17:49:58,589 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=13419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:50:42,675 INFO [train.py:904] (0/8) Epoch 2, batch 3300, loss[loss=0.268, simple_loss=0.3315, pruned_loss=0.1022, over 16824.00 frames. ], tot_loss[loss=0.2712, simple_loss=0.3349, pruned_loss=0.1038, over 3328856.86 frames. ], batch size: 83, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:51:02,649 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=13467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:51:48,093 INFO [train.py:904] (0/8) Epoch 2, batch 3350, loss[loss=0.2601, simple_loss=0.3228, pruned_loss=0.09875, over 16838.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3357, pruned_loss=0.1036, over 3307540.06 frames. ], batch size: 96, lr: 2.89e-02, grad_scale: 8.0 2023-04-27 17:52:13,284 INFO [optim.py:368] (0/8) 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 17:52:36,873 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8820, 5.7128, 5.5011, 5.5644, 5.6126, 6.0519, 5.8388, 5.6320], device='cuda:0'), covar=tensor([0.0624, 0.1043, 0.0946, 0.1162, 0.1922, 0.0680, 0.0678, 0.1422], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0287, 0.0260, 0.0244, 0.0324, 0.0269, 0.0215, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:52:43,031 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-27 17:52:56,206 INFO [train.py:904] (0/8) Epoch 2, batch 3400, loss[loss=0.23, simple_loss=0.3051, pruned_loss=0.07744, over 16812.00 frames. ], tot_loss[loss=0.2677, simple_loss=0.3325, pruned_loss=0.1015, over 3305623.33 frames. ], batch size: 42, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:05,275 INFO [train.py:904] (0/8) Epoch 2, batch 3450, loss[loss=0.2433, simple_loss=0.3161, pruned_loss=0.08521, over 16989.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3318, pruned_loss=0.1008, over 3316635.70 frames. ], batch size: 41, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:54:28,146 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 17:54:29,797 INFO [optim.py:368] (0/8) 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 17:55:11,523 INFO [train.py:904] (0/8) Epoch 2, batch 3500, loss[loss=0.2481, simple_loss=0.3237, pruned_loss=0.08622, over 17185.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3304, pruned_loss=0.09992, over 3318907.83 frames. ], batch size: 46, lr: 2.88e-02, grad_scale: 8.0 2023-04-27 17:56:21,459 INFO [train.py:904] (0/8) Epoch 2, batch 3550, loss[loss=0.2386, simple_loss=0.3096, pruned_loss=0.08383, over 16670.00 frames. ], tot_loss[loss=0.2647, simple_loss=0.3301, pruned_loss=0.09962, over 3309591.62 frames. ], batch size: 37, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:56:45,025 INFO [optim.py:368] (0/8) 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 17:57:10,804 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=13737.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 17:57:18,524 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6530, 4.7455, 4.7417, 4.7474, 4.6224, 5.2283, 5.0995, 4.6552], device='cuda:0'), covar=tensor([0.0772, 0.1159, 0.1008, 0.1549, 0.2489, 0.0767, 0.0763, 0.1764], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0284, 0.0258, 0.0240, 0.0316, 0.0262, 0.0208, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 17:57:29,800 INFO [train.py:904] (0/8) Epoch 2, batch 3600, loss[loss=0.2782, simple_loss=0.3223, pruned_loss=0.117, over 16822.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3275, pruned_loss=0.09873, over 3317916.29 frames. ], batch size: 83, lr: 2.87e-02, grad_scale: 8.0 2023-04-27 17:58:37,313 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=13798.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 17:58:41,001 INFO [train.py:904] (0/8) Epoch 2, batch 3650, loss[loss=0.2736, simple_loss=0.3166, pruned_loss=0.1153, over 16813.00 frames. ], tot_loss[loss=0.2616, simple_loss=0.3262, pruned_loss=0.09852, over 3319306.19 frames. ], batch size: 90, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 17:59:08,122 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.959e+02 4.047e+02 4.833e+02 5.657e+02 1.025e+03, threshold=9.667e+02, percent-clipped=2.0 2023-04-27 17:59:54,883 INFO [train.py:904] (0/8) Epoch 2, batch 3700, loss[loss=0.2745, simple_loss=0.3305, pruned_loss=0.1093, over 16232.00 frames. ], tot_loss[loss=0.2622, simple_loss=0.3246, pruned_loss=0.09984, over 3281467.71 frames. ], batch size: 165, lr: 2.86e-02, grad_scale: 8.0 2023-04-27 18:00:37,085 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3258, 1.7319, 2.1401, 2.1590, 2.5322, 2.1932, 1.6352, 2.3384], device='cuda:0'), covar=tensor([0.0047, 0.0206, 0.0111, 0.0122, 0.0041, 0.0114, 0.0180, 0.0043], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0110, 0.0092, 0.0083, 0.0057, 0.0059, 0.0092, 0.0055], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-27 18:00:54,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3575, 4.3178, 3.9823, 2.0082, 2.9810, 2.3222, 3.6608, 4.3608], device='cuda:0'), covar=tensor([0.0264, 0.0382, 0.0310, 0.1623, 0.0722, 0.1030, 0.0616, 0.0293], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0108, 0.0144, 0.0154, 0.0147, 0.0139, 0.0148, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 18:01:07,723 INFO [train.py:904] (0/8) Epoch 2, batch 3750, loss[loss=0.2731, simple_loss=0.3216, pruned_loss=0.1123, over 16714.00 frames. ], tot_loss[loss=0.2655, simple_loss=0.3261, pruned_loss=0.1024, over 3259682.84 frames. ], batch size: 83, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:01:33,976 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.812e+02 4.240e+02 4.945e+02 6.119e+02 1.020e+03, threshold=9.889e+02, percent-clipped=4.0 2023-04-27 18:01:51,329 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-27 18:02:24,609 INFO [train.py:904] (0/8) Epoch 2, batch 3800, loss[loss=0.2919, simple_loss=0.3596, pruned_loss=0.1121, over 16996.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3264, pruned_loss=0.1036, over 3268959.35 frames. ], batch size: 41, lr: 2.85e-02, grad_scale: 8.0 2023-04-27 18:02:53,121 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6875, 2.7704, 1.6100, 2.7010, 2.0912, 2.7996, 1.9233, 2.1639], device='cuda:0'), covar=tensor([0.0081, 0.0181, 0.1502, 0.0096, 0.0673, 0.0325, 0.1293, 0.0583], device='cuda:0'), in_proj_covar=tensor([0.0078, 0.0099, 0.0169, 0.0078, 0.0151, 0.0127, 0.0171, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 18:03:36,073 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-14000.pt 2023-04-27 18:03:40,205 INFO [train.py:904] (0/8) Epoch 2, batch 3850, loss[loss=0.2637, simple_loss=0.3194, pruned_loss=0.104, over 16447.00 frames. ], tot_loss[loss=0.2674, simple_loss=0.3265, pruned_loss=0.1042, over 3270479.35 frames. ], batch size: 165, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:04:07,045 INFO [optim.py:368] (0/8) 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 18:04:42,630 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 18:04:53,523 INFO [train.py:904] (0/8) Epoch 2, batch 3900, loss[loss=0.2701, simple_loss=0.3188, pruned_loss=0.1107, over 16722.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3252, pruned_loss=0.1042, over 3269223.95 frames. ], batch size: 83, lr: 2.84e-02, grad_scale: 8.0 2023-04-27 18:05:25,177 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14072.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:05:55,293 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14093.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:06:05,416 INFO [train.py:904] (0/8) Epoch 2, batch 3950, loss[loss=0.2691, simple_loss=0.32, pruned_loss=0.1091, over 16886.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3231, pruned_loss=0.1037, over 3285698.33 frames. ], batch size: 116, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:06:08,814 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14103.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:06:31,643 INFO [optim.py:368] (0/8) 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 18:06:34,029 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14120.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:06:52,460 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14133.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:07:17,638 INFO [train.py:904] (0/8) Epoch 2, batch 4000, loss[loss=0.2385, simple_loss=0.3065, pruned_loss=0.08525, over 17127.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3232, pruned_loss=0.104, over 3285734.06 frames. ], batch size: 47, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:07:37,740 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14164.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:08:02,241 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14181.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 18:08:10,202 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5209, 3.7927, 3.9836, 3.8882, 4.0639, 3.6850, 3.3502, 3.8913], device='cuda:0'), covar=tensor([0.0527, 0.0481, 0.0522, 0.0702, 0.0667, 0.0480, 0.1207, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0155, 0.0174, 0.0171, 0.0205, 0.0166, 0.0256, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 18:08:21,517 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6180, 3.4213, 3.6062, 3.8864, 3.9343, 3.6150, 3.8563, 3.8693], device='cuda:0'), covar=tensor([0.0475, 0.0537, 0.1100, 0.0394, 0.0389, 0.0914, 0.0407, 0.0347], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0267, 0.0377, 0.0277, 0.0225, 0.0214, 0.0212, 0.0216], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:08:31,023 INFO [train.py:904] (0/8) Epoch 2, batch 4050, loss[loss=0.2122, simple_loss=0.2911, pruned_loss=0.06662, over 16428.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3213, pruned_loss=0.1005, over 3281609.77 frames. ], batch size: 35, lr: 2.83e-02, grad_scale: 8.0 2023-04-27 18:08:56,317 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.111e+02 3.521e+02 4.461e+02 5.736e+02 1.012e+03, threshold=8.922e+02, percent-clipped=2.0 2023-04-27 18:09:43,229 INFO [train.py:904] (0/8) Epoch 2, batch 4100, loss[loss=0.2724, simple_loss=0.35, pruned_loss=0.0974, over 16712.00 frames. ], tot_loss[loss=0.2588, simple_loss=0.3212, pruned_loss=0.09815, over 3276160.28 frames. ], batch size: 89, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:10:59,428 INFO [train.py:904] (0/8) Epoch 2, batch 4150, loss[loss=0.3226, simple_loss=0.3867, pruned_loss=0.1293, over 16448.00 frames. ], tot_loss[loss=0.2689, simple_loss=0.3311, pruned_loss=0.1034, over 3233479.42 frames. ], batch size: 146, lr: 2.82e-02, grad_scale: 8.0 2023-04-27 18:11:25,251 INFO [optim.py:368] (0/8) 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 18:11:50,703 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 18:12:14,217 INFO [train.py:904] (0/8) Epoch 2, batch 4200, loss[loss=0.3391, simple_loss=0.3838, pruned_loss=0.1472, over 11301.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3391, pruned_loss=0.1062, over 3211706.73 frames. ], batch size: 250, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:12:18,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8759, 4.0545, 3.1263, 3.2103, 3.2313, 2.1841, 4.2355, 4.9765], device='cuda:0'), covar=tensor([0.1795, 0.0559, 0.1034, 0.0535, 0.1800, 0.1419, 0.0305, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0220, 0.0233, 0.0164, 0.0268, 0.0180, 0.0188, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:12:43,844 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6814, 2.6824, 1.7006, 2.6633, 2.0510, 2.6817, 1.8368, 2.2889], device='cuda:0'), covar=tensor([0.0064, 0.0198, 0.1102, 0.0073, 0.0636, 0.0280, 0.1037, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0094, 0.0161, 0.0076, 0.0149, 0.0119, 0.0170, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 18:13:15,572 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:26,255 INFO [train.py:904] (0/8) Epoch 2, batch 4250, loss[loss=0.2424, simple_loss=0.3296, pruned_loss=0.07766, over 16744.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3413, pruned_loss=0.1054, over 3213787.37 frames. ], batch size: 89, lr: 2.81e-02, grad_scale: 8.0 2023-04-27 18:13:34,207 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 18:13:44,103 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14413.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:13:52,843 INFO [optim.py:368] (0/8) 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 18:14:05,661 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14428.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:20,727 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2855, 4.1528, 3.8491, 1.7274, 2.9028, 2.0254, 3.5067, 4.3445], device='cuda:0'), covar=tensor([0.0295, 0.0462, 0.0370, 0.1820, 0.0772, 0.1114, 0.0773, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0102, 0.0144, 0.0155, 0.0145, 0.0135, 0.0149, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 18:14:24,082 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:14:38,559 INFO [train.py:904] (0/8) Epoch 2, batch 4300, loss[loss=0.2852, simple_loss=0.3535, pruned_loss=0.1085, over 17038.00 frames. ], tot_loss[loss=0.2743, simple_loss=0.3416, pruned_loss=0.1035, over 3223595.54 frames. ], batch size: 53, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:14:49,310 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14459.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:10,473 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:12,696 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14476.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:15:29,902 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14487.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:15:49,389 INFO [train.py:904] (0/8) Epoch 2, batch 4350, loss[loss=0.2872, simple_loss=0.3563, pruned_loss=0.1091, over 17200.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.3457, pruned_loss=0.1054, over 3209714.20 frames. ], batch size: 44, lr: 2.80e-02, grad_scale: 16.0 2023-04-27 18:16:08,791 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3601, 4.0673, 4.2014, 4.2325, 3.7577, 4.3534, 4.1238, 3.8774], device='cuda:0'), covar=tensor([0.0207, 0.0186, 0.0142, 0.0113, 0.0616, 0.0112, 0.0217, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0080, 0.0144, 0.0113, 0.0173, 0.0118, 0.0102, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:16:10,952 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 18:16:13,772 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.262e+02 4.011e+02 4.650e+02 6.201e+02 1.293e+03, threshold=9.301e+02, percent-clipped=5.0 2023-04-27 18:16:56,246 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14548.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:16:59,405 INFO [train.py:904] (0/8) Epoch 2, batch 4400, loss[loss=0.2419, simple_loss=0.3199, pruned_loss=0.08197, over 16678.00 frames. ], tot_loss[loss=0.2803, simple_loss=0.3478, pruned_loss=0.1064, over 3192539.89 frames. ], batch size: 134, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:17:53,691 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8968, 4.7099, 4.5029, 3.9912, 4.8204, 2.2086, 4.4189, 4.6861], device='cuda:0'), covar=tensor([0.0041, 0.0040, 0.0058, 0.0269, 0.0028, 0.1038, 0.0054, 0.0068], device='cuda:0'), in_proj_covar=tensor([0.0062, 0.0052, 0.0075, 0.0099, 0.0057, 0.0106, 0.0071, 0.0075], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:18:09,490 INFO [train.py:904] (0/8) Epoch 2, batch 4450, loss[loss=0.2704, simple_loss=0.3469, pruned_loss=0.09696, over 16832.00 frames. ], tot_loss[loss=0.2811, simple_loss=0.35, pruned_loss=0.1061, over 3196866.49 frames. ], batch size: 102, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:18:34,424 INFO [optim.py:368] (0/8) 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 18:19:03,990 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-27 18:19:17,773 INFO [train.py:904] (0/8) Epoch 2, batch 4500, loss[loss=0.2626, simple_loss=0.3303, pruned_loss=0.09743, over 16762.00 frames. ], tot_loss[loss=0.2793, simple_loss=0.3491, pruned_loss=0.1048, over 3198216.83 frames. ], batch size: 83, lr: 2.79e-02, grad_scale: 16.0 2023-04-27 18:19:20,373 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1656, 3.5211, 2.7488, 4.8663, 4.6582, 4.1722, 2.1259, 3.4396], device='cuda:0'), covar=tensor([0.1295, 0.0334, 0.1112, 0.0027, 0.0084, 0.0205, 0.1128, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0115, 0.0165, 0.0064, 0.0095, 0.0105, 0.0151, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-27 18:19:53,985 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 18:20:10,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7814, 3.8135, 3.7963, 1.4280, 4.0646, 3.9368, 3.1281, 3.1549], device='cuda:0'), covar=tensor([0.1142, 0.0089, 0.0148, 0.1760, 0.0050, 0.0046, 0.0290, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0078, 0.0078, 0.0149, 0.0072, 0.0072, 0.0099, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:20:28,462 INFO [train.py:904] (0/8) Epoch 2, batch 4550, loss[loss=0.322, simple_loss=0.3922, pruned_loss=0.1259, over 16821.00 frames. ], tot_loss[loss=0.2791, simple_loss=0.3493, pruned_loss=0.1045, over 3217978.21 frames. ], batch size: 42, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:20:54,777 INFO [optim.py:368] (0/8) 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 18:21:08,327 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14728.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:25,051 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=14740.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:21:43,184 INFO [train.py:904] (0/8) Epoch 2, batch 4600, loss[loss=0.2645, simple_loss=0.3427, pruned_loss=0.09309, over 17026.00 frames. ], tot_loss[loss=0.2775, simple_loss=0.3487, pruned_loss=0.1031, over 3233037.66 frames. ], batch size: 41, lr: 2.78e-02, grad_scale: 16.0 2023-04-27 18:21:53,278 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14759.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:09,285 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8459, 4.7831, 4.4496, 3.3699, 4.9773, 1.6254, 4.5924, 4.7262], device='cuda:0'), covar=tensor([0.0093, 0.0052, 0.0082, 0.0487, 0.0045, 0.1588, 0.0065, 0.0137], device='cuda:0'), in_proj_covar=tensor([0.0060, 0.0049, 0.0072, 0.0096, 0.0056, 0.0103, 0.0069, 0.0073], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:22:10,143 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14769.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,545 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14776.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:22:18,686 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=14776.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:22:27,187 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3588, 2.7459, 2.3379, 3.8412, 2.1275, 3.6121, 2.4638, 2.3917], device='cuda:0'), covar=tensor([0.0242, 0.0328, 0.0297, 0.0138, 0.1205, 0.0141, 0.0523, 0.0985], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0167, 0.0141, 0.0195, 0.0254, 0.0154, 0.0175, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 18:22:54,490 INFO [train.py:904] (0/8) Epoch 2, batch 4650, loss[loss=0.2684, simple_loss=0.3491, pruned_loss=0.09385, over 16215.00 frames. ], tot_loss[loss=0.276, simple_loss=0.3471, pruned_loss=0.1024, over 3226148.52 frames. ], batch size: 165, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:22:55,056 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=14801.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:03,207 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14807.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:23:04,653 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6174, 3.3380, 3.3194, 2.2658, 3.2325, 3.1951, 3.4590, 1.3008], device='cuda:0'), covar=tensor([0.0669, 0.0041, 0.0059, 0.0380, 0.0040, 0.0081, 0.0043, 0.0746], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0049, 0.0057, 0.0106, 0.0052, 0.0055, 0.0057, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 18:23:16,627 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5128, 3.3356, 2.8147, 1.5759, 2.6022, 1.9649, 3.0444, 3.2762], device='cuda:0'), covar=tensor([0.0282, 0.0382, 0.0511, 0.1913, 0.0810, 0.1147, 0.0701, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0104, 0.0147, 0.0154, 0.0145, 0.0136, 0.0148, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 18:23:20,767 INFO [optim.py:368] (0/8) 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 18:23:27,864 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=14824.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:23:54,904 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=14843.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:24:04,223 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.4753, 3.1143, 3.1261, 2.1546, 2.7427, 2.9344, 3.1080, 1.4705], device='cuda:0'), covar=tensor([0.0579, 0.0039, 0.0047, 0.0311, 0.0061, 0.0088, 0.0044, 0.0548], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0050, 0.0057, 0.0105, 0.0053, 0.0056, 0.0057, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 18:24:06,750 INFO [train.py:904] (0/8) Epoch 2, batch 4700, loss[loss=0.2677, simple_loss=0.3306, pruned_loss=0.1024, over 16557.00 frames. ], tot_loss[loss=0.2734, simple_loss=0.3444, pruned_loss=0.1012, over 3234044.65 frames. ], batch size: 68, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:24:17,921 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2663, 2.6067, 2.1663, 3.4934, 1.8827, 3.3244, 2.1739, 2.1973], device='cuda:0'), covar=tensor([0.0323, 0.0467, 0.0374, 0.0263, 0.1644, 0.0234, 0.0764, 0.1294], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0171, 0.0143, 0.0200, 0.0255, 0.0157, 0.0176, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 18:24:50,883 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-04-27 18:25:20,222 INFO [train.py:904] (0/8) Epoch 2, batch 4750, loss[loss=0.2744, simple_loss=0.3389, pruned_loss=0.1049, over 16579.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3405, pruned_loss=0.09924, over 3236822.13 frames. ], batch size: 75, lr: 2.77e-02, grad_scale: 16.0 2023-04-27 18:25:45,829 INFO [optim.py:368] (0/8) 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 18:26:30,657 INFO [train.py:904] (0/8) Epoch 2, batch 4800, loss[loss=0.2512, simple_loss=0.3321, pruned_loss=0.08518, over 16703.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.337, pruned_loss=0.09727, over 3222778.75 frames. ], batch size: 134, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:27:43,246 INFO [train.py:904] (0/8) Epoch 2, batch 4850, loss[loss=0.2524, simple_loss=0.3258, pruned_loss=0.08947, over 16591.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3387, pruned_loss=0.0973, over 3220315.52 frames. ], batch size: 57, lr: 2.76e-02, grad_scale: 16.0 2023-04-27 18:28:10,728 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2023-04-27 18:28:11,116 INFO [optim.py:368] (0/8) 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 18:28:33,245 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-27 18:28:58,608 INFO [train.py:904] (0/8) Epoch 2, batch 4900, loss[loss=0.2979, simple_loss=0.3573, pruned_loss=0.1192, over 12517.00 frames. ], tot_loss[loss=0.2656, simple_loss=0.3383, pruned_loss=0.09643, over 3196399.34 frames. ], batch size: 248, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:29:24,590 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15069.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:29:59,511 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.73 vs. limit=2.0 2023-04-27 18:30:04,910 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15096.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:10,702 INFO [train.py:904] (0/8) Epoch 2, batch 4950, loss[loss=0.2623, simple_loss=0.3426, pruned_loss=0.091, over 17251.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3386, pruned_loss=0.09601, over 3202738.19 frames. ], batch size: 52, lr: 2.75e-02, grad_scale: 16.0 2023-04-27 18:30:34,536 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15117.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:30:37,414 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.417e+02 3.788e+02 4.518e+02 5.433e+02 9.876e+02, threshold=9.036e+02, percent-clipped=3.0 2023-04-27 18:31:00,760 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 18:31:12,940 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15143.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:31:22,559 INFO [train.py:904] (0/8) Epoch 2, batch 5000, loss[loss=0.2642, simple_loss=0.3452, pruned_loss=0.09163, over 16922.00 frames. ], tot_loss[loss=0.2661, simple_loss=0.3401, pruned_loss=0.09603, over 3207309.94 frames. ], batch size: 109, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:31:28,096 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 18:32:01,074 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5318, 2.0604, 1.7575, 1.8638, 2.8255, 2.6682, 3.8066, 3.3892], device='cuda:0'), covar=tensor([0.0010, 0.0159, 0.0172, 0.0189, 0.0077, 0.0135, 0.0016, 0.0040], device='cuda:0'), in_proj_covar=tensor([0.0043, 0.0095, 0.0094, 0.0098, 0.0083, 0.0097, 0.0052, 0.0066], device='cuda:0'), out_proj_covar=tensor([6.7706e-05, 1.4943e-04, 1.4485e-04, 1.5902e-04, 1.3747e-04, 1.5916e-04, 8.4712e-05, 1.1111e-04], device='cuda:0') 2023-04-27 18:32:18,198 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9794, 3.8155, 3.5647, 4.2371, 4.2395, 3.9637, 4.1903, 4.1859], device='cuda:0'), covar=tensor([0.0414, 0.0479, 0.1617, 0.0463, 0.0432, 0.0503, 0.0463, 0.0367], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0263, 0.0370, 0.0273, 0.0213, 0.0198, 0.0202, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:32:20,900 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:26,389 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15195.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:32:34,520 INFO [train.py:904] (0/8) Epoch 2, batch 5050, loss[loss=0.3025, simple_loss=0.3659, pruned_loss=0.1196, over 15179.00 frames. ], tot_loss[loss=0.2666, simple_loss=0.341, pruned_loss=0.09609, over 3206939.34 frames. ], batch size: 190, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:00,587 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.434e+02 3.699e+02 4.750e+02 5.849e+02 1.293e+03, threshold=9.501e+02, percent-clipped=2.0 2023-04-27 18:33:45,533 INFO [train.py:904] (0/8) Epoch 2, batch 5100, loss[loss=0.2231, simple_loss=0.2993, pruned_loss=0.07347, over 16866.00 frames. ], tot_loss[loss=0.2644, simple_loss=0.3389, pruned_loss=0.09494, over 3202332.14 frames. ], batch size: 96, lr: 2.74e-02, grad_scale: 16.0 2023-04-27 18:33:52,831 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15256.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 18:34:54,171 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-27 18:34:56,476 INFO [train.py:904] (0/8) Epoch 2, batch 5150, loss[loss=0.2381, simple_loss=0.3247, pruned_loss=0.07577, over 16824.00 frames. ], tot_loss[loss=0.2626, simple_loss=0.3381, pruned_loss=0.09358, over 3192515.79 frames. ], batch size: 102, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:34:58,521 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 18:35:22,260 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.454e+02 3.560e+02 4.122e+02 5.032e+02 1.152e+03, threshold=8.244e+02, percent-clipped=1.0 2023-04-27 18:36:10,030 INFO [train.py:904] (0/8) Epoch 2, batch 5200, loss[loss=0.286, simple_loss=0.3633, pruned_loss=0.1043, over 16208.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3377, pruned_loss=0.09418, over 3193674.41 frames. ], batch size: 165, lr: 2.73e-02, grad_scale: 16.0 2023-04-27 18:36:35,104 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.53 vs. limit=5.0 2023-04-27 18:37:16,267 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15396.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:37:23,471 INFO [train.py:904] (0/8) Epoch 2, batch 5250, loss[loss=0.2497, simple_loss=0.313, pruned_loss=0.09323, over 16838.00 frames. ], tot_loss[loss=0.2612, simple_loss=0.3344, pruned_loss=0.09403, over 3207223.80 frames. ], batch size: 42, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:37:48,867 INFO [optim.py:368] (0/8) 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 18:38:06,536 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9016, 5.1813, 4.9229, 5.0676, 4.5684, 4.4684, 4.7419, 5.2973], device='cuda:0'), covar=tensor([0.0388, 0.0471, 0.0639, 0.0273, 0.0412, 0.0379, 0.0356, 0.0398], device='cuda:0'), in_proj_covar=tensor([0.0216, 0.0298, 0.0271, 0.0192, 0.0204, 0.0183, 0.0242, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:38:23,203 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15444.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:38:29,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5885, 4.5182, 4.9950, 5.1032, 4.9971, 4.5678, 4.5773, 4.6023], device='cuda:0'), covar=tensor([0.0171, 0.0237, 0.0206, 0.0201, 0.0314, 0.0163, 0.0543, 0.0212], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0149, 0.0169, 0.0161, 0.0194, 0.0156, 0.0244, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-27 18:38:32,791 INFO [train.py:904] (0/8) Epoch 2, batch 5300, loss[loss=0.2118, simple_loss=0.287, pruned_loss=0.06828, over 16473.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.3302, pruned_loss=0.09201, over 3213815.64 frames. ], batch size: 146, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:38:54,032 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9756, 3.8102, 3.7197, 3.8555, 3.3684, 3.7376, 3.5931, 3.5870], device='cuda:0'), covar=tensor([0.0231, 0.0170, 0.0202, 0.0125, 0.0717, 0.0182, 0.0476, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0094, 0.0160, 0.0124, 0.0190, 0.0131, 0.0111, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 18:38:57,083 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7333, 4.6466, 5.1085, 5.2282, 5.1631, 4.6909, 4.8146, 4.6608], device='cuda:0'), covar=tensor([0.0152, 0.0348, 0.0313, 0.0264, 0.0320, 0.0165, 0.0541, 0.0204], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0151, 0.0172, 0.0162, 0.0197, 0.0159, 0.0246, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-27 18:39:24,175 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8887, 2.5027, 2.1878, 3.1310, 2.1457, 3.0262, 2.4366, 2.0838], device='cuda:0'), covar=tensor([0.0275, 0.0348, 0.0305, 0.0193, 0.1086, 0.0165, 0.0501, 0.0947], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0173, 0.0146, 0.0201, 0.0252, 0.0162, 0.0179, 0.0234], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:39:27,323 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1526, 3.9995, 3.9903, 3.0961, 3.8831, 3.8343, 4.1408, 2.0166], device='cuda:0'), covar=tensor([0.0564, 0.0024, 0.0035, 0.0252, 0.0048, 0.0078, 0.0031, 0.0513], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0050, 0.0059, 0.0110, 0.0053, 0.0057, 0.0057, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 18:39:43,338 INFO [train.py:904] (0/8) Epoch 2, batch 5350, loss[loss=0.2436, simple_loss=0.3222, pruned_loss=0.08248, over 16553.00 frames. ], tot_loss[loss=0.2549, simple_loss=0.328, pruned_loss=0.09096, over 3214302.44 frames. ], batch size: 68, lr: 2.72e-02, grad_scale: 16.0 2023-04-27 18:40:09,986 INFO [optim.py:368] (0/8) 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 18:40:19,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7622, 3.3412, 2.2861, 4.3430, 4.2460, 4.0075, 1.8608, 3.1657], device='cuda:0'), covar=tensor([0.1402, 0.0363, 0.1293, 0.0059, 0.0144, 0.0261, 0.1178, 0.0538], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0120, 0.0166, 0.0068, 0.0104, 0.0114, 0.0152, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 18:40:53,797 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2043, 4.3537, 3.5921, 4.4770, 3.9257, 3.9863, 4.1607, 4.3578], device='cuda:0'), covar=tensor([0.1010, 0.1336, 0.2098, 0.0646, 0.0971, 0.0867, 0.0786, 0.1194], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0290, 0.0271, 0.0187, 0.0198, 0.0181, 0.0238, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:40:56,838 INFO [train.py:904] (0/8) Epoch 2, batch 5400, loss[loss=0.2878, simple_loss=0.3568, pruned_loss=0.1093, over 16247.00 frames. ], tot_loss[loss=0.2589, simple_loss=0.3323, pruned_loss=0.0927, over 3219012.62 frames. ], batch size: 165, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:40:57,159 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=15551.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 18:41:31,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8054, 4.3949, 4.5693, 4.7037, 4.0460, 4.5011, 4.5899, 4.3444], device='cuda:0'), covar=tensor([0.0269, 0.0237, 0.0179, 0.0109, 0.0741, 0.0211, 0.0195, 0.0220], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0097, 0.0160, 0.0124, 0.0190, 0.0133, 0.0111, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 18:41:50,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1298, 4.0173, 3.6784, 1.7365, 2.8298, 2.3871, 3.5674, 4.0528], device='cuda:0'), covar=tensor([0.0243, 0.0318, 0.0361, 0.1616, 0.0708, 0.0864, 0.0606, 0.0370], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0104, 0.0147, 0.0150, 0.0145, 0.0135, 0.0146, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 18:42:14,665 INFO [train.py:904] (0/8) Epoch 2, batch 5450, loss[loss=0.2798, simple_loss=0.3489, pruned_loss=0.1054, over 16592.00 frames. ], tot_loss[loss=0.2643, simple_loss=0.3369, pruned_loss=0.09584, over 3211446.06 frames. ], batch size: 57, lr: 2.71e-02, grad_scale: 16.0 2023-04-27 18:42:43,049 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.524e+02 4.170e+02 4.882e+02 6.060e+02 1.199e+03, threshold=9.765e+02, percent-clipped=3.0 2023-04-27 18:43:34,039 INFO [train.py:904] (0/8) Epoch 2, batch 5500, loss[loss=0.3867, simple_loss=0.4178, pruned_loss=0.1778, over 11948.00 frames. ], tot_loss[loss=0.2774, simple_loss=0.347, pruned_loss=0.1039, over 3188597.49 frames. ], batch size: 248, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:44:21,983 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9548, 3.8230, 3.0603, 2.9866, 3.0227, 2.3270, 4.1026, 4.7113], device='cuda:0'), covar=tensor([0.1612, 0.0648, 0.1075, 0.0629, 0.1946, 0.1099, 0.0299, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0227, 0.0242, 0.0176, 0.0280, 0.0183, 0.0197, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:44:50,703 INFO [train.py:904] (0/8) Epoch 2, batch 5550, loss[loss=0.4121, simple_loss=0.4297, pruned_loss=0.1973, over 11374.00 frames. ], tot_loss[loss=0.2922, simple_loss=0.3576, pruned_loss=0.1134, over 3156532.40 frames. ], batch size: 247, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:44:59,224 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9394, 3.9924, 4.3719, 4.4566, 4.4410, 3.9589, 4.0231, 4.1144], device='cuda:0'), covar=tensor([0.0235, 0.0252, 0.0342, 0.0353, 0.0357, 0.0250, 0.0695, 0.0307], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0147, 0.0168, 0.0165, 0.0196, 0.0159, 0.0248, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-27 18:45:19,405 INFO [optim.py:368] (0/8) 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 18:46:01,096 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9536, 2.6581, 2.6658, 1.7049, 2.8121, 2.7745, 2.3227, 2.4591], device='cuda:0'), covar=tensor([0.0811, 0.0134, 0.0196, 0.1280, 0.0112, 0.0099, 0.0392, 0.0365], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0082, 0.0078, 0.0152, 0.0073, 0.0071, 0.0106, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:46:11,054 INFO [train.py:904] (0/8) Epoch 2, batch 5600, loss[loss=0.3209, simple_loss=0.382, pruned_loss=0.1299, over 16897.00 frames. ], tot_loss[loss=0.303, simple_loss=0.3651, pruned_loss=0.1205, over 3117598.73 frames. ], batch size: 96, lr: 2.70e-02, grad_scale: 16.0 2023-04-27 18:46:24,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-27 18:47:34,307 INFO [train.py:904] (0/8) Epoch 2, batch 5650, loss[loss=0.3636, simple_loss=0.4039, pruned_loss=0.1616, over 16190.00 frames. ], tot_loss[loss=0.3154, simple_loss=0.3733, pruned_loss=0.1287, over 3061873.56 frames. ], batch size: 165, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:01,995 INFO [optim.py:368] (0/8) 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 18:48:53,344 INFO [train.py:904] (0/8) Epoch 2, batch 5700, loss[loss=0.3145, simple_loss=0.3918, pruned_loss=0.1186, over 16422.00 frames. ], tot_loss[loss=0.3195, simple_loss=0.376, pruned_loss=0.1315, over 3050145.19 frames. ], batch size: 75, lr: 2.69e-02, grad_scale: 16.0 2023-04-27 18:48:53,727 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=15851.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:49:02,807 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 18:49:28,546 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-04-27 18:49:44,713 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=15883.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:09,231 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=15899.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:50:12,116 INFO [train.py:904] (0/8) Epoch 2, batch 5750, loss[loss=0.2894, simple_loss=0.3648, pruned_loss=0.107, over 16624.00 frames. ], tot_loss[loss=0.3235, simple_loss=0.3796, pruned_loss=0.1337, over 3028754.13 frames. ], batch size: 76, lr: 2.69e-02, grad_scale: 8.0 2023-04-27 18:50:42,012 INFO [optim.py:368] (0/8) 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 18:51:09,916 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4244, 3.3826, 3.2594, 2.8928, 3.3223, 2.2046, 3.1353, 3.0964], device='cuda:0'), covar=tensor([0.0073, 0.0056, 0.0084, 0.0228, 0.0057, 0.0979, 0.0070, 0.0088], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0051, 0.0075, 0.0098, 0.0058, 0.0109, 0.0070, 0.0074], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:51:22,971 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=15944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:51:33,524 INFO [train.py:904] (0/8) Epoch 2, batch 5800, loss[loss=0.305, simple_loss=0.3733, pruned_loss=0.1184, over 16719.00 frames. ], tot_loss[loss=0.3204, simple_loss=0.3785, pruned_loss=0.1312, over 3034651.30 frames. ], batch size: 134, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:30,177 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6613, 4.4004, 4.4854, 4.5610, 3.9804, 4.4840, 4.4959, 4.2756], device='cuda:0'), covar=tensor([0.0291, 0.0163, 0.0152, 0.0104, 0.0682, 0.0172, 0.0154, 0.0191], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0088, 0.0145, 0.0113, 0.0176, 0.0121, 0.0103, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 18:52:51,631 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-16000.pt 2023-04-27 18:52:56,032 INFO [train.py:904] (0/8) Epoch 2, batch 5850, loss[loss=0.2899, simple_loss=0.3646, pruned_loss=0.1076, over 16790.00 frames. ], tot_loss[loss=0.3144, simple_loss=0.3747, pruned_loss=0.1271, over 3064828.09 frames. ], batch size: 102, lr: 2.68e-02, grad_scale: 8.0 2023-04-27 18:52:57,360 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16001.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:53:25,315 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.722e+02 4.866e+02 5.967e+02 7.152e+02 1.262e+03, threshold=1.193e+03, percent-clipped=0.0 2023-04-27 18:54:18,254 INFO [train.py:904] (0/8) Epoch 2, batch 5900, loss[loss=0.2681, simple_loss=0.3413, pruned_loss=0.09743, over 16204.00 frames. ], tot_loss[loss=0.3113, simple_loss=0.373, pruned_loss=0.1248, over 3090364.42 frames. ], batch size: 165, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:54:39,984 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16062.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:03,508 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6366, 3.0535, 2.9220, 2.1241, 2.9712, 2.8198, 3.0879, 1.4773], device='cuda:0'), covar=tensor([0.0570, 0.0063, 0.0082, 0.0332, 0.0068, 0.0140, 0.0051, 0.0597], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0050, 0.0056, 0.0107, 0.0050, 0.0058, 0.0056, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0001, 0.0003], device='cuda:0') 2023-04-27 18:55:07,082 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.75 vs. limit=2.0 2023-04-27 18:55:17,991 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16085.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:55:18,248 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-04-27 18:55:42,182 INFO [train.py:904] (0/8) Epoch 2, batch 5950, loss[loss=0.2676, simple_loss=0.3403, pruned_loss=0.09741, over 16555.00 frames. ], tot_loss[loss=0.3089, simple_loss=0.3728, pruned_loss=0.1225, over 3092821.00 frames. ], batch size: 62, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:56:13,225 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 18:56:13,607 INFO [optim.py:368] (0/8) 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 18:56:56,125 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:01,409 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16149.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:57:04,563 INFO [train.py:904] (0/8) Epoch 2, batch 6000, loss[loss=0.2882, simple_loss=0.356, pruned_loss=0.1102, over 16245.00 frames. ], tot_loss[loss=0.3093, simple_loss=0.3727, pruned_loss=0.123, over 3086551.34 frames. ], batch size: 165, lr: 2.67e-02, grad_scale: 8.0 2023-04-27 18:57:04,564 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 18:57:15,924 INFO [train.py:938] (0/8) Epoch 2, validation: loss=0.2302, simple_loss=0.3372, pruned_loss=0.06166, over 944034.00 frames. 2023-04-27 18:57:15,925 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 18:57:29,089 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-27 18:57:30,559 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4284, 4.5077, 4.4616, 3.1461, 4.5115, 4.6550, 4.7145, 2.1766], device='cuda:0'), covar=tensor([0.0487, 0.0016, 0.0033, 0.0289, 0.0024, 0.0036, 0.0014, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0050, 0.0054, 0.0105, 0.0049, 0.0055, 0.0055, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0001, 0.0003], device='cuda:0') 2023-04-27 18:58:04,755 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3819, 1.4340, 1.7938, 2.2177, 2.3566, 2.2000, 1.3478, 2.3595], device='cuda:0'), covar=tensor([0.0051, 0.0257, 0.0134, 0.0123, 0.0038, 0.0077, 0.0206, 0.0044], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0108, 0.0092, 0.0081, 0.0058, 0.0058, 0.0091, 0.0053], device='cuda:0'), out_proj_covar=tensor([1.2274e-04, 1.9124e-04, 1.6812e-04, 1.5142e-04, 9.9042e-05, 1.0449e-04, 1.5573e-04, 9.2256e-05], device='cuda:0') 2023-04-27 18:58:34,805 INFO [train.py:904] (0/8) Epoch 2, batch 6050, loss[loss=0.2806, simple_loss=0.3499, pruned_loss=0.1056, over 16677.00 frames. ], tot_loss[loss=0.3067, simple_loss=0.3701, pruned_loss=0.1217, over 3099276.20 frames. ], batch size: 57, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 18:58:48,000 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 18:59:06,028 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7524, 2.6435, 1.5846, 2.6178, 2.1591, 2.6474, 1.8460, 2.3912], device='cuda:0'), covar=tensor([0.0081, 0.0206, 0.1141, 0.0062, 0.0615, 0.0438, 0.1117, 0.0528], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0102, 0.0168, 0.0075, 0.0159, 0.0129, 0.0174, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 18:59:12,752 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 18:59:24,284 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8176, 3.7047, 1.3831, 3.6301, 2.5355, 3.7105, 1.7460, 2.6589], device='cuda:0'), covar=tensor([0.0040, 0.0145, 0.1818, 0.0041, 0.0768, 0.0282, 0.1571, 0.0654], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0103, 0.0170, 0.0076, 0.0160, 0.0130, 0.0176, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 18:59:33,563 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16239.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 18:59:52,061 INFO [train.py:904] (0/8) Epoch 2, batch 6100, loss[loss=0.3805, simple_loss=0.404, pruned_loss=0.1785, over 11665.00 frames. ], tot_loss[loss=0.3054, simple_loss=0.3695, pruned_loss=0.1207, over 3099575.80 frames. ], batch size: 247, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:00:00,566 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4288, 3.4698, 2.5969, 4.7703, 2.5629, 4.8814, 2.8528, 3.0084], device='cuda:0'), covar=tensor([0.0207, 0.0311, 0.0308, 0.0136, 0.1175, 0.0099, 0.0539, 0.0971], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0176, 0.0149, 0.0203, 0.0257, 0.0163, 0.0180, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:00:25,506 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4932, 3.5008, 1.5025, 3.3355, 2.4403, 3.5230, 1.7512, 2.5320], device='cuda:0'), covar=tensor([0.0050, 0.0125, 0.1593, 0.0046, 0.0754, 0.0277, 0.1419, 0.0623], device='cuda:0'), in_proj_covar=tensor([0.0075, 0.0103, 0.0170, 0.0076, 0.0161, 0.0131, 0.0177, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 19:00:40,169 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1819, 3.1444, 3.1444, 3.4367, 3.3793, 3.2228, 3.3792, 3.3543], device='cuda:0'), covar=tensor([0.0523, 0.0437, 0.0991, 0.0382, 0.0487, 0.0808, 0.0441, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0271, 0.0373, 0.0277, 0.0211, 0.0194, 0.0208, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:01:17,147 INFO [train.py:904] (0/8) Epoch 2, batch 6150, loss[loss=0.3391, simple_loss=0.3817, pruned_loss=0.1483, over 11513.00 frames. ], tot_loss[loss=0.3034, simple_loss=0.3672, pruned_loss=0.1198, over 3100292.84 frames. ], batch size: 247, lr: 2.66e-02, grad_scale: 8.0 2023-04-27 19:01:30,224 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0659, 1.6076, 1.3423, 1.4011, 1.7037, 1.6549, 1.7592, 1.8416], device='cuda:0'), covar=tensor([0.0019, 0.0109, 0.0135, 0.0121, 0.0061, 0.0113, 0.0035, 0.0054], device='cuda:0'), in_proj_covar=tensor([0.0040, 0.0095, 0.0095, 0.0099, 0.0085, 0.0097, 0.0052, 0.0070], device='cuda:0'), out_proj_covar=tensor([5.7746e-05, 1.4911e-04, 1.4468e-04, 1.5713e-04, 1.4031e-04, 1.5858e-04, 8.3219e-05, 1.1559e-04], device='cuda:0') 2023-04-27 19:01:45,850 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.137e+02 4.737e+02 5.827e+02 7.174e+02 1.739e+03, threshold=1.165e+03, percent-clipped=5.0 2023-04-27 19:02:34,196 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-27 19:02:34,542 INFO [train.py:904] (0/8) Epoch 2, batch 6200, loss[loss=0.2699, simple_loss=0.3434, pruned_loss=0.09817, over 16885.00 frames. ], tot_loss[loss=0.3014, simple_loss=0.3649, pruned_loss=0.1189, over 3105082.99 frames. ], batch size: 116, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:02:45,553 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16357.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:06,853 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16371.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:03:51,378 INFO [train.py:904] (0/8) Epoch 2, batch 6250, loss[loss=0.2708, simple_loss=0.3417, pruned_loss=0.09996, over 17007.00 frames. ], tot_loss[loss=0.3, simple_loss=0.3642, pruned_loss=0.1179, over 3119951.40 frames. ], batch size: 50, lr: 2.65e-02, grad_scale: 8.0 2023-04-27 19:04:05,016 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7097, 3.9949, 3.3618, 2.9964, 3.1172, 2.2667, 4.1946, 4.6825], device='cuda:0'), covar=tensor([0.1796, 0.0554, 0.0951, 0.0652, 0.1560, 0.1197, 0.0304, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0224, 0.0239, 0.0179, 0.0274, 0.0180, 0.0198, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:04:18,709 INFO [optim.py:368] (0/8) 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 19:04:37,400 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16432.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:04:51,252 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16441.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:05:05,089 INFO [train.py:904] (0/8) Epoch 2, batch 6300, loss[loss=0.3972, simple_loss=0.4182, pruned_loss=0.1881, over 11160.00 frames. ], tot_loss[loss=0.301, simple_loss=0.3649, pruned_loss=0.1185, over 3102906.98 frames. ], batch size: 248, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:18,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7412, 4.4583, 4.1524, 4.9773, 4.9524, 4.3951, 4.9854, 4.8609], device='cuda:0'), covar=tensor([0.0528, 0.0499, 0.1544, 0.0445, 0.0594, 0.0357, 0.0471, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0270, 0.0376, 0.0282, 0.0217, 0.0198, 0.0216, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:06:22,081 INFO [train.py:904] (0/8) Epoch 2, batch 6350, loss[loss=0.2691, simple_loss=0.3463, pruned_loss=0.09591, over 16872.00 frames. ], tot_loss[loss=0.3032, simple_loss=0.3662, pruned_loss=0.1201, over 3094939.86 frames. ], batch size: 96, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:06:29,799 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16505.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:06:52,066 INFO [optim.py:368] (0/8) 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 19:07:02,602 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-27 19:07:18,066 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.27 vs. limit=5.0 2023-04-27 19:07:21,376 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16539.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:07:39,373 INFO [train.py:904] (0/8) Epoch 2, batch 6400, loss[loss=0.279, simple_loss=0.3524, pruned_loss=0.1028, over 16298.00 frames. ], tot_loss[loss=0.3053, simple_loss=0.367, pruned_loss=0.1218, over 3085611.49 frames. ], batch size: 165, lr: 2.64e-02, grad_scale: 8.0 2023-04-27 19:08:34,559 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16587.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:08:55,796 INFO [train.py:904] (0/8) Epoch 2, batch 6450, loss[loss=0.2771, simple_loss=0.3414, pruned_loss=0.1064, over 16183.00 frames. ], tot_loss[loss=0.302, simple_loss=0.365, pruned_loss=0.1195, over 3093742.03 frames. ], batch size: 165, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:08:56,402 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-27 19:09:25,708 INFO [optim.py:368] (0/8) 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 19:10:13,792 INFO [train.py:904] (0/8) Epoch 2, batch 6500, loss[loss=0.2928, simple_loss=0.3389, pruned_loss=0.1233, over 11642.00 frames. ], tot_loss[loss=0.2973, simple_loss=0.3609, pruned_loss=0.1169, over 3113037.56 frames. ], batch size: 246, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:10:24,352 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16657.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:32,131 INFO [train.py:904] (0/8) Epoch 2, batch 6550, loss[loss=0.3031, simple_loss=0.3868, pruned_loss=0.1097, over 16663.00 frames. ], tot_loss[loss=0.3016, simple_loss=0.365, pruned_loss=0.1191, over 3101310.90 frames. ], batch size: 134, lr: 2.63e-02, grad_scale: 8.0 2023-04-27 19:11:37,691 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16705.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:11:51,713 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2123, 2.3250, 1.6741, 1.9122, 2.8137, 2.7088, 3.4067, 3.1111], device='cuda:0'), covar=tensor([0.0013, 0.0114, 0.0154, 0.0146, 0.0062, 0.0093, 0.0027, 0.0047], device='cuda:0'), in_proj_covar=tensor([0.0042, 0.0096, 0.0097, 0.0099, 0.0087, 0.0097, 0.0054, 0.0070], device='cuda:0'), out_proj_covar=tensor([6.1290e-05, 1.4911e-04, 1.4770e-04, 1.5477e-04, 1.4208e-04, 1.5720e-04, 8.5141e-05, 1.1592e-04], device='cuda:0') 2023-04-27 19:11:59,597 INFO [optim.py:368] (0/8) 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 19:12:00,121 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9502, 4.9820, 4.7213, 4.0696, 4.7924, 2.0474, 4.5315, 4.7967], device='cuda:0'), covar=tensor([0.0063, 0.0036, 0.0052, 0.0266, 0.0045, 0.1116, 0.0049, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0050, 0.0074, 0.0094, 0.0056, 0.0107, 0.0068, 0.0074], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:12:11,163 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=16727.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:32,998 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:12:47,637 INFO [train.py:904] (0/8) Epoch 2, batch 6600, loss[loss=0.3413, simple_loss=0.3908, pruned_loss=0.1458, over 15329.00 frames. ], tot_loss[loss=0.3052, simple_loss=0.3688, pruned_loss=0.1208, over 3110453.89 frames. ], batch size: 190, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:13:46,758 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16789.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:06,697 INFO [train.py:904] (0/8) Epoch 2, batch 6650, loss[loss=0.2754, simple_loss=0.3435, pruned_loss=0.1037, over 17122.00 frames. ], tot_loss[loss=0.3072, simple_loss=0.3696, pruned_loss=0.1224, over 3094330.73 frames. ], batch size: 47, lr: 2.62e-02, grad_scale: 8.0 2023-04-27 19:14:13,277 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=16805.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:14:35,747 INFO [optim.py:368] (0/8) 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 19:15:23,522 INFO [train.py:904] (0/8) Epoch 2, batch 6700, loss[loss=0.3411, simple_loss=0.3773, pruned_loss=0.1524, over 11416.00 frames. ], tot_loss[loss=0.3069, simple_loss=0.3685, pruned_loss=0.1226, over 3086750.12 frames. ], batch size: 247, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:15:26,816 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=16853.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:15:50,016 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16868.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:15:58,536 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1533, 4.0756, 4.2379, 4.5257, 4.5099, 4.1154, 4.4924, 4.4501], device='cuda:0'), covar=tensor([0.0528, 0.0414, 0.0808, 0.0304, 0.0370, 0.0477, 0.0312, 0.0265], device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0274, 0.0382, 0.0287, 0.0224, 0.0197, 0.0221, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:15:58,912 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-27 19:16:38,385 INFO [train.py:904] (0/8) Epoch 2, batch 6750, loss[loss=0.3152, simple_loss=0.3745, pruned_loss=0.128, over 15232.00 frames. ], tot_loss[loss=0.3065, simple_loss=0.3673, pruned_loss=0.1228, over 3081211.36 frames. ], batch size: 190, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:16:46,648 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=16906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:17:07,558 INFO [optim.py:368] (0/8) 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 19:17:16,438 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-04-27 19:17:22,301 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16929.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:17:26,467 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7335, 1.4860, 2.0728, 2.6100, 2.5686, 2.6532, 1.4788, 2.8121], device='cuda:0'), covar=tensor([0.0036, 0.0234, 0.0124, 0.0080, 0.0036, 0.0063, 0.0183, 0.0032], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0107, 0.0092, 0.0081, 0.0058, 0.0057, 0.0091, 0.0052], device='cuda:0'), out_proj_covar=tensor([1.1838e-04, 1.8677e-04, 1.6749e-04, 1.4754e-04, 1.0046e-04, 1.0280e-04, 1.5520e-04, 9.0199e-05], device='cuda:0') 2023-04-27 19:17:33,441 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2432, 3.1612, 1.4787, 3.2162, 2.2991, 3.2681, 1.6447, 2.4127], device='cuda:0'), covar=tensor([0.0041, 0.0172, 0.1241, 0.0048, 0.0603, 0.0201, 0.1239, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0104, 0.0167, 0.0072, 0.0159, 0.0135, 0.0176, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 19:17:53,323 INFO [train.py:904] (0/8) Epoch 2, batch 6800, loss[loss=0.3338, simple_loss=0.3843, pruned_loss=0.1417, over 11312.00 frames. ], tot_loss[loss=0.3056, simple_loss=0.3672, pruned_loss=0.122, over 3088247.99 frames. ], batch size: 246, lr: 2.61e-02, grad_scale: 8.0 2023-04-27 19:18:18,749 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=16967.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:19:03,050 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-27 19:19:10,229 INFO [train.py:904] (0/8) Epoch 2, batch 6850, loss[loss=0.4017, simple_loss=0.4221, pruned_loss=0.1906, over 11878.00 frames. ], tot_loss[loss=0.3082, simple_loss=0.3692, pruned_loss=0.1236, over 3078889.98 frames. ], batch size: 246, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:19:38,489 INFO [optim.py:368] (0/8) 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 19:19:49,702 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 2, batch 6900, loss[loss=0.3092, simple_loss=0.3838, pruned_loss=0.1173, over 16786.00 frames. ], tot_loss[loss=0.3083, simple_loss=0.371, pruned_loss=0.1228, over 3077245.81 frames. ], batch size: 102, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:21:00,394 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6025, 3.6796, 2.8951, 2.7954, 2.8796, 2.0690, 3.9557, 4.3981], device='cuda:0'), covar=tensor([0.1778, 0.0638, 0.1065, 0.0620, 0.1723, 0.1191, 0.0335, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0216, 0.0231, 0.0171, 0.0262, 0.0175, 0.0192, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:21:01,219 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17075.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:21:32,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8238, 3.8271, 3.7291, 3.1552, 3.7757, 1.7012, 3.4651, 3.6773], device='cuda:0'), covar=tensor([0.0076, 0.0057, 0.0082, 0.0298, 0.0058, 0.1424, 0.0080, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0063, 0.0052, 0.0078, 0.0096, 0.0057, 0.0111, 0.0070, 0.0077], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:21:40,742 INFO [train.py:904] (0/8) Epoch 2, batch 6950, loss[loss=0.3206, simple_loss=0.3741, pruned_loss=0.1336, over 16367.00 frames. ], tot_loss[loss=0.314, simple_loss=0.3746, pruned_loss=0.1267, over 3056828.39 frames. ], batch size: 146, lr: 2.60e-02, grad_scale: 8.0 2023-04-27 19:22:09,950 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.451e+02 5.662e+02 6.855e+02 8.747e+02 1.724e+03, threshold=1.371e+03, percent-clipped=6.0 2023-04-27 19:22:54,972 INFO [train.py:904] (0/8) Epoch 2, batch 7000, loss[loss=0.3179, simple_loss=0.3652, pruned_loss=0.1353, over 11591.00 frames. ], tot_loss[loss=0.3113, simple_loss=0.3735, pruned_loss=0.1246, over 3073673.50 frames. ], batch size: 246, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:23:27,406 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8364, 3.7257, 4.2632, 4.2937, 4.2459, 3.8146, 3.9956, 3.9562], device='cuda:0'), covar=tensor([0.0202, 0.0271, 0.0261, 0.0246, 0.0291, 0.0232, 0.0495, 0.0254], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0147, 0.0164, 0.0160, 0.0192, 0.0162, 0.0244, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-27 19:23:53,028 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17189.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:24:10,879 INFO [train.py:904] (0/8) Epoch 2, batch 7050, loss[loss=0.3132, simple_loss=0.3869, pruned_loss=0.1198, over 16824.00 frames. ], tot_loss[loss=0.3093, simple_loss=0.3733, pruned_loss=0.1227, over 3100628.86 frames. ], batch size: 102, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:24:37,728 INFO [optim.py:368] (0/8) 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 19:24:43,957 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17224.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 19:24:47,403 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0454, 1.4711, 1.2994, 1.3947, 1.7657, 1.5838, 1.7977, 1.7797], device='cuda:0'), covar=tensor([0.0018, 0.0103, 0.0146, 0.0136, 0.0061, 0.0113, 0.0040, 0.0060], device='cuda:0'), in_proj_covar=tensor([0.0043, 0.0099, 0.0099, 0.0102, 0.0087, 0.0101, 0.0052, 0.0071], device='cuda:0'), out_proj_covar=tensor([6.1130e-05, 1.5302e-04, 1.5028e-04, 1.6077e-04, 1.4151e-04, 1.6188e-04, 8.1389e-05, 1.1584e-04], device='cuda:0') 2023-04-27 19:25:20,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4821, 3.3284, 2.6153, 2.6757, 2.6208, 1.9941, 3.4480, 3.8875], device='cuda:0'), covar=tensor([0.1692, 0.0602, 0.1161, 0.0671, 0.1678, 0.1249, 0.0350, 0.0200], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0225, 0.0239, 0.0181, 0.0277, 0.0183, 0.0199, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:25:22,955 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:25:23,307 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-27 19:25:23,674 INFO [train.py:904] (0/8) Epoch 2, batch 7100, loss[loss=0.2572, simple_loss=0.3278, pruned_loss=0.09336, over 16781.00 frames. ], tot_loss[loss=0.3067, simple_loss=0.3706, pruned_loss=0.1213, over 3112705.41 frames. ], batch size: 83, lr: 2.59e-02, grad_scale: 8.0 2023-04-27 19:25:40,651 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17262.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:26:38,668 INFO [train.py:904] (0/8) Epoch 2, batch 7150, loss[loss=0.2603, simple_loss=0.3294, pruned_loss=0.09563, over 17021.00 frames. ], tot_loss[loss=0.3037, simple_loss=0.3676, pruned_loss=0.1199, over 3127481.79 frames. ], batch size: 55, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:27:07,406 INFO [optim.py:368] (0/8) 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 19:27:22,629 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6439, 3.7206, 1.5036, 3.7630, 2.3808, 3.7299, 1.8937, 2.5552], device='cuda:0'), covar=tensor([0.0034, 0.0116, 0.1723, 0.0031, 0.0780, 0.0244, 0.1358, 0.0728], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0107, 0.0170, 0.0072, 0.0161, 0.0136, 0.0178, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 19:27:52,991 INFO [train.py:904] (0/8) Epoch 2, batch 7200, loss[loss=0.2459, simple_loss=0.3292, pruned_loss=0.08131, over 16864.00 frames. ], tot_loss[loss=0.2999, simple_loss=0.3648, pruned_loss=0.1175, over 3106445.02 frames. ], batch size: 96, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:28:00,458 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5071, 4.1945, 4.0664, 1.8359, 3.1623, 2.6240, 3.8252, 4.1747], device='cuda:0'), covar=tensor([0.0194, 0.0413, 0.0334, 0.1619, 0.0647, 0.0874, 0.0616, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0107, 0.0151, 0.0155, 0.0147, 0.0138, 0.0150, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 19:28:10,381 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17362.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:28:17,981 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8078, 5.0653, 4.8100, 4.8931, 4.3205, 4.2444, 4.5604, 5.1766], device='cuda:0'), covar=tensor([0.0376, 0.0554, 0.0704, 0.0324, 0.0488, 0.0568, 0.0432, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0302, 0.0281, 0.0191, 0.0205, 0.0193, 0.0249, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:28:36,599 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7099, 2.9955, 2.4440, 4.0986, 2.1419, 3.9210, 2.6286, 2.4376], device='cuda:0'), covar=tensor([0.0242, 0.0388, 0.0346, 0.0160, 0.1323, 0.0170, 0.0603, 0.1070], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0186, 0.0157, 0.0216, 0.0265, 0.0171, 0.0188, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:28:44,536 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17383.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:29:13,592 INFO [train.py:904] (0/8) Epoch 2, batch 7250, loss[loss=0.2354, simple_loss=0.3037, pruned_loss=0.08356, over 17267.00 frames. ], tot_loss[loss=0.2958, simple_loss=0.3614, pruned_loss=0.1151, over 3095367.74 frames. ], batch size: 52, lr: 2.58e-02, grad_scale: 8.0 2023-04-27 19:29:42,546 INFO [optim.py:368] (0/8) 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 19:29:46,735 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17423.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:29:48,996 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8907, 3.8900, 1.5569, 3.9206, 2.4710, 3.8835, 1.9781, 2.8133], device='cuda:0'), covar=tensor([0.0044, 0.0115, 0.1599, 0.0032, 0.0722, 0.0277, 0.1265, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0075, 0.0109, 0.0173, 0.0073, 0.0161, 0.0138, 0.0180, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 19:30:19,174 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17444.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:30:29,435 INFO [train.py:904] (0/8) Epoch 2, batch 7300, loss[loss=0.2823, simple_loss=0.3537, pruned_loss=0.1055, over 16377.00 frames. ], tot_loss[loss=0.2945, simple_loss=0.36, pruned_loss=0.1144, over 3097081.88 frames. ], batch size: 146, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:30:35,647 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-27 19:31:04,287 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7098, 4.9710, 4.7019, 4.8101, 4.2626, 4.1627, 4.4502, 4.9779], device='cuda:0'), covar=tensor([0.0352, 0.0468, 0.0630, 0.0282, 0.0407, 0.0515, 0.0379, 0.0458], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0296, 0.0272, 0.0186, 0.0198, 0.0186, 0.0240, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:31:46,484 INFO [train.py:904] (0/8) Epoch 2, batch 7350, loss[loss=0.2522, simple_loss=0.332, pruned_loss=0.08617, over 16742.00 frames. ], tot_loss[loss=0.2938, simple_loss=0.3592, pruned_loss=0.1142, over 3088624.03 frames. ], batch size: 83, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:32:16,423 INFO [optim.py:368] (0/8) 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 19:32:22,410 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17524.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:32:29,593 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8338, 3.6405, 3.8648, 4.1746, 4.2110, 3.8470, 4.1387, 4.0987], device='cuda:0'), covar=tensor([0.0584, 0.0530, 0.1099, 0.0439, 0.0386, 0.0661, 0.0416, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0268, 0.0369, 0.0275, 0.0216, 0.0198, 0.0216, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:32:56,920 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17545.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:33:06,160 INFO [train.py:904] (0/8) Epoch 2, batch 7400, loss[loss=0.2605, simple_loss=0.342, pruned_loss=0.08951, over 16851.00 frames. ], tot_loss[loss=0.2964, simple_loss=0.3612, pruned_loss=0.1158, over 3089278.91 frames. ], batch size: 102, lr: 2.57e-02, grad_scale: 8.0 2023-04-27 19:33:25,272 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17562.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:33:40,921 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17572.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:34:15,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7987, 3.7664, 3.1758, 3.0156, 3.0544, 2.3175, 4.0870, 4.4997], device='cuda:0'), covar=tensor([0.1581, 0.0547, 0.0952, 0.0549, 0.1488, 0.1056, 0.0271, 0.0173], device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0229, 0.0242, 0.0185, 0.0284, 0.0185, 0.0203, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:34:27,191 INFO [train.py:904] (0/8) Epoch 2, batch 7450, loss[loss=0.2669, simple_loss=0.3261, pruned_loss=0.1038, over 16736.00 frames. ], tot_loss[loss=0.2979, simple_loss=0.3623, pruned_loss=0.1167, over 3091288.29 frames. ], batch size: 39, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:34:43,768 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.824e+02 5.406e+02 6.493e+02 7.703e+02 1.630e+03, threshold=1.299e+03, percent-clipped=5.0 2023-04-27 19:35:49,123 INFO [train.py:904] (0/8) Epoch 2, batch 7500, loss[loss=0.2818, simple_loss=0.3425, pruned_loss=0.1105, over 16997.00 frames. ], tot_loss[loss=0.299, simple_loss=0.3636, pruned_loss=0.1172, over 3068169.30 frames. ], batch size: 55, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:36:03,067 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 19:37:05,025 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17700.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:05,684 INFO [train.py:904] (0/8) Epoch 2, batch 7550, loss[loss=0.2996, simple_loss=0.3696, pruned_loss=0.1149, over 16683.00 frames. ], tot_loss[loss=0.3012, simple_loss=0.3644, pruned_loss=0.119, over 3037657.40 frames. ], batch size: 134, lr: 2.56e-02, grad_scale: 8.0 2023-04-27 19:37:32,570 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17718.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:34,649 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.677e+02 5.137e+02 6.469e+02 8.298e+02 1.927e+03, threshold=1.294e+03, percent-clipped=3.0 2023-04-27 19:37:50,595 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17730.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:37:56,801 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3157, 3.3306, 2.8411, 2.5876, 2.6542, 2.1010, 3.4714, 3.9210], device='cuda:0'), covar=tensor([0.1964, 0.0733, 0.1011, 0.0676, 0.1686, 0.1188, 0.0371, 0.0216], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0227, 0.0240, 0.0183, 0.0278, 0.0182, 0.0202, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:38:04,543 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=17739.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:38:22,212 INFO [train.py:904] (0/8) Epoch 2, batch 7600, loss[loss=0.3009, simple_loss=0.3622, pruned_loss=0.1198, over 16466.00 frames. ], tot_loss[loss=0.3005, simple_loss=0.3635, pruned_loss=0.1187, over 3051093.71 frames. ], batch size: 146, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:38:37,826 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17761.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:38:58,571 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-27 19:39:24,545 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=17791.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:39:27,896 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-27 19:39:39,714 INFO [train.py:904] (0/8) Epoch 2, batch 7650, loss[loss=0.3753, simple_loss=0.4095, pruned_loss=0.1706, over 11442.00 frames. ], tot_loss[loss=0.3017, simple_loss=0.3641, pruned_loss=0.1197, over 3047484.83 frames. ], batch size: 248, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:40:10,730 INFO [optim.py:368] (0/8) 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 19:40:40,036 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8775, 3.5458, 2.1364, 4.7489, 4.5733, 4.0625, 2.2123, 2.8636], device='cuda:0'), covar=tensor([0.1499, 0.0392, 0.1551, 0.0063, 0.0158, 0.0235, 0.1112, 0.0816], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0122, 0.0166, 0.0068, 0.0109, 0.0118, 0.0153, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 19:40:49,726 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=17845.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:40:58,874 INFO [train.py:904] (0/8) Epoch 2, batch 7700, loss[loss=0.2651, simple_loss=0.3372, pruned_loss=0.09654, over 16402.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3644, pruned_loss=0.1201, over 3081761.49 frames. ], batch size: 68, lr: 2.55e-02, grad_scale: 8.0 2023-04-27 19:42:04,292 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=17893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:42:16,830 INFO [train.py:904] (0/8) Epoch 2, batch 7750, loss[loss=0.2699, simple_loss=0.3404, pruned_loss=0.09974, over 16452.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3647, pruned_loss=0.1199, over 3069960.13 frames. ], batch size: 68, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:42:28,291 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8315, 3.6401, 3.7488, 2.7492, 3.7194, 3.5472, 3.8635, 1.8154], device='cuda:0'), covar=tensor([0.0491, 0.0030, 0.0039, 0.0254, 0.0034, 0.0095, 0.0026, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0053, 0.0059, 0.0110, 0.0053, 0.0061, 0.0058, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 19:42:46,602 INFO [optim.py:368] (0/8) 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 19:43:09,445 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0958, 2.7186, 2.2609, 3.3976, 3.2544, 3.2131, 2.0000, 2.6408], device='cuda:0'), covar=tensor([0.1140, 0.0305, 0.1051, 0.0077, 0.0224, 0.0242, 0.0934, 0.0614], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0124, 0.0165, 0.0068, 0.0110, 0.0120, 0.0154, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 19:43:18,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6119, 2.8818, 2.2586, 3.9635, 2.0427, 3.7936, 2.5279, 2.3042], device='cuda:0'), covar=tensor([0.0295, 0.0387, 0.0378, 0.0180, 0.1375, 0.0171, 0.0619, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0187, 0.0157, 0.0217, 0.0264, 0.0170, 0.0189, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:43:31,904 INFO [train.py:904] (0/8) Epoch 2, batch 7800, loss[loss=0.2992, simple_loss=0.3819, pruned_loss=0.1083, over 16670.00 frames. ], tot_loss[loss=0.302, simple_loss=0.3648, pruned_loss=0.1196, over 3090107.40 frames. ], batch size: 89, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:43:47,040 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=17960.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:44:16,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5274, 5.8330, 5.4341, 5.5579, 4.8529, 4.8197, 5.2436, 5.9084], device='cuda:0'), covar=tensor([0.0353, 0.0394, 0.0728, 0.0355, 0.0505, 0.0393, 0.0390, 0.0418], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0307, 0.0282, 0.0196, 0.0207, 0.0198, 0.0251, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:44:48,060 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-18000.pt 2023-04-27 19:44:52,430 INFO [train.py:904] (0/8) Epoch 2, batch 7850, loss[loss=0.2759, simple_loss=0.3511, pruned_loss=0.1003, over 16692.00 frames. ], tot_loss[loss=0.3025, simple_loss=0.3661, pruned_loss=0.1195, over 3100065.31 frames. ], batch size: 124, lr: 2.54e-02, grad_scale: 16.0 2023-04-27 19:45:17,993 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 19:45:22,125 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 19:45:22,890 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18021.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:45:48,800 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18039.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 19:46:05,631 INFO [train.py:904] (0/8) Epoch 2, batch 7900, loss[loss=0.3019, simple_loss=0.3683, pruned_loss=0.1177, over 16884.00 frames. ], tot_loss[loss=0.3027, simple_loss=0.3661, pruned_loss=0.1197, over 3089184.26 frames. ], batch size: 116, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:46:13,998 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18056.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:46:18,164 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4184, 4.4260, 4.8489, 5.0046, 4.9848, 4.4468, 4.6251, 4.4762], device='cuda:0'), covar=tensor([0.0180, 0.0231, 0.0261, 0.0249, 0.0238, 0.0203, 0.0467, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0153, 0.0168, 0.0164, 0.0194, 0.0166, 0.0257, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-27 19:46:29,401 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18066.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:03,312 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18086.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:47:04,547 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18087.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:47:19,227 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8622, 3.8483, 1.6438, 3.8822, 2.6874, 4.0310, 1.8030, 2.7162], device='cuda:0'), covar=tensor([0.0037, 0.0182, 0.1719, 0.0036, 0.0717, 0.0214, 0.1622, 0.0605], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0113, 0.0175, 0.0075, 0.0161, 0.0140, 0.0181, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 19:47:25,404 INFO [train.py:904] (0/8) Epoch 2, batch 7950, loss[loss=0.2774, simple_loss=0.3483, pruned_loss=0.1032, over 16397.00 frames. ], tot_loss[loss=0.302, simple_loss=0.3653, pruned_loss=0.1193, over 3091743.28 frames. ], batch size: 146, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:47:56,298 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.877e+02 5.212e+02 6.244e+02 7.914e+02 2.379e+03, threshold=1.249e+03, percent-clipped=3.0 2023-04-27 19:48:00,632 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 19:48:42,388 INFO [train.py:904] (0/8) Epoch 2, batch 8000, loss[loss=0.3211, simple_loss=0.3808, pruned_loss=0.1307, over 16854.00 frames. ], tot_loss[loss=0.3023, simple_loss=0.3652, pruned_loss=0.1197, over 3077519.73 frames. ], batch size: 116, lr: 2.53e-02, grad_scale: 8.0 2023-04-27 19:49:02,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1811, 3.4235, 2.6034, 4.5444, 2.0809, 4.4121, 2.9800, 2.6020], device='cuda:0'), covar=tensor([0.0254, 0.0377, 0.0386, 0.0165, 0.1469, 0.0152, 0.0562, 0.1263], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0186, 0.0159, 0.0217, 0.0262, 0.0170, 0.0189, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 19:49:16,786 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18174.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:18,056 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18175.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:49:56,242 INFO [train.py:904] (0/8) Epoch 2, batch 8050, loss[loss=0.2936, simple_loss=0.3615, pruned_loss=0.1128, over 15303.00 frames. ], tot_loss[loss=0.3007, simple_loss=0.3642, pruned_loss=0.1186, over 3106437.66 frames. ], batch size: 190, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:50:24,925 INFO [optim.py:368] (0/8) 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 19:50:46,731 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18235.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:48,706 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18236.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:50:57,370 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8486, 1.5300, 2.0798, 2.5132, 2.3489, 2.8078, 1.5212, 2.7194], device='cuda:0'), covar=tensor([0.0040, 0.0262, 0.0132, 0.0101, 0.0054, 0.0048, 0.0183, 0.0034], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0112, 0.0096, 0.0082, 0.0063, 0.0056, 0.0099, 0.0056], device='cuda:0'), out_proj_covar=tensor([1.1941e-04, 1.9323e-04, 1.7120e-04, 1.4787e-04, 1.0685e-04, 9.7543e-05, 1.6948e-04, 9.3599e-05], device='cuda:0') 2023-04-27 19:51:11,139 INFO [train.py:904] (0/8) Epoch 2, batch 8100, loss[loss=0.2834, simple_loss=0.3574, pruned_loss=0.1047, over 16931.00 frames. ], tot_loss[loss=0.2995, simple_loss=0.3636, pruned_loss=0.1178, over 3096660.82 frames. ], batch size: 109, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:51:20,321 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 19:51:38,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9445, 3.2919, 3.3148, 1.4485, 3.4301, 3.4576, 2.9714, 2.6235], device='cuda:0'), covar=tensor([0.0869, 0.0118, 0.0171, 0.1369, 0.0081, 0.0063, 0.0237, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0088, 0.0084, 0.0150, 0.0076, 0.0069, 0.0109, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 19:51:52,905 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1414, 4.1317, 2.9159, 5.3879, 5.2665, 4.7453, 2.7230, 3.5505], device='cuda:0'), covar=tensor([0.1213, 0.0280, 0.0994, 0.0040, 0.0102, 0.0171, 0.0798, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0122, 0.0164, 0.0069, 0.0112, 0.0120, 0.0155, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 19:52:29,029 INFO [train.py:904] (0/8) Epoch 2, batch 8150, loss[loss=0.2453, simple_loss=0.3214, pruned_loss=0.08457, over 17221.00 frames. ], tot_loss[loss=0.2981, simple_loss=0.3617, pruned_loss=0.1173, over 3097010.65 frames. ], batch size: 45, lr: 2.52e-02, grad_scale: 8.0 2023-04-27 19:52:52,894 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18316.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:53:00,714 INFO [optim.py:368] (0/8) 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 19:53:48,026 INFO [train.py:904] (0/8) Epoch 2, batch 8200, loss[loss=0.3545, simple_loss=0.3892, pruned_loss=0.1599, over 11623.00 frames. ], tot_loss[loss=0.2964, simple_loss=0.3593, pruned_loss=0.1167, over 3079801.86 frames. ], batch size: 247, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:53:56,656 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18356.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:54:06,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1395, 4.0954, 1.8247, 4.2609, 2.6054, 4.2747, 2.4563, 2.7799], device='cuda:0'), covar=tensor([0.0044, 0.0158, 0.1541, 0.0026, 0.0749, 0.0226, 0.1209, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0113, 0.0170, 0.0074, 0.0156, 0.0141, 0.0176, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 19:54:07,414 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-27 19:54:46,381 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18386.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:10,283 INFO [train.py:904] (0/8) Epoch 2, batch 8250, loss[loss=0.2939, simple_loss=0.3632, pruned_loss=0.1123, over 15349.00 frames. ], tot_loss[loss=0.2932, simple_loss=0.3578, pruned_loss=0.1143, over 3063991.74 frames. ], batch size: 190, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:55:15,140 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18404.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:55:44,640 INFO [optim.py:368] (0/8) 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 19:56:05,995 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18434.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:56:33,583 INFO [train.py:904] (0/8) Epoch 2, batch 8300, loss[loss=0.2534, simple_loss=0.3253, pruned_loss=0.09078, over 12265.00 frames. ], tot_loss[loss=0.2852, simple_loss=0.3528, pruned_loss=0.1088, over 3067095.93 frames. ], batch size: 247, lr: 2.51e-02, grad_scale: 4.0 2023-04-27 19:57:37,611 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18490.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:57:54,229 INFO [train.py:904] (0/8) Epoch 2, batch 8350, loss[loss=0.3038, simple_loss=0.3547, pruned_loss=0.1265, over 11786.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3496, pruned_loss=0.1047, over 3061032.25 frames. ], batch size: 247, lr: 2.50e-02, grad_scale: 4.0 2023-04-27 19:58:06,076 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18508.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 19:58:28,821 INFO [optim.py:368] (0/8) 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 19:58:41,678 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18530.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:58:43,426 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18531.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:15,569 INFO [train.py:904] (0/8) Epoch 2, batch 8400, loss[loss=0.2406, simple_loss=0.324, pruned_loss=0.07862, over 16869.00 frames. ], tot_loss[loss=0.2745, simple_loss=0.3456, pruned_loss=0.1017, over 3039731.18 frames. ], batch size: 116, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 19:59:16,069 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18551.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 19:59:45,488 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18569.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 19:59:45,585 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4172, 3.0636, 2.4723, 2.3967, 2.3275, 2.0207, 2.8793, 3.3512], device='cuda:0'), covar=tensor([0.1528, 0.0624, 0.0968, 0.0624, 0.1584, 0.1394, 0.0388, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0212, 0.0227, 0.0173, 0.0240, 0.0180, 0.0190, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:00:35,049 INFO [train.py:904] (0/8) Epoch 2, batch 8450, loss[loss=0.2351, simple_loss=0.3203, pruned_loss=0.07498, over 16201.00 frames. ], tot_loss[loss=0.2708, simple_loss=0.3433, pruned_loss=0.09914, over 3048100.01 frames. ], batch size: 165, lr: 2.50e-02, grad_scale: 8.0 2023-04-27 20:01:00,300 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18616.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:01:09,359 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.749e+02 4.009e+02 5.012e+02 6.279e+02 1.629e+03, threshold=1.002e+03, percent-clipped=8.0 2023-04-27 20:01:55,565 INFO [train.py:904] (0/8) Epoch 2, batch 8500, loss[loss=0.2576, simple_loss=0.3321, pruned_loss=0.09154, over 15238.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3374, pruned_loss=0.09455, over 3058411.33 frames. ], batch size: 190, lr: 2.49e-02, grad_scale: 8.0 2023-04-27 20:02:17,757 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18664.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:33,901 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:02:38,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6092, 3.6181, 3.6922, 3.6559, 3.7729, 4.1231, 4.0132, 3.7149], device='cuda:0'), covar=tensor([0.1805, 0.1312, 0.1058, 0.1876, 0.2274, 0.1005, 0.0825, 0.2107], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0251, 0.0239, 0.0229, 0.0288, 0.0260, 0.0202, 0.0295], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:02:49,540 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6989, 3.4799, 3.7137, 3.9574, 3.9936, 3.5732, 3.9899, 3.9377], device='cuda:0'), covar=tensor([0.0540, 0.0541, 0.0950, 0.0390, 0.0362, 0.0790, 0.0346, 0.0341], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0279, 0.0368, 0.0278, 0.0212, 0.0196, 0.0219, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:03:19,439 INFO [train.py:904] (0/8) Epoch 2, batch 8550, loss[loss=0.2929, simple_loss=0.3642, pruned_loss=0.1108, over 16204.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3352, pruned_loss=0.09345, over 3043679.91 frames. ], batch size: 165, lr: 2.49e-02, grad_scale: 4.0 2023-04-27 20:04:03,005 INFO [optim.py:368] (0/8) 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 20:04:26,456 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18735.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:04:58,652 INFO [train.py:904] (0/8) Epoch 2, batch 8600, loss[loss=0.3042, simple_loss=0.3669, pruned_loss=0.1207, over 16390.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3361, pruned_loss=0.09271, over 3047302.24 frames. ], batch size: 146, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:05:42,762 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7500, 4.0500, 3.8127, 3.8932, 3.5236, 3.6692, 3.7478, 3.9909], device='cuda:0'), covar=tensor([0.0489, 0.0748, 0.0751, 0.0391, 0.0553, 0.0694, 0.0555, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0207, 0.0291, 0.0256, 0.0188, 0.0197, 0.0184, 0.0241, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:06:06,718 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5965, 3.2984, 2.5379, 4.4980, 4.3408, 4.1928, 1.8394, 3.3208], device='cuda:0'), covar=tensor([0.1731, 0.0417, 0.1386, 0.0070, 0.0172, 0.0277, 0.1366, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0124, 0.0169, 0.0071, 0.0113, 0.0124, 0.0157, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:06:37,154 INFO [train.py:904] (0/8) Epoch 2, batch 8650, loss[loss=0.2254, simple_loss=0.3169, pruned_loss=0.0669, over 16718.00 frames. ], tot_loss[loss=0.257, simple_loss=0.3334, pruned_loss=0.09029, over 3038716.80 frames. ], batch size: 134, lr: 2.49e-02, grad_scale: 2.0 2023-04-27 20:07:29,103 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.199e+02 3.846e+02 4.801e+02 6.827e+02 6.621e+03, threshold=9.601e+02, percent-clipped=16.0 2023-04-27 20:07:44,377 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18830.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:07:46,103 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=18831.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:14,573 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18846.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:08:14,651 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8025, 3.5679, 3.8224, 4.0784, 4.1165, 3.7438, 4.1184, 4.0656], device='cuda:0'), covar=tensor([0.0461, 0.0513, 0.0931, 0.0357, 0.0391, 0.0664, 0.0379, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0282, 0.0373, 0.0278, 0.0215, 0.0198, 0.0221, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:08:23,076 INFO [train.py:904] (0/8) Epoch 2, batch 8700, loss[loss=0.2092, simple_loss=0.2979, pruned_loss=0.0603, over 16701.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.329, pruned_loss=0.08757, over 3055915.34 frames. ], batch size: 83, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:08:49,483 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=18864.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:09:14,856 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18878.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:09:16,689 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=18879.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:10:01,024 INFO [train.py:904] (0/8) Epoch 2, batch 8750, loss[loss=0.243, simple_loss=0.3311, pruned_loss=0.07751, over 16450.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3265, pruned_loss=0.08568, over 3041844.16 frames. ], batch size: 68, lr: 2.48e-02, grad_scale: 2.0 2023-04-27 20:10:05,340 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7636, 2.6868, 1.5680, 2.8304, 2.1275, 2.8056, 1.8522, 2.4665], device='cuda:0'), covar=tensor([0.0081, 0.0250, 0.1253, 0.0069, 0.0667, 0.0321, 0.1065, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0076, 0.0110, 0.0170, 0.0075, 0.0155, 0.0135, 0.0178, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 20:10:57,518 INFO [optim.py:368] (0/8) 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 20:11:10,325 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=18930.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:11:36,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5458, 4.1352, 4.0196, 1.8018, 4.1289, 4.2311, 3.4681, 3.4618], device='cuda:0'), covar=tensor([0.0694, 0.0077, 0.0129, 0.1322, 0.0081, 0.0062, 0.0224, 0.0260], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0084, 0.0083, 0.0157, 0.0074, 0.0072, 0.0110, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 20:11:53,581 INFO [train.py:904] (0/8) Epoch 2, batch 8800, loss[loss=0.2757, simple_loss=0.3467, pruned_loss=0.1023, over 16752.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3254, pruned_loss=0.08463, over 3052961.99 frames. ], batch size: 134, lr: 2.48e-02, grad_scale: 4.0 2023-04-27 20:13:18,191 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=18991.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:13:38,730 INFO [train.py:904] (0/8) Epoch 2, batch 8850, loss[loss=0.2252, simple_loss=0.3024, pruned_loss=0.07402, over 12562.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3268, pruned_loss=0.08339, over 3045024.41 frames. ], batch size: 247, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:14:28,297 INFO [optim.py:368] (0/8) 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 20:14:43,430 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19030.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:15:27,245 INFO [train.py:904] (0/8) Epoch 2, batch 8900, loss[loss=0.268, simple_loss=0.3444, pruned_loss=0.09576, over 16792.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3259, pruned_loss=0.08195, over 3046731.74 frames. ], batch size: 124, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:15:30,995 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8250, 2.8769, 2.4596, 3.6798, 3.6709, 3.7338, 1.6889, 2.7677], device='cuda:0'), covar=tensor([0.1807, 0.0506, 0.1297, 0.0121, 0.0204, 0.0239, 0.1581, 0.0758], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0126, 0.0170, 0.0071, 0.0111, 0.0120, 0.0156, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:16:40,477 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2364, 4.5394, 4.2503, 4.3142, 4.0071, 4.0915, 4.0231, 4.5810], device='cuda:0'), covar=tensor([0.0521, 0.0703, 0.0847, 0.0357, 0.0501, 0.0669, 0.0538, 0.0592], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0292, 0.0259, 0.0188, 0.0201, 0.0186, 0.0243, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:17:12,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0798, 3.7722, 3.5500, 1.9099, 2.8623, 2.3798, 3.3551, 3.9406], device='cuda:0'), covar=tensor([0.0312, 0.0564, 0.0420, 0.1534, 0.0741, 0.0966, 0.0836, 0.0525], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0101, 0.0153, 0.0150, 0.0141, 0.0135, 0.0145, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:17:32,222 INFO [train.py:904] (0/8) Epoch 2, batch 8950, loss[loss=0.2677, simple_loss=0.3371, pruned_loss=0.09917, over 12891.00 frames. ], tot_loss[loss=0.2454, simple_loss=0.3259, pruned_loss=0.08246, over 3058318.20 frames. ], batch size: 247, lr: 2.47e-02, grad_scale: 4.0 2023-04-27 20:17:50,823 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3082, 3.0939, 3.2067, 3.5053, 3.4871, 3.1997, 3.4509, 3.4893], device='cuda:0'), covar=tensor([0.0422, 0.0466, 0.0901, 0.0339, 0.0375, 0.1010, 0.0492, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0280, 0.0369, 0.0271, 0.0210, 0.0192, 0.0220, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:18:20,857 INFO [optim.py:368] (0/8) 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 20:18:32,653 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19129.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:19:11,616 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19146.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:19:20,098 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-04-27 20:19:21,329 INFO [train.py:904] (0/8) Epoch 2, batch 9000, loss[loss=0.2178, simple_loss=0.3032, pruned_loss=0.06624, over 16621.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3225, pruned_loss=0.08037, over 3069581.86 frames. ], batch size: 134, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:19:21,330 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 20:19:31,140 INFO [train.py:938] (0/8) Epoch 2, validation: loss=0.2044, simple_loss=0.3047, pruned_loss=0.05205, over 944034.00 frames. 2023-04-27 20:19:31,142 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 20:20:00,952 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19164.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:20:52,274 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19190.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 20:21:02,064 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19194.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:21:13,860 INFO [train.py:904] (0/8) Epoch 2, batch 9050, loss[loss=0.2335, simple_loss=0.3068, pruned_loss=0.08013, over 16780.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3242, pruned_loss=0.08136, over 3086232.70 frames. ], batch size: 124, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:21:37,541 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19212.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:21:57,926 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8786, 3.2005, 2.6076, 4.1830, 4.0186, 3.9801, 1.9524, 3.1943], device='cuda:0'), covar=tensor([0.1318, 0.0349, 0.1043, 0.0053, 0.0175, 0.0274, 0.1163, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0122, 0.0165, 0.0067, 0.0110, 0.0119, 0.0153, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:22:01,200 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.755e+02 4.212e+02 4.597e+02 5.989e+02 9.578e+02, threshold=9.193e+02, percent-clipped=0.0 2023-04-27 20:22:58,697 INFO [train.py:904] (0/8) Epoch 2, batch 9100, loss[loss=0.2463, simple_loss=0.3315, pruned_loss=0.08059, over 16211.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3235, pruned_loss=0.08173, over 3084675.94 frames. ], batch size: 165, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:24:07,758 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-04-27 20:24:22,923 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19286.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:24:55,901 INFO [train.py:904] (0/8) Epoch 2, batch 9150, loss[loss=0.2646, simple_loss=0.3504, pruned_loss=0.08939, over 16855.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3245, pruned_loss=0.0816, over 3076835.78 frames. ], batch size: 42, lr: 2.46e-02, grad_scale: 2.0 2023-04-27 20:25:01,476 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.48 vs. limit=5.0 2023-04-27 20:25:49,332 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.714e+02 4.631e+02 5.471e+02 6.935e+02 2.154e+03, threshold=1.094e+03, percent-clipped=5.0 2023-04-27 20:25:58,991 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19330.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:26:40,145 INFO [train.py:904] (0/8) Epoch 2, batch 9200, loss[loss=0.2005, simple_loss=0.2762, pruned_loss=0.06243, over 12005.00 frames. ], tot_loss[loss=0.2387, simple_loss=0.3185, pruned_loss=0.07945, over 3078687.57 frames. ], batch size: 247, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:27:20,655 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.77 vs. limit=2.0 2023-04-27 20:27:24,621 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4431, 3.4321, 3.2299, 3.3167, 3.0426, 3.3138, 3.1141, 3.1691], device='cuda:0'), covar=tensor([0.0270, 0.0175, 0.0185, 0.0129, 0.0507, 0.0179, 0.0605, 0.0221], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0094, 0.0140, 0.0114, 0.0165, 0.0123, 0.0100, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:27:30,643 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19378.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:03,161 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19394.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:28:16,039 INFO [train.py:904] (0/8) Epoch 2, batch 9250, loss[loss=0.2276, simple_loss=0.311, pruned_loss=0.07206, over 15359.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3182, pruned_loss=0.07951, over 3069541.23 frames. ], batch size: 190, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:29:05,884 INFO [optim.py:368] (0/8) 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 20:29:14,570 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9698, 1.6450, 1.5653, 1.4454, 1.8210, 1.5803, 1.7997, 1.8536], device='cuda:0'), covar=tensor([0.0021, 0.0095, 0.0097, 0.0129, 0.0065, 0.0106, 0.0032, 0.0054], device='cuda:0'), in_proj_covar=tensor([0.0048, 0.0103, 0.0100, 0.0105, 0.0092, 0.0102, 0.0055, 0.0074], device='cuda:0'), out_proj_covar=tensor([6.6529e-05, 1.5664e-04, 1.4873e-04, 1.5900e-04, 1.4511e-04, 1.6025e-04, 8.1035e-05, 1.1749e-04], device='cuda:0') 2023-04-27 20:30:06,039 INFO [train.py:904] (0/8) Epoch 2, batch 9300, loss[loss=0.2019, simple_loss=0.2864, pruned_loss=0.0587, over 16354.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3156, pruned_loss=0.07748, over 3069894.83 frames. ], batch size: 146, lr: 2.45e-02, grad_scale: 4.0 2023-04-27 20:30:14,378 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19455.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:31:20,777 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19485.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 20:31:49,027 INFO [train.py:904] (0/8) Epoch 2, batch 9350, loss[loss=0.2123, simple_loss=0.2928, pruned_loss=0.06595, over 17061.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.3165, pruned_loss=0.07801, over 3073764.70 frames. ], batch size: 53, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:32:10,039 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4955, 1.4211, 1.6414, 2.2935, 2.4968, 2.4526, 1.4780, 2.2483], device='cuda:0'), covar=tensor([0.0042, 0.0189, 0.0126, 0.0077, 0.0032, 0.0055, 0.0168, 0.0062], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0110, 0.0094, 0.0081, 0.0064, 0.0059, 0.0098, 0.0055], device='cuda:0'), out_proj_covar=tensor([1.2374e-04, 1.8750e-04, 1.6634e-04, 1.4217e-04, 1.0538e-04, 9.8435e-05, 1.6416e-04, 9.0263e-05], device='cuda:0') 2023-04-27 20:32:37,320 INFO [optim.py:368] (0/8) 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 20:33:28,468 INFO [train.py:904] (0/8) Epoch 2, batch 9400, loss[loss=0.2516, simple_loss=0.3461, pruned_loss=0.07854, over 16657.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3165, pruned_loss=0.07796, over 3057635.50 frames. ], batch size: 134, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:33:33,602 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7580, 5.0258, 4.7631, 4.8177, 4.3566, 4.2579, 4.6022, 5.0214], device='cuda:0'), covar=tensor([0.0337, 0.0524, 0.0671, 0.0268, 0.0446, 0.0526, 0.0369, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0211, 0.0293, 0.0250, 0.0183, 0.0196, 0.0183, 0.0231, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:34:02,444 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4586, 1.7769, 1.5569, 1.3508, 2.1608, 1.8663, 2.2665, 2.2636], device='cuda:0'), covar=tensor([0.0020, 0.0152, 0.0146, 0.0185, 0.0084, 0.0146, 0.0034, 0.0069], device='cuda:0'), in_proj_covar=tensor([0.0050, 0.0103, 0.0101, 0.0106, 0.0092, 0.0102, 0.0055, 0.0074], device='cuda:0'), out_proj_covar=tensor([6.8507e-05, 1.5730e-04, 1.4908e-04, 1.6041e-04, 1.4347e-04, 1.5860e-04, 7.9870e-05, 1.1675e-04], device='cuda:0') 2023-04-27 20:34:30,655 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9841, 3.8199, 3.5538, 1.7772, 2.9841, 2.2547, 3.3584, 3.7975], device='cuda:0'), covar=tensor([0.0285, 0.0369, 0.0338, 0.1470, 0.0600, 0.0869, 0.0691, 0.0428], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0099, 0.0150, 0.0147, 0.0136, 0.0132, 0.0143, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:34:39,283 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19586.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:08,268 INFO [train.py:904] (0/8) Epoch 2, batch 9450, loss[loss=0.2252, simple_loss=0.3121, pruned_loss=0.06909, over 15455.00 frames. ], tot_loss[loss=0.238, simple_loss=0.3185, pruned_loss=0.07878, over 3053833.40 frames. ], batch size: 191, lr: 2.44e-02, grad_scale: 4.0 2023-04-27 20:35:42,246 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=19618.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:35:56,421 INFO [optim.py:368] (0/8) 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 20:36:10,027 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1482, 4.0998, 3.8924, 4.0214, 3.4401, 3.9526, 3.9528, 3.8191], device='cuda:0'), covar=tensor([0.0271, 0.0135, 0.0219, 0.0126, 0.0748, 0.0192, 0.0307, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0093, 0.0138, 0.0116, 0.0168, 0.0123, 0.0101, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:36:16,145 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19634.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:36:48,235 INFO [train.py:904] (0/8) Epoch 2, batch 9500, loss[loss=0.2373, simple_loss=0.3159, pruned_loss=0.0794, over 16589.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3168, pruned_loss=0.0775, over 3056287.41 frames. ], batch size: 62, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:37:45,594 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=19679.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:38:30,027 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 20:38:34,560 INFO [train.py:904] (0/8) Epoch 2, batch 9550, loss[loss=0.2702, simple_loss=0.3497, pruned_loss=0.09538, over 15525.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3176, pruned_loss=0.07847, over 3067506.14 frames. ], batch size: 192, lr: 2.43e-02, grad_scale: 4.0 2023-04-27 20:38:55,043 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2118, 3.2466, 1.3818, 3.2507, 2.1686, 3.2711, 1.8221, 2.5983], device='cuda:0'), covar=tensor([0.0062, 0.0187, 0.1672, 0.0044, 0.0788, 0.0341, 0.1344, 0.0564], device='cuda:0'), in_proj_covar=tensor([0.0075, 0.0113, 0.0169, 0.0075, 0.0152, 0.0135, 0.0176, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 20:39:23,657 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.908e+02 4.753e+02 5.956e+02 1.328e+03, threshold=9.507e+02, percent-clipped=2.0 2023-04-27 20:39:55,505 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-27 20:40:12,657 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19750.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:40:13,484 INFO [train.py:904] (0/8) Epoch 2, batch 9600, loss[loss=0.2552, simple_loss=0.3295, pruned_loss=0.09041, over 16926.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3182, pruned_loss=0.07854, over 3069882.62 frames. ], batch size: 109, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:41:22,053 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=19785.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 20:42:00,146 INFO [train.py:904] (0/8) Epoch 2, batch 9650, loss[loss=0.2389, simple_loss=0.3121, pruned_loss=0.08282, over 12223.00 frames. ], tot_loss[loss=0.2399, simple_loss=0.3208, pruned_loss=0.07955, over 3063521.90 frames. ], batch size: 248, lr: 2.43e-02, grad_scale: 8.0 2023-04-27 20:42:15,814 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8915, 4.0244, 1.5113, 3.7930, 2.4674, 3.9285, 1.8445, 2.8833], device='cuda:0'), covar=tensor([0.0035, 0.0118, 0.1762, 0.0042, 0.0887, 0.0263, 0.1484, 0.0562], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0113, 0.0167, 0.0073, 0.0151, 0.0133, 0.0173, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 20:42:54,375 INFO [optim.py:368] (0/8) 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 20:43:09,977 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=19833.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 20:43:48,664 INFO [train.py:904] (0/8) Epoch 2, batch 9700, loss[loss=0.223, simple_loss=0.3083, pruned_loss=0.06886, over 16908.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.32, pruned_loss=0.07963, over 3064798.79 frames. ], batch size: 116, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:45:08,489 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-27 20:45:33,548 INFO [train.py:904] (0/8) Epoch 2, batch 9750, loss[loss=0.2032, simple_loss=0.2872, pruned_loss=0.05965, over 16608.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3181, pruned_loss=0.07929, over 3055286.02 frames. ], batch size: 62, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:46:21,261 INFO [optim.py:368] (0/8) 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 20:47:15,372 INFO [train.py:904] (0/8) Epoch 2, batch 9800, loss[loss=0.214, simple_loss=0.3088, pruned_loss=0.05958, over 16751.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.3166, pruned_loss=0.07688, over 3074338.29 frames. ], batch size: 83, lr: 2.42e-02, grad_scale: 8.0 2023-04-27 20:48:00,424 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=19974.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:48:00,546 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4040, 4.1616, 4.0734, 4.2160, 3.7575, 4.2186, 4.1819, 3.9359], device='cuda:0'), covar=tensor([0.0288, 0.0226, 0.0204, 0.0134, 0.0566, 0.0195, 0.0228, 0.0267], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0092, 0.0140, 0.0115, 0.0163, 0.0122, 0.0101, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 20:48:16,004 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8488, 2.8059, 2.2900, 3.8551, 3.5943, 3.7943, 1.8329, 2.7009], device='cuda:0'), covar=tensor([0.1212, 0.0414, 0.1115, 0.0050, 0.0140, 0.0222, 0.1040, 0.0607], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0124, 0.0163, 0.0065, 0.0109, 0.0116, 0.0151, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:49:00,733 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-20000.pt 2023-04-27 20:49:05,348 INFO [train.py:904] (0/8) Epoch 2, batch 9850, loss[loss=0.237, simple_loss=0.3237, pruned_loss=0.07514, over 16590.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3181, pruned_loss=0.07648, over 3085415.49 frames. ], batch size: 75, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:49:32,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7384, 3.3800, 3.2722, 2.2713, 3.1969, 3.2114, 3.5369, 1.7272], device='cuda:0'), covar=tensor([0.0516, 0.0024, 0.0040, 0.0283, 0.0041, 0.0055, 0.0020, 0.0423], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0054, 0.0059, 0.0105, 0.0053, 0.0057, 0.0057, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 20:49:46,071 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20020.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:49:55,605 INFO [optim.py:368] (0/8) 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 20:50:47,869 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1898, 2.1909, 2.0777, 2.0176, 2.8064, 2.7844, 3.5449, 3.2390], device='cuda:0'), covar=tensor([0.0015, 0.0169, 0.0144, 0.0176, 0.0081, 0.0096, 0.0023, 0.0049], device='cuda:0'), in_proj_covar=tensor([0.0051, 0.0104, 0.0101, 0.0103, 0.0093, 0.0104, 0.0055, 0.0074], device='cuda:0'), out_proj_covar=tensor([6.8975e-05, 1.5813e-04, 1.4749e-04, 1.5468e-04, 1.4205e-04, 1.6116e-04, 7.9442e-05, 1.1568e-04], device='cuda:0') 2023-04-27 20:50:52,723 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9542, 2.3923, 2.2555, 3.2334, 3.0830, 3.2173, 2.0561, 2.5201], device='cuda:0'), covar=tensor([0.1586, 0.0578, 0.1232, 0.0089, 0.0210, 0.0331, 0.1178, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0126, 0.0164, 0.0065, 0.0111, 0.0118, 0.0153, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:50:58,132 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20050.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:50:58,955 INFO [train.py:904] (0/8) Epoch 2, batch 9900, loss[loss=0.2433, simple_loss=0.338, pruned_loss=0.07428, over 16958.00 frames. ], tot_loss[loss=0.2363, simple_loss=0.319, pruned_loss=0.07676, over 3076238.29 frames. ], batch size: 109, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:52:12,583 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20081.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:31,321 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:51,007 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:52:56,920 INFO [train.py:904] (0/8) Epoch 2, batch 9950, loss[loss=0.2466, simple_loss=0.3294, pruned_loss=0.08189, over 16338.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.321, pruned_loss=0.07686, over 3077848.98 frames. ], batch size: 146, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:53:23,645 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20112.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:53:55,572 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.288e+02 4.201e+02 4.972e+02 6.369e+02 1.989e+03, threshold=9.943e+02, percent-clipped=8.0 2023-04-27 20:54:55,894 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:54:56,745 INFO [train.py:904] (0/8) Epoch 2, batch 10000, loss[loss=0.2077, simple_loss=0.3032, pruned_loss=0.05613, over 16228.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3184, pruned_loss=0.07572, over 3094684.42 frames. ], batch size: 165, lr: 2.41e-02, grad_scale: 8.0 2023-04-27 20:55:41,254 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20173.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:56:37,816 INFO [train.py:904] (0/8) Epoch 2, batch 10050, loss[loss=0.2467, simple_loss=0.326, pruned_loss=0.08373, over 16929.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3186, pruned_loss=0.07597, over 3083774.87 frames. ], batch size: 109, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:56:44,663 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:57:08,797 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6767, 1.3282, 1.8110, 2.4866, 2.4355, 2.3506, 1.4443, 2.3908], device='cuda:0'), covar=tensor([0.0038, 0.0234, 0.0114, 0.0090, 0.0043, 0.0079, 0.0173, 0.0055], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0115, 0.0101, 0.0087, 0.0068, 0.0061, 0.0102, 0.0056], device='cuda:0'), out_proj_covar=tensor([1.3023e-04, 1.9320e-04, 1.7457e-04, 1.5114e-04, 1.1129e-04, 1.0022e-04, 1.6783e-04, 9.0345e-05], device='cuda:0') 2023-04-27 20:57:20,865 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5814, 3.4609, 3.1686, 1.7881, 2.6233, 2.1497, 2.8756, 3.3622], device='cuda:0'), covar=tensor([0.0302, 0.0377, 0.0402, 0.1508, 0.0722, 0.0941, 0.0753, 0.0497], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0099, 0.0152, 0.0148, 0.0138, 0.0132, 0.0143, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 20:57:24,816 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 2, batch 10100, loss[loss=0.2285, simple_loss=0.3035, pruned_loss=0.07671, over 12969.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.32, pruned_loss=0.07729, over 3068106.28 frames. ], batch size: 248, lr: 2.40e-02, grad_scale: 8.0 2023-04-27 20:58:38,032 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20265.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:58:58,377 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20274.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 20:59:30,269 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-2.pt 2023-04-27 20:59:54,876 INFO [train.py:904] (0/8) Epoch 3, batch 0, loss[loss=0.2485, simple_loss=0.3102, pruned_loss=0.09339, over 17011.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3102, pruned_loss=0.09339, over 17011.00 frames. ], batch size: 41, lr: 2.28e-02, grad_scale: 8.0 2023-04-27 20:59:54,877 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 21:00:02,297 INFO [train.py:938] (0/8) Epoch 3, validation: loss=0.2012, simple_loss=0.3019, pruned_loss=0.05024, over 944034.00 frames. 2023-04-27 21:00:02,303 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 21:00:13,644 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9067, 4.6192, 4.8460, 5.1025, 5.2263, 4.6153, 5.3166, 5.1269], device='cuda:0'), covar=tensor([0.0573, 0.0639, 0.1102, 0.0539, 0.0456, 0.0314, 0.0397, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0280, 0.0362, 0.0276, 0.0210, 0.0193, 0.0216, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:00:31,874 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20322.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:00:38,801 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.937e+02 4.805e+02 5.776e+02 7.144e+02 1.042e+03, threshold=1.155e+03, percent-clipped=3.0 2023-04-27 21:01:12,901 INFO [train.py:904] (0/8) Epoch 3, batch 50, loss[loss=0.2783, simple_loss=0.3469, pruned_loss=0.1049, over 17093.00 frames. ], tot_loss[loss=0.2866, simple_loss=0.346, pruned_loss=0.1136, over 746669.98 frames. ], batch size: 53, lr: 2.28e-02, grad_scale: 2.0 2023-04-27 21:01:45,735 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20376.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:02,745 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20388.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:16,437 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7361, 3.6368, 3.5401, 2.3634, 3.4360, 3.3591, 3.6661, 1.7072], device='cuda:0'), covar=tensor([0.0462, 0.0038, 0.0074, 0.0278, 0.0051, 0.0096, 0.0049, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0056, 0.0061, 0.0108, 0.0055, 0.0060, 0.0059, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:02:19,624 INFO [train.py:904] (0/8) Epoch 3, batch 100, loss[loss=0.2087, simple_loss=0.2868, pruned_loss=0.06528, over 17217.00 frames. ], tot_loss[loss=0.2749, simple_loss=0.3379, pruned_loss=0.106, over 1324432.17 frames. ], batch size: 44, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:02:35,453 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7027, 5.0064, 5.0469, 5.0138, 4.9230, 5.5192, 5.2848, 5.0110], device='cuda:0'), covar=tensor([0.0883, 0.1275, 0.1016, 0.1411, 0.1894, 0.0786, 0.0872, 0.1645], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0302, 0.0278, 0.0266, 0.0335, 0.0293, 0.0225, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:02:43,353 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20418.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:02:54,677 INFO [optim.py:368] (0/8) 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 21:03:13,733 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8839, 2.7546, 2.4228, 4.2746, 1.8843, 4.2380, 2.3730, 2.3987], device='cuda:0'), covar=tensor([0.0295, 0.0568, 0.0394, 0.0198, 0.1607, 0.0165, 0.0781, 0.1353], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0204, 0.0167, 0.0228, 0.0273, 0.0175, 0.0201, 0.0257], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:03:18,105 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20445.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:24,373 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20449.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:27,033 INFO [train.py:904] (0/8) Epoch 3, batch 150, loss[loss=0.234, simple_loss=0.3177, pruned_loss=0.07516, over 17111.00 frames. ], tot_loss[loss=0.267, simple_loss=0.3326, pruned_loss=0.1007, over 1772295.25 frames. ], batch size: 47, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:03:49,102 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20468.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:03:54,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1939, 1.6616, 2.2441, 2.7060, 2.8830, 3.1731, 1.7049, 2.9905], device='cuda:0'), covar=tensor([0.0039, 0.0199, 0.0122, 0.0095, 0.0045, 0.0067, 0.0165, 0.0054], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0117, 0.0104, 0.0092, 0.0071, 0.0065, 0.0103, 0.0058], device='cuda:0'), out_proj_covar=tensor([1.3500e-04, 1.9712e-04, 1.7981e-04, 1.6023e-04, 1.1506e-04, 1.0737e-04, 1.6986e-04, 9.5703e-05], device='cuda:0') 2023-04-27 21:03:57,406 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1661, 5.5259, 5.1474, 5.4224, 4.8811, 4.6449, 4.9834, 5.6304], device='cuda:0'), covar=tensor([0.0465, 0.0573, 0.0815, 0.0334, 0.0535, 0.0498, 0.0528, 0.0524], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0349, 0.0307, 0.0213, 0.0232, 0.0215, 0.0275, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:04:05,879 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20479.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:04:35,433 INFO [train.py:904] (0/8) Epoch 3, batch 200, loss[loss=0.3425, simple_loss=0.3706, pruned_loss=0.1572, over 16898.00 frames. ], tot_loss[loss=0.2693, simple_loss=0.3343, pruned_loss=0.1021, over 2110648.93 frames. ], batch size: 109, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:09,763 INFO [optim.py:368] (0/8) 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 21:05:43,635 INFO [train.py:904] (0/8) Epoch 3, batch 250, loss[loss=0.2645, simple_loss=0.3092, pruned_loss=0.1099, over 16490.00 frames. ], tot_loss[loss=0.2664, simple_loss=0.3307, pruned_loss=0.1011, over 2375704.24 frames. ], batch size: 146, lr: 2.27e-02, grad_scale: 2.0 2023-04-27 21:05:58,418 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20560.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:35,601 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=20588.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:06:54,935 INFO [train.py:904] (0/8) Epoch 3, batch 300, loss[loss=0.2101, simple_loss=0.2907, pruned_loss=0.06473, over 17260.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.3267, pruned_loss=0.09841, over 2581653.24 frames. ], batch size: 43, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:07:29,015 INFO [optim.py:368] (0/8) 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 21:07:59,184 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=20649.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:08:01,110 INFO [train.py:904] (0/8) Epoch 3, batch 350, loss[loss=0.2665, simple_loss=0.3146, pruned_loss=0.1093, over 16232.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3217, pruned_loss=0.09462, over 2757093.86 frames. ], batch size: 165, lr: 2.26e-02, grad_scale: 2.0 2023-04-27 21:08:36,843 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20676.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:09,382 INFO [train.py:904] (0/8) Epoch 3, batch 400, loss[loss=0.1993, simple_loss=0.2861, pruned_loss=0.05626, over 17137.00 frames. ], tot_loss[loss=0.251, simple_loss=0.3174, pruned_loss=0.09227, over 2886554.02 frames. ], batch size: 48, lr: 2.26e-02, grad_scale: 4.0 2023-04-27 21:09:41,231 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20724.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:09:45,248 INFO [optim.py:368] (0/8) 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 21:10:08,722 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20744.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:10,761 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20745.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:17,853 INFO [train.py:904] (0/8) Epoch 3, batch 450, loss[loss=0.2537, simple_loss=0.3133, pruned_loss=0.09702, over 16835.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3164, pruned_loss=0.09226, over 2982586.41 frames. ], batch size: 102, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:10:31,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6770, 3.6541, 1.6626, 3.7598, 2.5226, 3.7240, 1.9372, 2.8736], device='cuda:0'), covar=tensor([0.0050, 0.0197, 0.1383, 0.0037, 0.0662, 0.0355, 0.1087, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0127, 0.0171, 0.0078, 0.0157, 0.0153, 0.0177, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 21:10:41,886 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20768.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:10:49,448 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:14,832 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20793.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:11:24,058 INFO [train.py:904] (0/8) Epoch 3, batch 500, loss[loss=0.2073, simple_loss=0.2833, pruned_loss=0.06561, over 16838.00 frames. ], tot_loss[loss=0.2493, simple_loss=0.3156, pruned_loss=0.09148, over 3060838.68 frames. ], batch size: 42, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:11:46,148 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20816.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:12:00,590 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 3, batch 550, loss[loss=0.2231, simple_loss=0.2975, pruned_loss=0.07432, over 17226.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3144, pruned_loss=0.09055, over 3109301.78 frames. ], batch size: 44, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:12:45,316 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=20860.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:13:40,835 INFO [train.py:904] (0/8) Epoch 3, batch 600, loss[loss=0.2752, simple_loss=0.3569, pruned_loss=0.09674, over 17291.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3133, pruned_loss=0.08977, over 3154122.39 frames. ], batch size: 52, lr: 2.25e-02, grad_scale: 4.0 2023-04-27 21:13:50,525 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=20908.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:15,689 INFO [optim.py:368] (0/8) 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 21:14:36,130 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7303, 2.5966, 2.4453, 4.0902, 2.0278, 3.7885, 2.3126, 2.4026], device='cuda:0'), covar=tensor([0.0263, 0.0563, 0.0382, 0.0177, 0.1395, 0.0198, 0.0758, 0.1036], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0209, 0.0174, 0.0232, 0.0279, 0.0181, 0.0205, 0.0269], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:14:38,332 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=20944.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:14:48,037 INFO [train.py:904] (0/8) Epoch 3, batch 650, loss[loss=0.2385, simple_loss=0.3174, pruned_loss=0.07985, over 16617.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.312, pruned_loss=0.08922, over 3199434.90 frames. ], batch size: 57, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:15:39,370 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 21:15:57,320 INFO [train.py:904] (0/8) Epoch 3, batch 700, loss[loss=0.2692, simple_loss=0.3163, pruned_loss=0.1111, over 16933.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3099, pruned_loss=0.08773, over 3232579.67 frames. ], batch size: 109, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:16:30,870 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.136e+02 3.749e+02 4.303e+02 5.331e+02 1.033e+03, threshold=8.606e+02, percent-clipped=0.0 2023-04-27 21:16:55,497 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21044.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:16:58,152 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 21:17:03,825 INFO [train.py:904] (0/8) Epoch 3, batch 750, loss[loss=0.2533, simple_loss=0.3089, pruned_loss=0.0988, over 16776.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3108, pruned_loss=0.08799, over 3256383.84 frames. ], batch size: 102, lr: 2.24e-02, grad_scale: 4.0 2023-04-27 21:17:06,026 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21052.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:34,401 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21074.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:17:48,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5567, 4.6202, 4.5320, 4.7441, 4.5437, 5.1572, 4.8947, 4.5332], device='cuda:0'), covar=tensor([0.0921, 0.1080, 0.1115, 0.1434, 0.2322, 0.0862, 0.0878, 0.1916], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0308, 0.0285, 0.0271, 0.0355, 0.0310, 0.0238, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:17:57,837 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21092.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:10,823 INFO [train.py:904] (0/8) Epoch 3, batch 800, loss[loss=0.2405, simple_loss=0.3201, pruned_loss=0.08039, over 16696.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3117, pruned_loss=0.08905, over 3272544.47 frames. ], batch size: 62, lr: 2.24e-02, grad_scale: 8.0 2023-04-27 21:18:26,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3943, 2.0886, 1.9833, 2.2116, 2.7992, 2.7713, 3.8086, 3.2742], device='cuda:0'), covar=tensor([0.0018, 0.0164, 0.0166, 0.0162, 0.0091, 0.0130, 0.0034, 0.0060], device='cuda:0'), in_proj_covar=tensor([0.0058, 0.0111, 0.0111, 0.0112, 0.0103, 0.0113, 0.0066, 0.0084], device='cuda:0'), out_proj_covar=tensor([8.1245e-05, 1.6684e-04, 1.6087e-04, 1.6491e-04, 1.5857e-04, 1.7239e-04, 9.9045e-05, 1.3130e-04], device='cuda:0') 2023-04-27 21:18:27,394 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21113.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:34,855 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:39,529 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21122.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:18:47,527 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.745e+02 4.034e+02 4.669e+02 5.861e+02 1.175e+03, threshold=9.338e+02, percent-clipped=5.0 2023-04-27 21:19:20,183 INFO [train.py:904] (0/8) Epoch 3, batch 850, loss[loss=0.2536, simple_loss=0.3087, pruned_loss=0.09923, over 16739.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3097, pruned_loss=0.08812, over 3287322.12 frames. ], batch size: 124, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:19:58,147 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3886, 5.2481, 5.1556, 4.5016, 5.0690, 2.5960, 4.7995, 5.3007], device='cuda:0'), covar=tensor([0.0048, 0.0045, 0.0055, 0.0242, 0.0047, 0.0928, 0.0063, 0.0066], device='cuda:0'), in_proj_covar=tensor([0.0075, 0.0061, 0.0096, 0.0110, 0.0070, 0.0119, 0.0083, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:19:58,198 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21179.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:20:27,607 INFO [train.py:904] (0/8) Epoch 3, batch 900, loss[loss=0.2563, simple_loss=0.3153, pruned_loss=0.09863, over 16775.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3086, pruned_loss=0.08717, over 3296070.72 frames. ], batch size: 116, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:21:03,018 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.660e+02 3.821e+02 4.761e+02 5.619e+02 1.173e+03, threshold=9.521e+02, percent-clipped=3.0 2023-04-27 21:21:26,784 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21244.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:21:35,850 INFO [train.py:904] (0/8) Epoch 3, batch 950, loss[loss=0.1855, simple_loss=0.2619, pruned_loss=0.05458, over 16763.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3082, pruned_loss=0.08626, over 3305706.29 frames. ], batch size: 39, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:22:32,677 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4730, 3.1726, 2.6380, 2.3543, 2.3320, 1.9917, 3.1717, 3.6504], device='cuda:0'), covar=tensor([0.1854, 0.0666, 0.1084, 0.0977, 0.2016, 0.1318, 0.0451, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0234, 0.0248, 0.0196, 0.0267, 0.0188, 0.0210, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:22:34,057 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21292.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:22:44,942 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2375, 1.5974, 2.4375, 3.0063, 2.8574, 3.7227, 1.8286, 3.1813], device='cuda:0'), covar=tensor([0.0049, 0.0195, 0.0108, 0.0089, 0.0051, 0.0030, 0.0155, 0.0035], device='cuda:0'), in_proj_covar=tensor([0.0083, 0.0118, 0.0105, 0.0099, 0.0077, 0.0066, 0.0106, 0.0060], device='cuda:0'), out_proj_covar=tensor([1.3829e-04, 1.9764e-04, 1.7997e-04, 1.6991e-04, 1.2372e-04, 1.0963e-04, 1.7266e-04, 9.8717e-05], device='cuda:0') 2023-04-27 21:22:45,630 INFO [train.py:904] (0/8) Epoch 3, batch 1000, loss[loss=0.2286, simple_loss=0.3002, pruned_loss=0.07854, over 17217.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.307, pruned_loss=0.08631, over 3305942.24 frames. ], batch size: 44, lr: 2.23e-02, grad_scale: 8.0 2023-04-27 21:23:00,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1385, 3.4302, 3.3636, 1.5956, 3.4693, 3.5510, 2.9000, 2.8372], device='cuda:0'), covar=tensor([0.0739, 0.0098, 0.0197, 0.1212, 0.0083, 0.0065, 0.0363, 0.0348], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0082, 0.0084, 0.0151, 0.0079, 0.0074, 0.0114, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-27 21:23:01,377 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21312.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:21,767 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.770e+02 3.937e+02 4.742e+02 6.159e+02 9.838e+02, threshold=9.485e+02, percent-clipped=2.0 2023-04-27 21:23:24,094 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21328.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:23:24,459 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-27 21:23:55,383 INFO [train.py:904] (0/8) Epoch 3, batch 1050, loss[loss=0.2546, simple_loss=0.3305, pruned_loss=0.0894, over 17085.00 frames. ], tot_loss[loss=0.2406, simple_loss=0.3077, pruned_loss=0.08674, over 3304878.45 frames. ], batch size: 53, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:24:25,419 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21373.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:24:26,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1190, 5.8063, 5.7091, 5.6623, 5.5514, 6.1162, 5.8307, 5.6228], device='cuda:0'), covar=tensor([0.0476, 0.0984, 0.0803, 0.1018, 0.2137, 0.0680, 0.0656, 0.1396], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0304, 0.0287, 0.0269, 0.0355, 0.0308, 0.0244, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:24:46,284 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21389.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:02,822 INFO [train.py:904] (0/8) Epoch 3, batch 1100, loss[loss=0.2132, simple_loss=0.2881, pruned_loss=0.06911, over 17169.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3082, pruned_loss=0.08704, over 3300374.15 frames. ], batch size: 46, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:25:12,311 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21408.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:25:38,415 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.639e+02 4.233e+02 5.080e+02 6.558e+02 1.311e+03, threshold=1.016e+03, percent-clipped=3.0 2023-04-27 21:25:57,449 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1745, 2.8189, 2.5144, 4.4708, 1.9726, 4.4782, 2.4376, 2.3462], device='cuda:0'), covar=tensor([0.0283, 0.0658, 0.0435, 0.0180, 0.1689, 0.0151, 0.0851, 0.1593], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0214, 0.0179, 0.0242, 0.0282, 0.0184, 0.0206, 0.0275], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:26:10,402 INFO [train.py:904] (0/8) Epoch 3, batch 1150, loss[loss=0.2016, simple_loss=0.2778, pruned_loss=0.06267, over 17213.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3075, pruned_loss=0.08561, over 3302650.01 frames. ], batch size: 44, lr: 2.22e-02, grad_scale: 4.0 2023-04-27 21:26:42,094 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21474.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:26:47,020 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4465, 2.9777, 2.5988, 2.3409, 2.2893, 2.0670, 2.9119, 3.1365], device='cuda:0'), covar=tensor([0.1431, 0.0616, 0.0884, 0.0737, 0.1597, 0.1079, 0.0399, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0234, 0.0244, 0.0193, 0.0266, 0.0185, 0.0209, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:27:19,797 INFO [train.py:904] (0/8) Epoch 3, batch 1200, loss[loss=0.2753, simple_loss=0.3128, pruned_loss=0.1189, over 16872.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3071, pruned_loss=0.08567, over 3306607.42 frames. ], batch size: 90, lr: 2.22e-02, grad_scale: 8.0 2023-04-27 21:27:28,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5106, 3.4897, 1.6936, 3.6142, 2.4433, 3.6045, 1.8570, 2.7731], device='cuda:0'), covar=tensor([0.0073, 0.0227, 0.1346, 0.0056, 0.0700, 0.0307, 0.1178, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0130, 0.0169, 0.0082, 0.0156, 0.0159, 0.0179, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-27 21:27:56,781 INFO [optim.py:368] (0/8) 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 21:28:27,752 INFO [train.py:904] (0/8) Epoch 3, batch 1250, loss[loss=0.262, simple_loss=0.3094, pruned_loss=0.1073, over 16926.00 frames. ], tot_loss[loss=0.2393, simple_loss=0.3072, pruned_loss=0.08575, over 3315881.32 frames. ], batch size: 109, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:28:59,773 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4456, 4.2326, 4.2949, 4.3404, 3.8840, 4.3341, 4.2916, 4.0817], device='cuda:0'), covar=tensor([0.0330, 0.0238, 0.0163, 0.0129, 0.0699, 0.0192, 0.0279, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0125, 0.0189, 0.0154, 0.0223, 0.0165, 0.0132, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:29:30,699 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21595.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:29:39,782 INFO [train.py:904] (0/8) Epoch 3, batch 1300, loss[loss=0.2225, simple_loss=0.3017, pruned_loss=0.07164, over 17201.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3058, pruned_loss=0.08492, over 3315553.27 frames. ], batch size: 44, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:29:56,022 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21613.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:30:17,383 INFO [optim.py:368] (0/8) 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 21:30:49,967 INFO [train.py:904] (0/8) Epoch 3, batch 1350, loss[loss=0.1983, simple_loss=0.2717, pruned_loss=0.06243, over 15775.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3059, pruned_loss=0.0845, over 3316724.76 frames. ], batch size: 35, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:30:56,178 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21656.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:13,058 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21668.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:20,899 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:35,231 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21684.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:31:58,968 INFO [train.py:904] (0/8) Epoch 3, batch 1400, loss[loss=0.1933, simple_loss=0.2634, pruned_loss=0.06157, over 16771.00 frames. ], tot_loss[loss=0.2372, simple_loss=0.3057, pruned_loss=0.08436, over 3316757.58 frames. ], batch size: 39, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:32:09,171 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21708.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:32:35,079 INFO [optim.py:368] (0/8) 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 21:33:07,174 INFO [train.py:904] (0/8) Epoch 3, batch 1450, loss[loss=0.1978, simple_loss=0.2691, pruned_loss=0.06322, over 16362.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3048, pruned_loss=0.08368, over 3315035.05 frames. ], batch size: 36, lr: 2.21e-02, grad_scale: 8.0 2023-04-27 21:33:13,878 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21756.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:24,047 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21763.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:33:39,398 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21774.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:14,646 INFO [train.py:904] (0/8) Epoch 3, batch 1500, loss[loss=0.2321, simple_loss=0.3173, pruned_loss=0.07351, over 17051.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.305, pruned_loss=0.08413, over 3316708.63 frames. ], batch size: 53, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:34:43,976 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=21822.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:47,175 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21824.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:34:52,716 INFO [optim.py:368] (0/8) 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 21:35:13,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3133, 4.1432, 3.4612, 1.9298, 2.6269, 2.1936, 3.7096, 4.1176], device='cuda:0'), covar=tensor([0.0183, 0.0388, 0.0453, 0.1427, 0.0692, 0.0972, 0.0440, 0.0384], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0120, 0.0153, 0.0144, 0.0137, 0.0129, 0.0144, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 21:35:16,472 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9760, 2.0181, 1.8828, 1.7265, 2.6582, 2.5546, 3.5607, 3.0563], device='cuda:0'), covar=tensor([0.0029, 0.0165, 0.0167, 0.0201, 0.0087, 0.0146, 0.0039, 0.0071], device='cuda:0'), in_proj_covar=tensor([0.0060, 0.0115, 0.0114, 0.0117, 0.0107, 0.0116, 0.0072, 0.0088], device='cuda:0'), out_proj_covar=tensor([8.6248e-05, 1.7191e-04, 1.6320e-04, 1.7295e-04, 1.6227e-04, 1.7626e-04, 1.0718e-04, 1.3680e-04], device='cuda:0') 2023-04-27 21:35:23,695 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21850.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:35:24,408 INFO [train.py:904] (0/8) Epoch 3, batch 1550, loss[loss=0.2587, simple_loss=0.3358, pruned_loss=0.09083, over 16721.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3071, pruned_loss=0.08625, over 3320507.78 frames. ], batch size: 62, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:35:36,326 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-04-27 21:35:40,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6330, 2.5094, 2.3223, 2.3160, 3.0544, 2.8769, 3.8790, 3.4699], device='cuda:0'), covar=tensor([0.0020, 0.0125, 0.0137, 0.0154, 0.0068, 0.0129, 0.0060, 0.0062], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0115, 0.0114, 0.0117, 0.0107, 0.0117, 0.0073, 0.0089], device='cuda:0'), out_proj_covar=tensor([8.6925e-05, 1.7223e-04, 1.6381e-04, 1.7347e-04, 1.6267e-04, 1.7752e-04, 1.0822e-04, 1.3817e-04], device='cuda:0') 2023-04-27 21:36:31,492 INFO [train.py:904] (0/8) Epoch 3, batch 1600, loss[loss=0.2743, simple_loss=0.3434, pruned_loss=0.1025, over 16603.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3103, pruned_loss=0.08731, over 3308714.02 frames. ], batch size: 68, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:36:46,204 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=21911.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:37:01,607 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8245, 3.8779, 1.8873, 3.8867, 2.5008, 3.8713, 1.9904, 2.9113], device='cuda:0'), covar=tensor([0.0052, 0.0171, 0.1383, 0.0065, 0.0756, 0.0333, 0.1231, 0.0516], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0136, 0.0171, 0.0084, 0.0157, 0.0165, 0.0179, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-27 21:37:07,929 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.625e+02 3.946e+02 4.752e+02 6.203e+02 1.123e+03, threshold=9.504e+02, percent-clipped=2.0 2023-04-27 21:37:38,903 INFO [train.py:904] (0/8) Epoch 3, batch 1650, loss[loss=0.1994, simple_loss=0.2768, pruned_loss=0.06106, over 17191.00 frames. ], tot_loss[loss=0.2441, simple_loss=0.3119, pruned_loss=0.08815, over 3314362.91 frames. ], batch size: 43, lr: 2.20e-02, grad_scale: 8.0 2023-04-27 21:37:39,890 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21951.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:01,307 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21968.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:03,301 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=21969.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:04,748 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=21970.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:16,734 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6234, 4.3665, 4.3683, 1.9433, 4.5435, 4.4391, 3.5421, 3.5697], device='cuda:0'), covar=tensor([0.0669, 0.0080, 0.0141, 0.1157, 0.0054, 0.0055, 0.0216, 0.0282], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0082, 0.0080, 0.0148, 0.0075, 0.0074, 0.0112, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-27 21:38:24,243 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=21984.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:38:44,042 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6117, 4.1978, 4.5017, 3.1723, 4.1983, 4.6274, 4.3206, 2.5531], device='cuda:0'), covar=tensor([0.0371, 0.0037, 0.0025, 0.0226, 0.0028, 0.0022, 0.0020, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0058, 0.0062, 0.0111, 0.0056, 0.0061, 0.0062, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:38:46,348 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-22000.pt 2023-04-27 21:38:50,893 INFO [train.py:904] (0/8) Epoch 3, batch 1700, loss[loss=0.3935, simple_loss=0.4237, pruned_loss=0.1816, over 12497.00 frames. ], tot_loss[loss=0.247, simple_loss=0.315, pruned_loss=0.08953, over 3311520.13 frames. ], batch size: 248, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:38:52,747 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-27 21:38:59,140 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8750, 3.7099, 2.6769, 4.7396, 4.4948, 4.3464, 1.7484, 3.2572], device='cuda:0'), covar=tensor([0.1224, 0.0331, 0.1075, 0.0048, 0.0208, 0.0213, 0.1159, 0.0588], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0129, 0.0162, 0.0073, 0.0142, 0.0134, 0.0153, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 21:39:11,669 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22016.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:28,010 INFO [optim.py:368] (0/8) 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 21:39:30,872 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22031.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:33,271 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22032.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:39:57,287 INFO [train.py:904] (0/8) Epoch 3, batch 1750, loss[loss=0.2172, simple_loss=0.2879, pruned_loss=0.07328, over 16837.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.3167, pruned_loss=0.0904, over 3305659.13 frames. ], batch size: 42, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:40:40,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6904, 2.6373, 2.4030, 4.1055, 1.8947, 3.7891, 2.2212, 2.5071], device='cuda:0'), covar=tensor([0.0299, 0.0596, 0.0416, 0.0165, 0.1621, 0.0204, 0.0857, 0.1086], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0215, 0.0180, 0.0241, 0.0284, 0.0186, 0.0206, 0.0276], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:41:02,176 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-27 21:41:06,921 INFO [train.py:904] (0/8) Epoch 3, batch 1800, loss[loss=0.2242, simple_loss=0.3072, pruned_loss=0.07058, over 17219.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.317, pruned_loss=0.08999, over 3308302.53 frames. ], batch size: 52, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:41:30,489 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 21:41:45,848 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22130.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:41:49,402 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7432, 4.5766, 1.9113, 4.7333, 2.7534, 4.6444, 2.4224, 3.3316], device='cuda:0'), covar=tensor([0.0024, 0.0116, 0.1266, 0.0024, 0.0655, 0.0218, 0.1007, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0135, 0.0171, 0.0082, 0.0155, 0.0163, 0.0178, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-27 21:42:14,096 INFO [train.py:904] (0/8) Epoch 3, batch 1850, loss[loss=0.2402, simple_loss=0.3264, pruned_loss=0.07706, over 17034.00 frames. ], tot_loss[loss=0.2482, simple_loss=0.3172, pruned_loss=0.08959, over 3307027.33 frames. ], batch size: 50, lr: 2.19e-02, grad_scale: 4.0 2023-04-27 21:42:47,438 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22176.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:09,365 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22191.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:22,348 INFO [train.py:904] (0/8) Epoch 3, batch 1900, loss[loss=0.224, simple_loss=0.3, pruned_loss=0.07402, over 17125.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3167, pruned_loss=0.08909, over 3305782.37 frames. ], batch size: 48, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:43:31,016 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22206.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:43:42,603 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 21:44:06,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0607, 5.4409, 5.0956, 5.2857, 4.6854, 4.5924, 4.9051, 5.5126], device='cuda:0'), covar=tensor([0.0548, 0.0548, 0.0689, 0.0318, 0.0544, 0.0543, 0.0451, 0.0572], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0378, 0.0328, 0.0232, 0.0246, 0.0228, 0.0297, 0.0257], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:44:12,264 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:12,281 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22237.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:32,425 INFO [train.py:904] (0/8) Epoch 3, batch 1950, loss[loss=0.2726, simple_loss=0.3308, pruned_loss=0.1072, over 16770.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3173, pruned_loss=0.08878, over 3294822.10 frames. ], batch size: 116, lr: 2.18e-02, grad_scale: 4.0 2023-04-27 21:44:32,732 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:44:58,069 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22269.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:06,180 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22276.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:08,594 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 21:45:30,750 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2238, 3.9736, 3.6208, 2.0221, 3.0993, 2.1447, 3.9275, 3.9871], device='cuda:0'), covar=tensor([0.0169, 0.0341, 0.0363, 0.1343, 0.0525, 0.0931, 0.0356, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0121, 0.0153, 0.0146, 0.0137, 0.0130, 0.0147, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 21:45:35,275 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22298.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:36,247 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22299.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:45:38,318 INFO [train.py:904] (0/8) Epoch 3, batch 2000, loss[loss=0.262, simple_loss=0.3117, pruned_loss=0.1061, over 16887.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.317, pruned_loss=0.08889, over 3302952.25 frames. ], batch size: 96, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:45:56,839 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3172, 4.1878, 3.1903, 5.2467, 5.2645, 4.6458, 2.5122, 3.7319], device='cuda:0'), covar=tensor([0.1098, 0.0318, 0.0802, 0.0085, 0.0136, 0.0227, 0.0958, 0.0468], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0129, 0.0160, 0.0075, 0.0142, 0.0134, 0.0152, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 21:46:01,056 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22317.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:13,072 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22326.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:46:17,542 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.887e+02 3.938e+02 4.847e+02 6.098e+02 1.059e+03, threshold=9.693e+02, percent-clipped=1.0 2023-04-27 21:46:46,970 INFO [train.py:904] (0/8) Epoch 3, batch 2050, loss[loss=0.2258, simple_loss=0.3089, pruned_loss=0.07132, over 17090.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3161, pruned_loss=0.0877, over 3318256.47 frames. ], batch size: 50, lr: 2.18e-02, grad_scale: 8.0 2023-04-27 21:47:06,388 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4112, 5.2033, 5.0586, 4.5308, 5.0310, 2.6685, 4.8678, 5.3268], device='cuda:0'), covar=tensor([0.0045, 0.0044, 0.0059, 0.0260, 0.0047, 0.0964, 0.0063, 0.0069], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0065, 0.0102, 0.0117, 0.0074, 0.0119, 0.0092, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:47:33,610 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22385.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:47:55,919 INFO [train.py:904] (0/8) Epoch 3, batch 2100, loss[loss=0.2252, simple_loss=0.3093, pruned_loss=0.07056, over 17122.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3171, pruned_loss=0.08899, over 3314443.76 frames. ], batch size: 49, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:48:17,982 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.99 vs. limit=5.0 2023-04-27 21:48:20,693 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22419.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:48:21,861 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4808, 2.3185, 1.9835, 2.2141, 2.8969, 2.7403, 3.9435, 3.2962], device='cuda:0'), covar=tensor([0.0020, 0.0121, 0.0133, 0.0124, 0.0062, 0.0099, 0.0028, 0.0059], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0117, 0.0115, 0.0117, 0.0109, 0.0116, 0.0075, 0.0093], device='cuda:0'), out_proj_covar=tensor([8.8985e-05, 1.7438e-04, 1.6411e-04, 1.7223e-04, 1.6445e-04, 1.7513e-04, 1.1092e-04, 1.4358e-04], device='cuda:0') 2023-04-27 21:48:26,427 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4346, 3.3784, 2.5323, 2.2377, 2.4042, 1.8470, 3.4202, 3.5469], device='cuda:0'), covar=tensor([0.1802, 0.0539, 0.1053, 0.0954, 0.1878, 0.1363, 0.0330, 0.0410], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0234, 0.0248, 0.0203, 0.0276, 0.0187, 0.0211, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:48:31,790 INFO [optim.py:368] (0/8) 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 21:48:44,124 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.82 vs. limit=5.0 2023-04-27 21:48:55,101 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22446.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:00,598 INFO [train.py:904] (0/8) Epoch 3, batch 2150, loss[loss=0.2532, simple_loss=0.3394, pruned_loss=0.08352, over 16777.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3179, pruned_loss=0.08885, over 3322458.82 frames. ], batch size: 57, lr: 2.17e-02, grad_scale: 8.0 2023-04-27 21:49:22,490 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22467.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:49:29,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1845, 5.0065, 4.9492, 5.0356, 4.4753, 4.9851, 4.9416, 4.5871], device='cuda:0'), covar=tensor([0.0274, 0.0149, 0.0148, 0.0111, 0.0660, 0.0180, 0.0198, 0.0275], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0127, 0.0186, 0.0154, 0.0219, 0.0165, 0.0136, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:49:38,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3366, 4.3565, 4.2410, 1.8027, 2.9914, 2.5670, 3.7917, 4.3337], device='cuda:0'), covar=tensor([0.0315, 0.0581, 0.0316, 0.1617, 0.0755, 0.0911, 0.0748, 0.0611], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0122, 0.0156, 0.0145, 0.0137, 0.0130, 0.0148, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 21:49:47,671 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22486.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:08,039 INFO [train.py:904] (0/8) Epoch 3, batch 2200, loss[loss=0.3271, simple_loss=0.3684, pruned_loss=0.1429, over 12001.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3198, pruned_loss=0.09023, over 3320153.81 frames. ], batch size: 248, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:50:14,262 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22506.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:50:46,036 INFO [optim.py:368] (0/8) 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 21:50:48,911 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 3, batch 2250, loss[loss=0.227, simple_loss=0.302, pruned_loss=0.07602, over 17182.00 frames. ], tot_loss[loss=0.2506, simple_loss=0.3201, pruned_loss=0.0905, over 3314976.81 frames. ], batch size: 46, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:51:18,566 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22554.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:51:42,438 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22571.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:09,946 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22593.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:52:20,021 INFO [train.py:904] (0/8) Epoch 3, batch 2300, loss[loss=0.2565, simple_loss=0.3128, pruned_loss=0.1001, over 16786.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3193, pruned_loss=0.08987, over 3312239.57 frames. ], batch size: 83, lr: 2.17e-02, grad_scale: 4.0 2023-04-27 21:52:57,418 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22626.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:01,631 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.100e+02 3.931e+02 4.776e+02 5.724e+02 1.077e+03, threshold=9.553e+02, percent-clipped=1.0 2023-04-27 21:53:12,003 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22639.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:53:30,357 INFO [train.py:904] (0/8) Epoch 3, batch 2350, loss[loss=0.2344, simple_loss=0.3087, pruned_loss=0.08004, over 16138.00 frames. ], tot_loss[loss=0.2491, simple_loss=0.3191, pruned_loss=0.08958, over 3318125.81 frames. ], batch size: 35, lr: 2.16e-02, grad_scale: 4.0 2023-04-27 21:54:01,191 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22674.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:54:03,553 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.8458, 6.0584, 5.7178, 5.8820, 5.3912, 4.9601, 5.6246, 6.1842], device='cuda:0'), covar=tensor([0.0428, 0.0481, 0.0893, 0.0376, 0.0496, 0.0389, 0.0406, 0.0559], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0384, 0.0337, 0.0237, 0.0251, 0.0232, 0.0302, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:54:31,354 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9196, 3.9265, 3.7212, 3.8340, 3.3915, 3.8769, 3.4885, 3.6328], device='cuda:0'), covar=tensor([0.0357, 0.0169, 0.0189, 0.0136, 0.0763, 0.0178, 0.0762, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0128, 0.0189, 0.0156, 0.0220, 0.0167, 0.0137, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:54:35,060 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22700.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 21:54:35,723 INFO [train.py:904] (0/8) Epoch 3, batch 2400, loss[loss=0.2546, simple_loss=0.3335, pruned_loss=0.08791, over 17180.00 frames. ], tot_loss[loss=0.2498, simple_loss=0.3201, pruned_loss=0.08973, over 3323395.70 frames. ], batch size: 46, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:54:39,862 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7746, 2.6263, 2.6458, 1.9557, 2.6820, 2.6602, 2.5263, 1.8502], device='cuda:0'), covar=tensor([0.0290, 0.0056, 0.0039, 0.0216, 0.0039, 0.0049, 0.0036, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0057, 0.0060, 0.0109, 0.0053, 0.0062, 0.0061, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:55:14,043 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7873, 4.8277, 5.4156, 5.3419, 5.3794, 4.8588, 4.9014, 4.7943], device='cuda:0'), covar=tensor([0.0213, 0.0212, 0.0245, 0.0295, 0.0280, 0.0230, 0.0632, 0.0260], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0180, 0.0197, 0.0194, 0.0235, 0.0196, 0.0302, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 21:55:17,392 INFO [optim.py:368] (0/8) 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 21:55:20,231 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-27 21:55:32,306 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:55:46,921 INFO [train.py:904] (0/8) Epoch 3, batch 2450, loss[loss=0.2552, simple_loss=0.3387, pruned_loss=0.0858, over 16684.00 frames. ], tot_loss[loss=0.2501, simple_loss=0.3208, pruned_loss=0.08974, over 3305860.87 frames. ], batch size: 62, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:56:23,460 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-27 21:56:32,180 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1295, 3.9142, 4.1052, 2.9308, 3.9288, 4.1907, 4.0200, 2.2546], device='cuda:0'), covar=tensor([0.0395, 0.0076, 0.0033, 0.0240, 0.0032, 0.0038, 0.0023, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0056, 0.0060, 0.0108, 0.0053, 0.0061, 0.0061, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 21:56:33,343 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22786.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:46,197 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22794.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:56:49,271 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4791, 5.7577, 5.5185, 5.6940, 5.1920, 4.8402, 5.3043, 5.9715], device='cuda:0'), covar=tensor([0.0420, 0.0620, 0.0737, 0.0296, 0.0444, 0.0450, 0.0465, 0.0515], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0385, 0.0332, 0.0232, 0.0245, 0.0233, 0.0298, 0.0258], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:56:55,063 INFO [train.py:904] (0/8) Epoch 3, batch 2500, loss[loss=0.2062, simple_loss=0.2853, pruned_loss=0.06357, over 17234.00 frames. ], tot_loss[loss=0.2489, simple_loss=0.3197, pruned_loss=0.08905, over 3312331.09 frames. ], batch size: 45, lr: 2.16e-02, grad_scale: 8.0 2023-04-27 21:56:59,285 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-27 21:57:05,333 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0324, 5.3645, 5.1072, 5.2042, 4.7840, 4.5233, 4.8612, 5.4314], device='cuda:0'), covar=tensor([0.0511, 0.0589, 0.0776, 0.0367, 0.0481, 0.0603, 0.0488, 0.0595], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0391, 0.0337, 0.0237, 0.0249, 0.0238, 0.0303, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 21:57:27,654 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=22825.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:33,677 INFO [optim.py:368] (0/8) 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 21:57:36,900 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22832.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:57:41,075 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22834.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:00,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5292, 2.1560, 1.7013, 2.0180, 2.6637, 2.5883, 3.1023, 2.8027], device='cuda:0'), covar=tensor([0.0043, 0.0141, 0.0173, 0.0151, 0.0080, 0.0133, 0.0057, 0.0089], device='cuda:0'), in_proj_covar=tensor([0.0059, 0.0118, 0.0114, 0.0116, 0.0109, 0.0116, 0.0077, 0.0096], device='cuda:0'), out_proj_covar=tensor([8.5433e-05, 1.7440e-04, 1.6301e-04, 1.7102e-04, 1.6530e-04, 1.7431e-04, 1.1400e-04, 1.4769e-04], device='cuda:0') 2023-04-27 21:58:03,439 INFO [train.py:904] (0/8) Epoch 3, batch 2550, loss[loss=0.2717, simple_loss=0.3345, pruned_loss=0.1044, over 16315.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3192, pruned_loss=0.08866, over 3312992.71 frames. ], batch size: 165, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:58:10,315 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22855.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:31,364 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22871.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:44,552 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22880.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:58:52,588 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=22886.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:01,518 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=22893.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:13,845 INFO [train.py:904] (0/8) Epoch 3, batch 2600, loss[loss=0.2666, simple_loss=0.326, pruned_loss=0.1036, over 16702.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3186, pruned_loss=0.08757, over 3313928.32 frames. ], batch size: 124, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 21:59:38,827 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22919.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 21:59:53,535 INFO [optim.py:368] (0/8) 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 22:00:02,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4653, 3.0545, 2.5966, 2.2906, 2.2154, 1.9724, 2.9943, 3.1983], device='cuda:0'), covar=tensor([0.1250, 0.0477, 0.0764, 0.0797, 0.1483, 0.1161, 0.0284, 0.0314], device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0242, 0.0251, 0.0205, 0.0286, 0.0189, 0.0215, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:00:09,153 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=22941.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:00:22,559 INFO [train.py:904] (0/8) Epoch 3, batch 2650, loss[loss=0.2568, simple_loss=0.3393, pruned_loss=0.08716, over 17134.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3196, pruned_loss=0.08769, over 3307722.23 frames. ], batch size: 49, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:00:43,363 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 22:00:55,066 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-27 22:01:20,709 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-27 22:01:22,675 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=22995.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:01:30,285 INFO [train.py:904] (0/8) Epoch 3, batch 2700, loss[loss=0.2772, simple_loss=0.3577, pruned_loss=0.09836, over 16625.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3193, pruned_loss=0.08701, over 3317533.68 frames. ], batch size: 68, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:02:09,773 INFO [optim.py:368] (0/8) 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 22:02:23,978 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23041.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:02:37,964 INFO [train.py:904] (0/8) Epoch 3, batch 2750, loss[loss=0.2594, simple_loss=0.3278, pruned_loss=0.09553, over 16182.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3191, pruned_loss=0.08621, over 3322596.56 frames. ], batch size: 35, lr: 2.15e-02, grad_scale: 8.0 2023-04-27 22:02:40,459 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-27 22:03:28,908 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23089.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:03:45,511 INFO [train.py:904] (0/8) Epoch 3, batch 2800, loss[loss=0.2388, simple_loss=0.297, pruned_loss=0.09032, over 16536.00 frames. ], tot_loss[loss=0.2447, simple_loss=0.3178, pruned_loss=0.08578, over 3328514.33 frames. ], batch size: 146, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:03:47,382 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-27 22:03:52,774 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2503, 2.2548, 2.2650, 3.6367, 1.8749, 3.2640, 2.1887, 2.1001], device='cuda:0'), covar=tensor([0.0356, 0.0670, 0.0416, 0.0195, 0.1533, 0.0256, 0.0876, 0.1251], device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0221, 0.0187, 0.0252, 0.0291, 0.0197, 0.0213, 0.0286], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:03:54,901 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23108.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:06,977 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8933, 3.9043, 3.1024, 2.7277, 2.9878, 2.2870, 3.7896, 4.4017], device='cuda:0'), covar=tensor([0.1458, 0.0446, 0.0850, 0.0810, 0.1611, 0.1070, 0.0309, 0.0338], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0239, 0.0246, 0.0201, 0.0278, 0.0186, 0.0210, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:04:25,544 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.403e+02 4.548e+02 5.612e+02 1.011e+03, threshold=9.095e+02, percent-clipped=1.0 2023-04-27 22:04:54,810 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23150.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:04:55,606 INFO [train.py:904] (0/8) Epoch 3, batch 2850, loss[loss=0.2385, simple_loss=0.3179, pruned_loss=0.07959, over 16609.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3165, pruned_loss=0.08494, over 3323013.06 frames. ], batch size: 62, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:05:20,424 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23169.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:36,696 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23181.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:05:56,761 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7490, 3.7623, 2.9442, 2.5856, 2.6719, 2.0602, 3.5878, 3.8992], device='cuda:0'), covar=tensor([0.1579, 0.0458, 0.0905, 0.0873, 0.1560, 0.1385, 0.0336, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0240, 0.0246, 0.0201, 0.0278, 0.0187, 0.0211, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:06:03,320 INFO [train.py:904] (0/8) Epoch 3, batch 2900, loss[loss=0.2596, simple_loss=0.3214, pruned_loss=0.09895, over 16538.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.316, pruned_loss=0.08596, over 3329724.12 frames. ], batch size: 68, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:06:24,552 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 22:06:25,793 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23217.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:42,321 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:06:43,092 INFO [optim.py:368] (0/8) 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 22:07:01,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3569, 4.2495, 4.7625, 4.7931, 4.7806, 4.3786, 4.4477, 4.3856], device='cuda:0'), covar=tensor([0.0226, 0.0302, 0.0280, 0.0309, 0.0357, 0.0244, 0.0608, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0183, 0.0203, 0.0204, 0.0245, 0.0200, 0.0310, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 22:07:12,209 INFO [train.py:904] (0/8) Epoch 3, batch 2950, loss[loss=0.2806, simple_loss=0.3325, pruned_loss=0.1143, over 16925.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3163, pruned_loss=0.08768, over 3324339.15 frames. ], batch size: 96, lr: 2.14e-02, grad_scale: 8.0 2023-04-27 22:07:49,193 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23278.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:07,106 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23290.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:12,641 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23295.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:08:19,861 INFO [train.py:904] (0/8) Epoch 3, batch 3000, loss[loss=0.2493, simple_loss=0.3278, pruned_loss=0.08541, over 17098.00 frames. ], tot_loss[loss=0.2485, simple_loss=0.3179, pruned_loss=0.08949, over 3319184.83 frames. ], batch size: 53, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:08:19,861 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 22:08:30,495 INFO [train.py:938] (0/8) Epoch 3, validation: loss=0.1721, simple_loss=0.279, pruned_loss=0.03262, over 944034.00 frames. 2023-04-27 22:08:30,496 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 22:09:10,137 INFO [optim.py:368] (0/8) 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 22:09:25,507 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23343.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:09:35,714 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2565, 1.7328, 2.1717, 2.8380, 2.8956, 3.3750, 1.7622, 2.9983], device='cuda:0'), covar=tensor([0.0041, 0.0175, 0.0139, 0.0090, 0.0051, 0.0058, 0.0168, 0.0061], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0123, 0.0111, 0.0105, 0.0084, 0.0070, 0.0112, 0.0065], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-27 22:09:37,537 INFO [train.py:904] (0/8) Epoch 3, batch 3050, loss[loss=0.2806, simple_loss=0.3195, pruned_loss=0.1208, over 16946.00 frames. ], tot_loss[loss=0.2484, simple_loss=0.3178, pruned_loss=0.08949, over 3323173.50 frames. ], batch size: 109, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:10:04,910 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7371, 4.4519, 1.6288, 4.7343, 2.6816, 4.6467, 2.4022, 3.2842], device='cuda:0'), covar=tensor([0.0031, 0.0159, 0.1621, 0.0024, 0.0743, 0.0291, 0.1128, 0.0486], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0137, 0.0175, 0.0087, 0.0163, 0.0174, 0.0184, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-27 22:10:44,562 INFO [train.py:904] (0/8) Epoch 3, batch 3100, loss[loss=0.2654, simple_loss=0.3172, pruned_loss=0.1067, over 16722.00 frames. ], tot_loss[loss=0.2476, simple_loss=0.3169, pruned_loss=0.08915, over 3324639.48 frames. ], batch size: 102, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:11:28,259 INFO [optim.py:368] (0/8) 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 22:11:53,307 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23450.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:11:54,098 INFO [train.py:904] (0/8) Epoch 3, batch 3150, loss[loss=0.2473, simple_loss=0.3306, pruned_loss=0.08206, over 16532.00 frames. ], tot_loss[loss=0.2478, simple_loss=0.3165, pruned_loss=0.08953, over 3318278.95 frames. ], batch size: 62, lr: 2.13e-02, grad_scale: 4.0 2023-04-27 22:12:12,240 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23464.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:34,993 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23481.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:41,255 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.68 vs. limit=2.0 2023-04-27 22:12:44,083 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=23488.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:12:58,227 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23498.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:01,352 INFO [train.py:904] (0/8) Epoch 3, batch 3200, loss[loss=0.2144, simple_loss=0.2829, pruned_loss=0.07291, over 16850.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3152, pruned_loss=0.08854, over 3324466.49 frames. ], batch size: 102, lr: 2.13e-02, grad_scale: 8.0 2023-04-27 22:13:14,115 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-27 22:13:39,248 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23529.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:13:42,225 INFO [optim.py:368] (0/8) 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 22:13:51,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0871, 4.1111, 3.9888, 4.0773, 3.1721, 4.0749, 3.9194, 3.7260], device='cuda:0'), covar=tensor([0.0560, 0.0310, 0.0316, 0.0207, 0.1520, 0.0348, 0.0570, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0138, 0.0196, 0.0163, 0.0231, 0.0175, 0.0143, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 22:14:06,532 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=23549.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:14:08,951 INFO [train.py:904] (0/8) Epoch 3, batch 3250, loss[loss=0.1797, simple_loss=0.2585, pruned_loss=0.05044, over 17263.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3148, pruned_loss=0.08793, over 3324804.18 frames. ], batch size: 43, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:14:38,904 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23573.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:14:56,594 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23585.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:15:12,977 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-27 22:15:18,380 INFO [train.py:904] (0/8) Epoch 3, batch 3300, loss[loss=0.3278, simple_loss=0.3745, pruned_loss=0.1405, over 16263.00 frames. ], tot_loss[loss=0.2473, simple_loss=0.3169, pruned_loss=0.08883, over 3320700.22 frames. ], batch size: 165, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:15:57,200 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.549e+02 3.983e+02 4.910e+02 5.801e+02 1.123e+03, threshold=9.819e+02, percent-clipped=5.0 2023-04-27 22:16:24,702 INFO [train.py:904] (0/8) Epoch 3, batch 3350, loss[loss=0.213, simple_loss=0.2934, pruned_loss=0.06626, over 17207.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.3168, pruned_loss=0.08871, over 3316030.70 frames. ], batch size: 44, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:17:16,317 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-04-27 22:17:33,473 INFO [train.py:904] (0/8) Epoch 3, batch 3400, loss[loss=0.267, simple_loss=0.3227, pruned_loss=0.1056, over 16899.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3168, pruned_loss=0.08783, over 3313815.85 frames. ], batch size: 109, lr: 2.12e-02, grad_scale: 8.0 2023-04-27 22:17:35,121 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5343, 2.5447, 2.3315, 2.1044, 3.0000, 2.5786, 3.8592, 3.2802], device='cuda:0'), covar=tensor([0.0019, 0.0117, 0.0140, 0.0176, 0.0075, 0.0137, 0.0040, 0.0070], device='cuda:0'), in_proj_covar=tensor([0.0064, 0.0120, 0.0116, 0.0118, 0.0112, 0.0121, 0.0081, 0.0098], device='cuda:0'), out_proj_covar=tensor([9.3604e-05, 1.7448e-04, 1.6431e-04, 1.7157e-04, 1.6724e-04, 1.7838e-04, 1.1964e-04, 1.4999e-04], device='cuda:0') 2023-04-27 22:17:56,612 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3894, 4.3980, 3.8934, 1.8644, 2.9329, 2.6986, 3.6638, 4.1863], device='cuda:0'), covar=tensor([0.0297, 0.0458, 0.0369, 0.1499, 0.0697, 0.0810, 0.0679, 0.0683], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0124, 0.0153, 0.0143, 0.0135, 0.0127, 0.0147, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 22:18:13,371 INFO [optim.py:368] (0/8) 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 22:18:40,218 INFO [train.py:904] (0/8) Epoch 3, batch 3450, loss[loss=0.2364, simple_loss=0.3094, pruned_loss=0.08168, over 17170.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3142, pruned_loss=0.08658, over 3305944.39 frames. ], batch size: 46, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:18:58,670 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23764.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:19:00,879 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-27 22:19:47,208 INFO [train.py:904] (0/8) Epoch 3, batch 3500, loss[loss=0.2465, simple_loss=0.3085, pruned_loss=0.09229, over 16512.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.312, pruned_loss=0.08511, over 3300447.33 frames. ], batch size: 146, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:19:53,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3540, 2.3051, 2.2883, 3.6096, 1.7864, 3.3795, 2.0283, 2.0591], device='cuda:0'), covar=tensor([0.0383, 0.0878, 0.0508, 0.0277, 0.2054, 0.0301, 0.1118, 0.1636], device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0224, 0.0189, 0.0254, 0.0297, 0.0200, 0.0217, 0.0290], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:20:04,595 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23812.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:20:09,417 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4521, 3.5662, 3.5877, 1.7401, 3.8055, 3.7532, 3.1729, 2.8451], device='cuda:0'), covar=tensor([0.0627, 0.0100, 0.0145, 0.1026, 0.0054, 0.0050, 0.0244, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0079, 0.0078, 0.0137, 0.0074, 0.0073, 0.0108, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-27 22:20:31,312 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.876e+02 4.693e+02 5.719e+02 1.171e+03, threshold=9.385e+02, percent-clipped=5.0 2023-04-27 22:20:48,134 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=23844.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 22:20:59,081 INFO [train.py:904] (0/8) Epoch 3, batch 3550, loss[loss=0.2603, simple_loss=0.3213, pruned_loss=0.09962, over 16171.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3104, pruned_loss=0.08428, over 3300066.62 frames. ], batch size: 165, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:21:03,557 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-27 22:21:12,175 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-04-27 22:21:29,380 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23873.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:21:45,172 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=23885.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:05,758 INFO [train.py:904] (0/8) Epoch 3, batch 3600, loss[loss=0.2223, simple_loss=0.308, pruned_loss=0.06827, over 17036.00 frames. ], tot_loss[loss=0.2376, simple_loss=0.3088, pruned_loss=0.08321, over 3308811.11 frames. ], batch size: 50, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:22:29,300 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8067, 3.5148, 2.6821, 5.0728, 4.9180, 4.3092, 1.9202, 3.3173], device='cuda:0'), covar=tensor([0.1437, 0.0492, 0.1171, 0.0067, 0.0225, 0.0348, 0.1270, 0.0607], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0130, 0.0162, 0.0077, 0.0151, 0.0142, 0.0154, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 22:22:33,416 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23921.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:22:47,019 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.441e+02 3.888e+02 4.838e+02 6.309e+02 1.139e+03, threshold=9.677e+02, percent-clipped=4.0 2023-04-27 22:22:50,343 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=23933.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:23:14,452 INFO [train.py:904] (0/8) Epoch 3, batch 3650, loss[loss=0.2296, simple_loss=0.2856, pruned_loss=0.08684, over 16470.00 frames. ], tot_loss[loss=0.2382, simple_loss=0.3081, pruned_loss=0.08418, over 3299744.22 frames. ], batch size: 68, lr: 2.11e-02, grad_scale: 8.0 2023-04-27 22:23:35,178 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0236, 3.7093, 2.5191, 4.7031, 4.5471, 4.3616, 1.7389, 3.1724], device='cuda:0'), covar=tensor([0.1238, 0.0320, 0.1108, 0.0061, 0.0212, 0.0282, 0.1215, 0.0575], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0131, 0.0164, 0.0076, 0.0150, 0.0142, 0.0155, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 22:24:25,491 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-24000.pt 2023-04-27 22:24:29,898 INFO [train.py:904] (0/8) Epoch 3, batch 3700, loss[loss=0.2596, simple_loss=0.3137, pruned_loss=0.1028, over 16489.00 frames. ], tot_loss[loss=0.2394, simple_loss=0.3068, pruned_loss=0.08603, over 3303735.92 frames. ], batch size: 75, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:25:13,798 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.162e+02 3.909e+02 4.503e+02 5.530e+02 1.035e+03, threshold=9.006e+02, percent-clipped=1.0 2023-04-27 22:25:22,527 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24037.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:25:42,969 INFO [train.py:904] (0/8) Epoch 3, batch 3750, loss[loss=0.2777, simple_loss=0.34, pruned_loss=0.1077, over 12008.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3083, pruned_loss=0.08839, over 3278787.25 frames. ], batch size: 247, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:26:46,927 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24098.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:26:50,667 INFO [train.py:904] (0/8) Epoch 3, batch 3800, loss[loss=0.2375, simple_loss=0.2952, pruned_loss=0.08987, over 16732.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.309, pruned_loss=0.09004, over 3270530.51 frames. ], batch size: 124, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:27:34,020 INFO [optim.py:368] (0/8) 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 22:27:44,065 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1724, 1.9696, 1.5929, 1.7686, 2.6046, 2.4594, 2.7074, 2.5951], device='cuda:0'), covar=tensor([0.0033, 0.0156, 0.0182, 0.0180, 0.0079, 0.0114, 0.0064, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0061, 0.0119, 0.0119, 0.0116, 0.0112, 0.0119, 0.0079, 0.0097], device='cuda:0'), out_proj_covar=tensor([9.0320e-05, 1.7364e-04, 1.6814e-04, 1.6854e-04, 1.6595e-04, 1.7614e-04, 1.1740e-04, 1.4746e-04], device='cuda:0') 2023-04-27 22:27:52,803 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24144.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:28:01,901 INFO [train.py:904] (0/8) Epoch 3, batch 3850, loss[loss=0.2571, simple_loss=0.3222, pruned_loss=0.09601, over 16408.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3085, pruned_loss=0.09006, over 3276051.82 frames. ], batch size: 68, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:28:14,889 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 22:29:00,728 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=24192.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:29:13,625 INFO [train.py:904] (0/8) Epoch 3, batch 3900, loss[loss=0.2386, simple_loss=0.3013, pruned_loss=0.0879, over 16904.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3073, pruned_loss=0.08956, over 3272010.78 frames. ], batch size: 116, lr: 2.10e-02, grad_scale: 8.0 2023-04-27 22:29:25,121 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-27 22:29:34,116 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-04-27 22:29:39,250 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0269, 3.3068, 3.8069, 2.4408, 3.4385, 3.7366, 3.5803, 1.9938], device='cuda:0'), covar=tensor([0.0335, 0.0085, 0.0023, 0.0218, 0.0041, 0.0039, 0.0029, 0.0292], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0053, 0.0057, 0.0107, 0.0054, 0.0059, 0.0058, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:29:56,954 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.402e+02 3.854e+02 4.579e+02 5.596e+02 1.788e+03, threshold=9.157e+02, percent-clipped=5.0 2023-04-27 22:30:00,296 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-27 22:30:22,694 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7002, 4.5743, 4.5500, 4.6010, 4.1347, 4.5705, 4.4467, 4.2588], device='cuda:0'), covar=tensor([0.0279, 0.0184, 0.0148, 0.0119, 0.0706, 0.0170, 0.0229, 0.0272], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0127, 0.0178, 0.0148, 0.0210, 0.0160, 0.0130, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:30:25,199 INFO [train.py:904] (0/8) Epoch 3, batch 3950, loss[loss=0.2659, simple_loss=0.3106, pruned_loss=0.1106, over 16892.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3063, pruned_loss=0.08979, over 3266623.56 frames. ], batch size: 109, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:30:25,645 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0524, 3.4423, 3.6668, 2.4675, 3.3781, 3.6535, 3.4507, 2.1557], device='cuda:0'), covar=tensor([0.0342, 0.0067, 0.0026, 0.0223, 0.0040, 0.0042, 0.0035, 0.0260], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0054, 0.0058, 0.0110, 0.0054, 0.0061, 0.0059, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:30:44,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6692, 4.6227, 5.2515, 5.2917, 5.2103, 4.6979, 4.7138, 4.5724], device='cuda:0'), covar=tensor([0.0239, 0.0362, 0.0266, 0.0318, 0.0261, 0.0206, 0.0645, 0.0264], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0188, 0.0201, 0.0200, 0.0240, 0.0204, 0.0306, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-27 22:30:44,687 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5299, 4.0668, 4.3771, 1.7512, 4.6195, 4.6120, 3.5628, 3.1992], device='cuda:0'), covar=tensor([0.0863, 0.0145, 0.0100, 0.1391, 0.0033, 0.0040, 0.0254, 0.0437], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0082, 0.0078, 0.0145, 0.0076, 0.0074, 0.0112, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-27 22:31:07,689 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9806, 3.3000, 2.9029, 4.4161, 4.2432, 4.1225, 1.4438, 3.6050], device='cuda:0'), covar=tensor([0.1309, 0.0390, 0.0838, 0.0057, 0.0152, 0.0248, 0.1282, 0.0383], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0130, 0.0162, 0.0074, 0.0142, 0.0137, 0.0153, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 22:31:34,854 INFO [train.py:904] (0/8) Epoch 3, batch 4000, loss[loss=0.2267, simple_loss=0.2989, pruned_loss=0.07724, over 16677.00 frames. ], tot_loss[loss=0.2429, simple_loss=0.3059, pruned_loss=0.08991, over 3273320.02 frames. ], batch size: 62, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:32:17,083 INFO [optim.py:368] (0/8) 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 22:32:45,207 INFO [train.py:904] (0/8) Epoch 3, batch 4050, loss[loss=0.209, simple_loss=0.2905, pruned_loss=0.06379, over 16796.00 frames. ], tot_loss[loss=0.2398, simple_loss=0.3046, pruned_loss=0.08751, over 3271563.58 frames. ], batch size: 116, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:32:50,582 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6691, 4.5837, 4.3541, 3.7960, 4.5148, 1.6980, 4.1997, 4.3486], device='cuda:0'), covar=tensor([0.0044, 0.0041, 0.0084, 0.0297, 0.0047, 0.1439, 0.0080, 0.0106], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0067, 0.0098, 0.0116, 0.0075, 0.0116, 0.0089, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:32:58,563 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9829, 1.5276, 2.0944, 2.8647, 2.7742, 3.0453, 1.6650, 3.0002], device='cuda:0'), covar=tensor([0.0036, 0.0200, 0.0123, 0.0065, 0.0046, 0.0039, 0.0175, 0.0021], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0122, 0.0111, 0.0102, 0.0088, 0.0069, 0.0114, 0.0065], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-27 22:33:36,538 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4272, 3.4908, 3.2096, 3.3176, 3.0015, 3.2701, 3.1253, 3.1691], device='cuda:0'), covar=tensor([0.0325, 0.0167, 0.0218, 0.0156, 0.0605, 0.0178, 0.1035, 0.0289], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0123, 0.0176, 0.0145, 0.0204, 0.0159, 0.0128, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:33:46,389 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24393.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:33:58,265 INFO [train.py:904] (0/8) Epoch 3, batch 4100, loss[loss=0.2369, simple_loss=0.3106, pruned_loss=0.08159, over 17122.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3048, pruned_loss=0.08529, over 3269627.11 frames. ], batch size: 47, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:34:42,998 INFO [optim.py:368] (0/8) 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 22:35:00,703 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-27 22:35:13,095 INFO [train.py:904] (0/8) Epoch 3, batch 4150, loss[loss=0.3123, simple_loss=0.3669, pruned_loss=0.1288, over 10967.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3142, pruned_loss=0.08982, over 3244001.20 frames. ], batch size: 246, lr: 2.09e-02, grad_scale: 8.0 2023-04-27 22:36:27,621 INFO [train.py:904] (0/8) Epoch 3, batch 4200, loss[loss=0.2929, simple_loss=0.3699, pruned_loss=0.1079, over 16910.00 frames. ], tot_loss[loss=0.2532, simple_loss=0.3223, pruned_loss=0.09212, over 3232719.39 frames. ], batch size: 116, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:37:10,964 INFO [optim.py:368] (0/8) 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 22:37:35,095 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24547.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:37:40,131 INFO [train.py:904] (0/8) Epoch 3, batch 4250, loss[loss=0.2762, simple_loss=0.3345, pruned_loss=0.109, over 12136.00 frames. ], tot_loss[loss=0.2546, simple_loss=0.325, pruned_loss=0.09207, over 3211848.28 frames. ], batch size: 246, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:38:06,854 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24569.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:38:53,671 INFO [train.py:904] (0/8) Epoch 3, batch 4300, loss[loss=0.2786, simple_loss=0.3403, pruned_loss=0.1085, over 11354.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3257, pruned_loss=0.09066, over 3196805.07 frames. ], batch size: 246, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:39:04,966 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24608.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:26,755 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1397, 3.8975, 4.0761, 4.3733, 4.3559, 3.9810, 4.3195, 4.3444], device='cuda:0'), covar=tensor([0.0494, 0.0529, 0.0973, 0.0303, 0.0325, 0.0616, 0.0377, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0331, 0.0430, 0.0332, 0.0247, 0.0234, 0.0257, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:39:37,567 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24630.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:39:38,221 INFO [optim.py:368] (0/8) 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 22:39:42,129 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-27 22:40:06,101 INFO [train.py:904] (0/8) Epoch 3, batch 4350, loss[loss=0.2521, simple_loss=0.3343, pruned_loss=0.08492, over 16523.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3292, pruned_loss=0.09196, over 3198030.50 frames. ], batch size: 68, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:40:44,537 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-27 22:41:10,483 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=24693.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:41:21,571 INFO [train.py:904] (0/8) Epoch 3, batch 4400, loss[loss=0.2519, simple_loss=0.3421, pruned_loss=0.08088, over 16562.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3311, pruned_loss=0.09297, over 3189875.85 frames. ], batch size: 75, lr: 2.08e-02, grad_scale: 8.0 2023-04-27 22:41:35,299 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-27 22:42:05,390 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.377e+02 3.698e+02 4.447e+02 5.386e+02 9.920e+02, threshold=8.895e+02, percent-clipped=4.0 2023-04-27 22:42:20,530 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=24741.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:42:34,834 INFO [train.py:904] (0/8) Epoch 3, batch 4450, loss[loss=0.2409, simple_loss=0.3289, pruned_loss=0.07651, over 16651.00 frames. ], tot_loss[loss=0.2604, simple_loss=0.3341, pruned_loss=0.09333, over 3202402.83 frames. ], batch size: 134, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:42:55,224 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24765.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:43:03,650 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-27 22:43:47,154 INFO [train.py:904] (0/8) Epoch 3, batch 4500, loss[loss=0.2691, simple_loss=0.3438, pruned_loss=0.09724, over 16237.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.3329, pruned_loss=0.09216, over 3212916.55 frames. ], batch size: 165, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:44:10,161 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-27 22:44:11,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6760, 4.4690, 4.6696, 4.9757, 4.9878, 4.3834, 5.0486, 4.9769], device='cuda:0'), covar=tensor([0.0570, 0.0580, 0.1042, 0.0320, 0.0285, 0.0527, 0.0321, 0.0288], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0341, 0.0448, 0.0340, 0.0252, 0.0240, 0.0268, 0.0269], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:44:22,439 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24826.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:29,096 INFO [optim.py:368] (0/8) 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 22:44:52,900 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24848.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:44:56,540 INFO [train.py:904] (0/8) Epoch 3, batch 4550, loss[loss=0.2895, simple_loss=0.3572, pruned_loss=0.1109, over 16760.00 frames. ], tot_loss[loss=0.2576, simple_loss=0.3321, pruned_loss=0.09156, over 3231742.73 frames. ], batch size: 124, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:45:02,277 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2260, 4.9811, 5.0147, 5.0911, 4.3868, 5.0849, 4.9213, 4.6930], device='cuda:0'), covar=tensor([0.0276, 0.0149, 0.0151, 0.0113, 0.0787, 0.0158, 0.0135, 0.0207], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0116, 0.0167, 0.0138, 0.0198, 0.0147, 0.0123, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:46:07,325 INFO [train.py:904] (0/8) Epoch 3, batch 4600, loss[loss=0.248, simple_loss=0.3306, pruned_loss=0.08269, over 16794.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3323, pruned_loss=0.09126, over 3228875.16 frames. ], batch size: 83, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:46:10,669 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24903.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:20,508 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=24909.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:44,276 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=24925.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:46:52,472 INFO [optim.py:368] (0/8) 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 22:47:20,340 INFO [train.py:904] (0/8) Epoch 3, batch 4650, loss[loss=0.2216, simple_loss=0.2999, pruned_loss=0.07169, over 17190.00 frames. ], tot_loss[loss=0.2562, simple_loss=0.331, pruned_loss=0.09068, over 3217534.75 frames. ], batch size: 46, lr: 2.07e-02, grad_scale: 8.0 2023-04-27 22:47:28,680 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-27 22:47:50,699 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1907, 3.1873, 2.6380, 4.7860, 4.4987, 4.0796, 1.8502, 3.3606], device='cuda:0'), covar=tensor([0.1157, 0.0432, 0.1062, 0.0047, 0.0105, 0.0247, 0.1142, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0127, 0.0161, 0.0068, 0.0126, 0.0132, 0.0149, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 22:48:24,109 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=24995.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:48:32,751 INFO [train.py:904] (0/8) Epoch 3, batch 4700, loss[loss=0.2756, simple_loss=0.3432, pruned_loss=0.104, over 11413.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3274, pruned_loss=0.08887, over 3213886.44 frames. ], batch size: 247, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:48:54,288 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8447, 2.2930, 1.6415, 2.0121, 2.8967, 2.6733, 3.8493, 3.5436], device='cuda:0'), covar=tensor([0.0009, 0.0148, 0.0251, 0.0178, 0.0084, 0.0127, 0.0015, 0.0047], device='cuda:0'), in_proj_covar=tensor([0.0052, 0.0113, 0.0120, 0.0116, 0.0111, 0.0121, 0.0073, 0.0094], device='cuda:0'), out_proj_covar=tensor([7.4454e-05, 1.6282e-04, 1.6972e-04, 1.6828e-04, 1.6311e-04, 1.7756e-04, 1.0484e-04, 1.4109e-04], device='cuda:0') 2023-04-27 22:49:07,130 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 22:49:17,310 INFO [optim.py:368] (0/8) 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 22:49:24,848 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7869, 3.5091, 3.3756, 2.2524, 3.2332, 3.3049, 3.2561, 1.7273], device='cuda:0'), covar=tensor([0.0374, 0.0022, 0.0030, 0.0239, 0.0033, 0.0061, 0.0039, 0.0332], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0049, 0.0057, 0.0108, 0.0052, 0.0061, 0.0057, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:49:45,111 INFO [train.py:904] (0/8) Epoch 3, batch 4750, loss[loss=0.2242, simple_loss=0.3004, pruned_loss=0.07399, over 16445.00 frames. ], tot_loss[loss=0.2495, simple_loss=0.3239, pruned_loss=0.08756, over 3209579.54 frames. ], batch size: 68, lr: 2.06e-02, grad_scale: 4.0 2023-04-27 22:49:47,852 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 22:49:52,723 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25056.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:50:58,612 INFO [train.py:904] (0/8) Epoch 3, batch 4800, loss[loss=0.2531, simple_loss=0.3319, pruned_loss=0.08715, over 16177.00 frames. ], tot_loss[loss=0.2456, simple_loss=0.3202, pruned_loss=0.08547, over 3201662.74 frames. ], batch size: 165, lr: 2.06e-02, grad_scale: 8.0 2023-04-27 22:51:04,633 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-27 22:51:28,213 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25121.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:51:47,436 INFO [optim.py:368] (0/8) 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 22:52:13,136 INFO [train.py:904] (0/8) Epoch 3, batch 4850, loss[loss=0.2742, simple_loss=0.3478, pruned_loss=0.1003, over 16316.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.322, pruned_loss=0.08539, over 3194850.29 frames. ], batch size: 146, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:52:59,313 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2962, 4.9080, 5.1023, 5.1122, 4.6317, 5.0493, 5.0896, 4.7373], device='cuda:0'), covar=tensor([0.0265, 0.0144, 0.0122, 0.0108, 0.0607, 0.0129, 0.0136, 0.0243], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0114, 0.0162, 0.0133, 0.0188, 0.0146, 0.0119, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:53:25,031 INFO [train.py:904] (0/8) Epoch 3, batch 4900, loss[loss=0.2602, simple_loss=0.3443, pruned_loss=0.08802, over 16394.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3227, pruned_loss=0.08519, over 3189649.24 frames. ], batch size: 146, lr: 2.06e-02, grad_scale: 2.0 2023-04-27 22:53:27,938 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25203.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:29,422 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25204.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:53:46,334 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7456, 3.5877, 3.7264, 3.6314, 3.6143, 4.1198, 3.8919, 3.6415], device='cuda:0'), covar=tensor([0.1337, 0.1483, 0.1095, 0.1720, 0.2529, 0.1059, 0.0940, 0.1960], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0297, 0.0277, 0.0266, 0.0345, 0.0298, 0.0244, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 22:53:58,889 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25225.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:54:10,580 INFO [optim.py:368] (0/8) 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 22:54:14,866 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3990, 2.4400, 2.1360, 3.9223, 1.8587, 3.4606, 2.1597, 2.1724], device='cuda:0'), covar=tensor([0.0351, 0.0767, 0.0517, 0.0167, 0.1802, 0.0247, 0.0990, 0.1298], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0232, 0.0194, 0.0251, 0.0304, 0.0202, 0.0219, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:54:35,311 INFO [train.py:904] (0/8) Epoch 3, batch 4950, loss[loss=0.2646, simple_loss=0.3369, pruned_loss=0.09613, over 16920.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3232, pruned_loss=0.08541, over 3178822.66 frames. ], batch size: 109, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:54:35,559 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25251.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:08,065 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25273.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:55:21,231 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-27 22:55:48,346 INFO [train.py:904] (0/8) Epoch 3, batch 5000, loss[loss=0.2496, simple_loss=0.3398, pruned_loss=0.07968, over 16204.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.324, pruned_loss=0.08479, over 3191296.75 frames. ], batch size: 165, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:35,250 INFO [optim.py:368] (0/8) 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 22:56:59,654 INFO [train.py:904] (0/8) Epoch 3, batch 5050, loss[loss=0.2636, simple_loss=0.3291, pruned_loss=0.09905, over 16348.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.324, pruned_loss=0.08417, over 3200273.51 frames. ], batch size: 35, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:56:59,974 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25351.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 22:57:36,602 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8129, 4.6492, 5.1392, 5.1425, 5.2536, 4.7341, 4.5910, 4.5609], device='cuda:0'), covar=tensor([0.0271, 0.0388, 0.0336, 0.0432, 0.0450, 0.0330, 0.0910, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0171, 0.0187, 0.0183, 0.0231, 0.0190, 0.0290, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-27 22:58:08,624 INFO [train.py:904] (0/8) Epoch 3, batch 5100, loss[loss=0.2008, simple_loss=0.2936, pruned_loss=0.05405, over 16790.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3218, pruned_loss=0.0828, over 3206879.48 frames. ], batch size: 102, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:58:32,508 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5518, 2.6764, 2.2042, 4.2158, 1.8093, 3.7867, 2.2726, 2.2822], device='cuda:0'), covar=tensor([0.0356, 0.0694, 0.0525, 0.0149, 0.1824, 0.0227, 0.0995, 0.1376], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0231, 0.0196, 0.0256, 0.0302, 0.0206, 0.0223, 0.0302], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 22:58:38,845 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25421.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:58:47,859 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25428.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 22:58:57,533 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 3.320e+02 3.999e+02 5.208e+02 7.552e+02, threshold=7.999e+02, percent-clipped=0.0 2023-04-27 22:59:23,207 INFO [train.py:904] (0/8) Epoch 3, batch 5150, loss[loss=0.2578, simple_loss=0.3502, pruned_loss=0.08276, over 16876.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3223, pruned_loss=0.08224, over 3192471.49 frames. ], batch size: 102, lr: 2.05e-02, grad_scale: 2.0 2023-04-27 22:59:50,323 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25469.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 22:59:50,405 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2828, 5.5474, 5.1587, 5.3275, 4.9106, 4.5963, 5.0350, 5.6647], device='cuda:0'), covar=tensor([0.0422, 0.0494, 0.0778, 0.0314, 0.0419, 0.0481, 0.0407, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0355, 0.0313, 0.0225, 0.0229, 0.0224, 0.0287, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:00:19,347 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25489.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:00:35,992 INFO [train.py:904] (0/8) Epoch 3, batch 5200, loss[loss=0.2835, simple_loss=0.3471, pruned_loss=0.11, over 12293.00 frames. ], tot_loss[loss=0.2431, simple_loss=0.3219, pruned_loss=0.08219, over 3195646.90 frames. ], batch size: 246, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:00:40,439 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:01:22,144 INFO [optim.py:368] (0/8) 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 23:01:45,799 INFO [train.py:904] (0/8) Epoch 3, batch 5250, loss[loss=0.2531, simple_loss=0.3242, pruned_loss=0.09101, over 16944.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3194, pruned_loss=0.08195, over 3202089.27 frames. ], batch size: 109, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:01:47,294 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25552.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:02:00,436 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 23:02:56,119 INFO [train.py:904] (0/8) Epoch 3, batch 5300, loss[loss=0.2048, simple_loss=0.2838, pruned_loss=0.06296, over 16884.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3156, pruned_loss=0.08047, over 3205764.75 frames. ], batch size: 96, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:03:43,234 INFO [optim.py:368] (0/8) 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 23:04:08,028 INFO [train.py:904] (0/8) Epoch 3, batch 5350, loss[loss=0.2497, simple_loss=0.3249, pruned_loss=0.08726, over 16664.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3149, pruned_loss=0.08045, over 3192858.64 frames. ], batch size: 134, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:04:08,351 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=25651.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:04:30,670 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4385, 4.4245, 4.2740, 3.0882, 4.0297, 4.3051, 4.2625, 2.2052], device='cuda:0'), covar=tensor([0.0339, 0.0018, 0.0038, 0.0210, 0.0034, 0.0047, 0.0035, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0051, 0.0059, 0.0107, 0.0051, 0.0062, 0.0057, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:04:44,393 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0521, 4.0197, 4.0859, 4.1437, 4.0164, 4.6011, 4.3452, 4.1255], device='cuda:0'), covar=tensor([0.1255, 0.1271, 0.1054, 0.1555, 0.2133, 0.0918, 0.0967, 0.2169], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0297, 0.0273, 0.0264, 0.0344, 0.0299, 0.0241, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:05:16,511 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=25699.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:05:19,290 INFO [train.py:904] (0/8) Epoch 3, batch 5400, loss[loss=0.2554, simple_loss=0.3346, pruned_loss=0.08814, over 16719.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3178, pruned_loss=0.08143, over 3190435.76 frames. ], batch size: 124, lr: 2.04e-02, grad_scale: 4.0 2023-04-27 23:06:07,787 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 3.733e+02 4.548e+02 5.512e+02 9.876e+02, threshold=9.097e+02, percent-clipped=3.0 2023-04-27 23:06:34,330 INFO [train.py:904] (0/8) Epoch 3, batch 5450, loss[loss=0.3241, simple_loss=0.3841, pruned_loss=0.132, over 16662.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3215, pruned_loss=0.08376, over 3193289.85 frames. ], batch size: 134, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:07:24,680 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=25784.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:07:49,285 INFO [train.py:904] (0/8) Epoch 3, batch 5500, loss[loss=0.2944, simple_loss=0.3653, pruned_loss=0.1118, over 16822.00 frames. ], tot_loss[loss=0.2573, simple_loss=0.3313, pruned_loss=0.09165, over 3167859.32 frames. ], batch size: 102, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:08:16,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0385, 2.5146, 2.3536, 3.2973, 3.0834, 3.2286, 1.8826, 2.7588], device='cuda:0'), covar=tensor([0.1085, 0.0378, 0.0940, 0.0078, 0.0222, 0.0270, 0.1011, 0.0568], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0130, 0.0166, 0.0071, 0.0135, 0.0142, 0.0156, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 23:08:39,233 INFO [optim.py:368] (0/8) 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 23:09:06,200 INFO [train.py:904] (0/8) Epoch 3, batch 5550, loss[loss=0.3675, simple_loss=0.4142, pruned_loss=0.1604, over 16891.00 frames. ], tot_loss[loss=0.269, simple_loss=0.3402, pruned_loss=0.09896, over 3145170.93 frames. ], batch size: 116, lr: 2.03e-02, grad_scale: 4.0 2023-04-27 23:09:17,558 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=25858.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:09:56,085 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 23:10:25,195 INFO [train.py:904] (0/8) Epoch 3, batch 5600, loss[loss=0.2496, simple_loss=0.3317, pruned_loss=0.0838, over 16701.00 frames. ], tot_loss[loss=0.2798, simple_loss=0.3475, pruned_loss=0.106, over 3096097.22 frames. ], batch size: 83, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:10:56,220 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=25919.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:11:21,470 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.427e+02 5.585e+02 6.873e+02 8.559e+02 2.132e+03, threshold=1.375e+03, percent-clipped=5.0 2023-04-27 23:11:48,521 INFO [train.py:904] (0/8) Epoch 3, batch 5650, loss[loss=0.3556, simple_loss=0.4011, pruned_loss=0.155, over 16255.00 frames. ], tot_loss[loss=0.291, simple_loss=0.3554, pruned_loss=0.1134, over 3056992.56 frames. ], batch size: 165, lr: 2.03e-02, grad_scale: 8.0 2023-04-27 23:13:05,604 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-26000.pt 2023-04-27 23:13:10,054 INFO [train.py:904] (0/8) Epoch 3, batch 5700, loss[loss=0.2743, simple_loss=0.3587, pruned_loss=0.09489, over 16659.00 frames. ], tot_loss[loss=0.2925, simple_loss=0.3564, pruned_loss=0.1143, over 3071135.60 frames. ], batch size: 89, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:14:00,537 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.120e+02 5.064e+02 6.111e+02 7.661e+02 1.195e+03, threshold=1.222e+03, percent-clipped=0.0 2023-04-27 23:14:27,448 INFO [train.py:904] (0/8) Epoch 3, batch 5750, loss[loss=0.3552, simple_loss=0.4096, pruned_loss=0.1504, over 15244.00 frames. ], tot_loss[loss=0.2983, simple_loss=0.3609, pruned_loss=0.1179, over 3023024.59 frames. ], batch size: 190, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:22,095 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26084.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:15:47,186 INFO [train.py:904] (0/8) Epoch 3, batch 5800, loss[loss=0.2737, simple_loss=0.3474, pruned_loss=0.1, over 16392.00 frames. ], tot_loss[loss=0.2966, simple_loss=0.3603, pruned_loss=0.1164, over 3022275.28 frames. ], batch size: 146, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:15:48,971 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26102.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:16:36,166 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26132.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:16:38,592 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.751e+02 5.049e+02 6.109e+02 8.114e+02 1.629e+03, threshold=1.222e+03, percent-clipped=2.0 2023-04-27 23:17:05,090 INFO [train.py:904] (0/8) Epoch 3, batch 5850, loss[loss=0.3098, simple_loss=0.3708, pruned_loss=0.1243, over 16160.00 frames. ], tot_loss[loss=0.2919, simple_loss=0.3571, pruned_loss=0.1133, over 3046117.79 frames. ], batch size: 165, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:17:24,099 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26163.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:31,359 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26168.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:17:56,679 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6064, 5.0477, 5.1477, 5.1444, 5.0312, 5.5777, 5.2215, 5.0318], device='cuda:0'), covar=tensor([0.0719, 0.1118, 0.0798, 0.1284, 0.1909, 0.0717, 0.0857, 0.1754], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0301, 0.0280, 0.0268, 0.0348, 0.0303, 0.0247, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:18:26,957 INFO [train.py:904] (0/8) Epoch 3, batch 5900, loss[loss=0.2754, simple_loss=0.3507, pruned_loss=0.1, over 16554.00 frames. ], tot_loss[loss=0.2909, simple_loss=0.3561, pruned_loss=0.1128, over 3036846.46 frames. ], batch size: 68, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:18:51,423 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26214.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:18:51,804 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-27 23:19:09,059 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6553, 4.6028, 4.4627, 3.8001, 4.5067, 1.7228, 4.2004, 4.4430], device='cuda:0'), covar=tensor([0.0050, 0.0050, 0.0083, 0.0275, 0.0047, 0.1361, 0.0073, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0061, 0.0095, 0.0110, 0.0071, 0.0120, 0.0082, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:19:15,302 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26229.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:22,631 INFO [optim.py:368] (0/8) 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 23:19:47,812 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26250.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:19:48,582 INFO [train.py:904] (0/8) Epoch 3, batch 5950, loss[loss=0.3023, simple_loss=0.3722, pruned_loss=0.1162, over 17064.00 frames. ], tot_loss[loss=0.2886, simple_loss=0.3565, pruned_loss=0.1103, over 3051666.48 frames. ], batch size: 55, lr: 2.02e-02, grad_scale: 8.0 2023-04-27 23:20:19,795 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0344, 4.7871, 4.6963, 5.2900, 5.3296, 4.6375, 5.3848, 5.2727], device='cuda:0'), covar=tensor([0.0667, 0.0661, 0.1594, 0.0464, 0.0530, 0.0581, 0.0470, 0.0463], device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0338, 0.0444, 0.0338, 0.0252, 0.0238, 0.0277, 0.0277], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:21:07,935 INFO [train.py:904] (0/8) Epoch 3, batch 6000, loss[loss=0.2447, simple_loss=0.3195, pruned_loss=0.085, over 16669.00 frames. ], tot_loss[loss=0.2869, simple_loss=0.3553, pruned_loss=0.1093, over 3089239.19 frames. ], batch size: 57, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:21:07,936 INFO [train.py:929] (0/8) Computing validation loss 2023-04-27 23:21:18,889 INFO [train.py:938] (0/8) Epoch 3, validation: loss=0.2097, simple_loss=0.3184, pruned_loss=0.05055, over 944034.00 frames. 2023-04-27 23:21:18,890 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-27 23:21:34,032 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26311.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:22:07,345 INFO [optim.py:368] (0/8) 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 23:22:12,992 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4149, 1.7642, 1.5331, 1.5177, 2.1256, 1.9805, 2.2217, 2.3301], device='cuda:0'), covar=tensor([0.0015, 0.0126, 0.0152, 0.0160, 0.0065, 0.0115, 0.0043, 0.0066], device='cuda:0'), in_proj_covar=tensor([0.0052, 0.0116, 0.0120, 0.0122, 0.0112, 0.0123, 0.0076, 0.0098], device='cuda:0'), out_proj_covar=tensor([7.0831e-05, 1.6617e-04, 1.6686e-04, 1.7410e-04, 1.6286e-04, 1.7805e-04, 1.0802e-04, 1.4540e-04], device='cuda:0') 2023-04-27 23:22:36,230 INFO [train.py:904] (0/8) Epoch 3, batch 6050, loss[loss=0.2249, simple_loss=0.3063, pruned_loss=0.07168, over 16179.00 frames. ], tot_loss[loss=0.2859, simple_loss=0.3536, pruned_loss=0.1091, over 3070646.58 frames. ], batch size: 35, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:22:48,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3768, 4.2733, 4.1351, 4.1870, 3.6966, 4.1982, 4.1497, 3.9198], device='cuda:0'), covar=tensor([0.0468, 0.0278, 0.0215, 0.0154, 0.0783, 0.0314, 0.0369, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0132, 0.0174, 0.0147, 0.0208, 0.0166, 0.0130, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:23:21,083 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5066, 2.5215, 2.1960, 3.9159, 1.7747, 3.5041, 2.2777, 2.3145], device='cuda:0'), covar=tensor([0.0386, 0.0860, 0.0613, 0.0238, 0.2156, 0.0287, 0.0997, 0.1341], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0235, 0.0197, 0.0256, 0.0313, 0.0209, 0.0224, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:23:51,481 INFO [train.py:904] (0/8) Epoch 3, batch 6100, loss[loss=0.3089, simple_loss=0.3564, pruned_loss=0.1307, over 11696.00 frames. ], tot_loss[loss=0.2832, simple_loss=0.3522, pruned_loss=0.1071, over 3087307.94 frames. ], batch size: 246, lr: 2.01e-02, grad_scale: 8.0 2023-04-27 23:23:57,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9794, 1.6477, 1.4436, 1.4688, 1.8254, 1.6908, 1.7623, 1.8814], device='cuda:0'), covar=tensor([0.0021, 0.0093, 0.0120, 0.0111, 0.0060, 0.0089, 0.0047, 0.0060], device='cuda:0'), in_proj_covar=tensor([0.0053, 0.0116, 0.0121, 0.0122, 0.0112, 0.0122, 0.0076, 0.0098], device='cuda:0'), out_proj_covar=tensor([7.1997e-05, 1.6584e-04, 1.6724e-04, 1.7438e-04, 1.6342e-04, 1.7557e-04, 1.0794e-04, 1.4568e-04], device='cuda:0') 2023-04-27 23:24:12,719 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26414.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:24:42,458 INFO [optim.py:368] (0/8) 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 23:25:04,789 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8064, 3.7890, 4.4099, 4.3424, 4.3463, 3.9080, 3.9901, 3.9160], device='cuda:0'), covar=tensor([0.0233, 0.0268, 0.0248, 0.0329, 0.0349, 0.0257, 0.0701, 0.0314], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0168, 0.0186, 0.0184, 0.0225, 0.0192, 0.0291, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-27 23:25:11,826 INFO [train.py:904] (0/8) Epoch 3, batch 6150, loss[loss=0.2543, simple_loss=0.3299, pruned_loss=0.08932, over 16817.00 frames. ], tot_loss[loss=0.2805, simple_loss=0.3494, pruned_loss=0.1058, over 3094862.04 frames. ], batch size: 102, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:25:23,316 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26458.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:25:50,229 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26475.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:26:28,212 INFO [train.py:904] (0/8) Epoch 3, batch 6200, loss[loss=0.3066, simple_loss=0.3773, pruned_loss=0.118, over 16241.00 frames. ], tot_loss[loss=0.2785, simple_loss=0.3471, pruned_loss=0.1049, over 3104349.50 frames. ], batch size: 165, lr: 2.01e-02, grad_scale: 4.0 2023-04-27 23:26:48,143 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26514.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:27:03,426 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26524.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:27:12,868 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2120, 4.1752, 1.7059, 4.2283, 2.6565, 4.2483, 1.9128, 2.8010], device='cuda:0'), covar=tensor([0.0049, 0.0144, 0.1549, 0.0032, 0.0667, 0.0294, 0.1471, 0.0607], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0131, 0.0171, 0.0079, 0.0160, 0.0163, 0.0179, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-27 23:27:18,526 INFO [optim.py:368] (0/8) 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 23:27:41,895 INFO [train.py:904] (0/8) Epoch 3, batch 6250, loss[loss=0.3358, simple_loss=0.3808, pruned_loss=0.1454, over 11783.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.3465, pruned_loss=0.104, over 3117244.77 frames. ], batch size: 248, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:27:58,188 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26562.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:28:14,684 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5025, 1.7833, 1.5070, 1.4568, 2.1992, 1.9028, 2.1416, 2.2340], device='cuda:0'), covar=tensor([0.0015, 0.0123, 0.0164, 0.0162, 0.0070, 0.0105, 0.0051, 0.0066], device='cuda:0'), in_proj_covar=tensor([0.0052, 0.0118, 0.0121, 0.0123, 0.0114, 0.0123, 0.0078, 0.0100], device='cuda:0'), out_proj_covar=tensor([7.1321e-05, 1.6814e-04, 1.6777e-04, 1.7570e-04, 1.6580e-04, 1.7632e-04, 1.1076e-04, 1.4723e-04], device='cuda:0') 2023-04-27 23:28:54,867 INFO [train.py:904] (0/8) Epoch 3, batch 6300, loss[loss=0.2742, simple_loss=0.3412, pruned_loss=0.1037, over 16405.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3463, pruned_loss=0.1033, over 3123871.37 frames. ], batch size: 75, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:29:02,863 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:29:48,253 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.782e+02 4.322e+02 5.254e+02 6.822e+02 1.499e+03, threshold=1.051e+03, percent-clipped=4.0 2023-04-27 23:30:11,910 INFO [train.py:904] (0/8) Epoch 3, batch 6350, loss[loss=0.2644, simple_loss=0.3407, pruned_loss=0.094, over 16927.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3474, pruned_loss=0.1045, over 3132304.90 frames. ], batch size: 109, lr: 2.00e-02, grad_scale: 4.0 2023-04-27 23:31:25,040 INFO [train.py:904] (0/8) Epoch 3, batch 6400, loss[loss=0.2218, simple_loss=0.3029, pruned_loss=0.0704, over 16865.00 frames. ], tot_loss[loss=0.2803, simple_loss=0.3482, pruned_loss=0.1062, over 3108941.64 frames. ], batch size: 96, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:12,499 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.974e+02 5.048e+02 6.146e+02 7.949e+02 2.287e+03, threshold=1.229e+03, percent-clipped=11.0 2023-04-27 23:32:35,294 INFO [train.py:904] (0/8) Epoch 3, batch 6450, loss[loss=0.2489, simple_loss=0.3197, pruned_loss=0.08906, over 15265.00 frames. ], tot_loss[loss=0.2784, simple_loss=0.3475, pruned_loss=0.1046, over 3113190.31 frames. ], batch size: 190, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:32:47,359 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26758.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:05,692 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=26770.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:33:26,733 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3948, 2.9803, 2.5243, 2.2929, 2.2474, 1.9754, 2.9101, 3.1313], device='cuda:0'), covar=tensor([0.1383, 0.0577, 0.0899, 0.0895, 0.1802, 0.1224, 0.0370, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0261, 0.0234, 0.0250, 0.0207, 0.0287, 0.0188, 0.0214, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:33:48,944 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1074, 2.5391, 2.4930, 3.2907, 3.0722, 3.2869, 1.8828, 2.8252], device='cuda:0'), covar=tensor([0.1051, 0.0344, 0.0869, 0.0087, 0.0218, 0.0272, 0.1054, 0.0507], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0133, 0.0166, 0.0072, 0.0137, 0.0144, 0.0154, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-27 23:33:53,761 INFO [train.py:904] (0/8) Epoch 3, batch 6500, loss[loss=0.257, simple_loss=0.3259, pruned_loss=0.09403, over 16896.00 frames. ], tot_loss[loss=0.2758, simple_loss=0.3448, pruned_loss=0.1034, over 3110429.66 frames. ], batch size: 116, lr: 2.00e-02, grad_scale: 8.0 2023-04-27 23:34:01,036 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26806.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:34:26,013 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3776, 3.2939, 2.5876, 2.3427, 2.5195, 2.2176, 3.4541, 3.6158], device='cuda:0'), covar=tensor([0.2175, 0.0825, 0.1204, 0.1102, 0.1624, 0.1221, 0.0421, 0.0510], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0237, 0.0252, 0.0211, 0.0294, 0.0190, 0.0216, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:34:29,622 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 23:35:12,121 INFO [train.py:904] (0/8) Epoch 3, batch 6550, loss[loss=0.2886, simple_loss=0.3724, pruned_loss=0.1024, over 17116.00 frames. ], tot_loss[loss=0.2771, simple_loss=0.3474, pruned_loss=0.1034, over 3135753.53 frames. ], batch size: 49, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:35:45,477 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26872.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:36:28,904 INFO [train.py:904] (0/8) Epoch 3, batch 6600, loss[loss=0.2482, simple_loss=0.3286, pruned_loss=0.08395, over 16482.00 frames. ], tot_loss[loss=0.2798, simple_loss=0.3501, pruned_loss=0.1047, over 3126892.90 frames. ], batch size: 68, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:36:33,615 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26904.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:36:37,669 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=26906.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:36:47,029 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1607, 4.1521, 4.7179, 4.6368, 4.6395, 4.2562, 4.2991, 4.1941], device='cuda:0'), covar=tensor([0.0202, 0.0225, 0.0222, 0.0323, 0.0334, 0.0236, 0.0811, 0.0350], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0169, 0.0188, 0.0186, 0.0224, 0.0192, 0.0290, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-27 23:36:52,036 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5192, 3.4117, 2.7860, 2.4292, 2.6328, 2.0805, 3.5604, 3.9030], device='cuda:0'), covar=tensor([0.1859, 0.0620, 0.1073, 0.0972, 0.1758, 0.1221, 0.0363, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0238, 0.0256, 0.0213, 0.0295, 0.0192, 0.0219, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:37:20,965 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.349e+02 4.809e+02 5.887e+02 7.395e+02 2.123e+03, threshold=1.177e+03, percent-clipped=7.0 2023-04-27 23:37:45,752 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=26950.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:46,408 INFO [train.py:904] (0/8) Epoch 3, batch 6650, loss[loss=0.26, simple_loss=0.3393, pruned_loss=0.09038, over 16822.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3513, pruned_loss=0.1061, over 3128420.55 frames. ], batch size: 102, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:37:51,231 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=26954.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:37:53,967 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3223, 4.2482, 1.8134, 4.4466, 2.7443, 4.3710, 2.0173, 3.0980], device='cuda:0'), covar=tensor([0.0041, 0.0172, 0.1593, 0.0031, 0.0667, 0.0254, 0.1365, 0.0554], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0130, 0.0173, 0.0080, 0.0160, 0.0160, 0.0179, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-27 23:38:00,528 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2862, 4.5748, 4.3048, 4.2970, 3.9791, 3.9679, 4.2325, 4.6035], device='cuda:0'), covar=tensor([0.0496, 0.0567, 0.0829, 0.0413, 0.0543, 0.0796, 0.0472, 0.0560], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0381, 0.0335, 0.0243, 0.0244, 0.0242, 0.0301, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:38:09,008 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=26965.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:38:13,986 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-04-27 23:39:02,827 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-04-27 23:39:03,595 INFO [train.py:904] (0/8) Epoch 3, batch 6700, loss[loss=0.2595, simple_loss=0.3382, pruned_loss=0.09041, over 16818.00 frames. ], tot_loss[loss=0.2822, simple_loss=0.3505, pruned_loss=0.1069, over 3105818.71 frames. ], batch size: 102, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:39:05,246 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3573, 4.2379, 4.1554, 4.1929, 3.7455, 4.2002, 4.1044, 3.9746], device='cuda:0'), covar=tensor([0.0289, 0.0160, 0.0183, 0.0128, 0.0726, 0.0201, 0.0309, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0128, 0.0167, 0.0140, 0.0197, 0.0160, 0.0126, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:39:18,528 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27011.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:39:57,110 INFO [optim.py:368] (0/8) 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 23:40:10,089 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-27 23:40:21,197 INFO [train.py:904] (0/8) Epoch 3, batch 6750, loss[loss=0.2639, simple_loss=0.3355, pruned_loss=0.09616, over 17138.00 frames. ], tot_loss[loss=0.2808, simple_loss=0.3486, pruned_loss=0.1065, over 3107700.48 frames. ], batch size: 48, lr: 1.99e-02, grad_scale: 8.0 2023-04-27 23:40:51,202 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27070.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:41:37,196 INFO [train.py:904] (0/8) Epoch 3, batch 6800, loss[loss=0.332, simple_loss=0.374, pruned_loss=0.145, over 11574.00 frames. ], tot_loss[loss=0.2793, simple_loss=0.3475, pruned_loss=0.1055, over 3113451.72 frames. ], batch size: 246, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:42:03,930 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27118.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:31,279 INFO [optim.py:368] (0/8) 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 23:42:51,194 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27148.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:42:55,402 INFO [train.py:904] (0/8) Epoch 3, batch 6850, loss[loss=0.251, simple_loss=0.3397, pruned_loss=0.08115, over 16840.00 frames. ], tot_loss[loss=0.2809, simple_loss=0.3493, pruned_loss=0.1063, over 3098512.41 frames. ], batch size: 42, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:44:10,178 INFO [train.py:904] (0/8) Epoch 3, batch 6900, loss[loss=0.3747, simple_loss=0.4029, pruned_loss=0.1733, over 11546.00 frames. ], tot_loss[loss=0.2814, simple_loss=0.3515, pruned_loss=0.1057, over 3100194.10 frames. ], batch size: 247, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:44:22,763 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27209.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:44:23,224 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-27 23:44:32,963 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-27 23:45:02,583 INFO [optim.py:368] (0/8) 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 23:45:28,464 INFO [train.py:904] (0/8) Epoch 3, batch 6950, loss[loss=0.2789, simple_loss=0.3541, pruned_loss=0.1018, over 16648.00 frames. ], tot_loss[loss=0.2828, simple_loss=0.3526, pruned_loss=0.1065, over 3116697.09 frames. ], batch size: 134, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:45:41,664 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27260.0, num_to_drop=1, layers_to_drop={3} 2023-04-27 23:46:43,371 INFO [train.py:904] (0/8) Epoch 3, batch 7000, loss[loss=0.3168, simple_loss=0.3884, pruned_loss=0.1226, over 16424.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3518, pruned_loss=0.1047, over 3141059.18 frames. ], batch size: 146, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:46:51,002 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27306.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:47:35,872 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 3, batch 7050, loss[loss=0.2795, simple_loss=0.3513, pruned_loss=0.1039, over 16301.00 frames. ], tot_loss[loss=0.2829, simple_loss=0.3539, pruned_loss=0.106, over 3133679.65 frames. ], batch size: 165, lr: 1.98e-02, grad_scale: 8.0 2023-04-27 23:48:22,052 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-27 23:49:19,641 INFO [train.py:904] (0/8) Epoch 3, batch 7100, loss[loss=0.2929, simple_loss=0.3621, pruned_loss=0.1119, over 16897.00 frames. ], tot_loss[loss=0.2834, simple_loss=0.3534, pruned_loss=0.1067, over 3110158.01 frames. ], batch size: 109, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:49:24,432 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7990, 5.1065, 4.8443, 4.8128, 4.4042, 4.4201, 4.5747, 5.1367], device='cuda:0'), covar=tensor([0.0437, 0.0532, 0.0758, 0.0357, 0.0493, 0.0578, 0.0465, 0.0520], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0371, 0.0333, 0.0239, 0.0239, 0.0240, 0.0299, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:50:12,111 INFO [optim.py:368] (0/8) 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 23:50:25,324 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0721, 2.5752, 2.2730, 3.3455, 3.1414, 3.1736, 1.8574, 2.7049], device='cuda:0'), covar=tensor([0.1011, 0.0357, 0.1002, 0.0078, 0.0231, 0.0344, 0.1028, 0.0618], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0131, 0.0161, 0.0070, 0.0133, 0.0142, 0.0151, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-27 23:50:36,347 INFO [train.py:904] (0/8) Epoch 3, batch 7150, loss[loss=0.2701, simple_loss=0.3468, pruned_loss=0.09664, over 16447.00 frames. ], tot_loss[loss=0.2818, simple_loss=0.3511, pruned_loss=0.1062, over 3112992.87 frames. ], batch size: 75, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:50:49,261 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 23:51:51,150 INFO [train.py:904] (0/8) Epoch 3, batch 7200, loss[loss=0.2501, simple_loss=0.3223, pruned_loss=0.08899, over 12220.00 frames. ], tot_loss[loss=0.2777, simple_loss=0.348, pruned_loss=0.1037, over 3104657.50 frames. ], batch size: 248, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:51:55,816 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=27504.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:52:27,097 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7219, 3.6587, 3.8228, 3.6079, 3.7143, 4.1042, 3.9139, 3.6512], device='cuda:0'), covar=tensor([0.1474, 0.1337, 0.0952, 0.1702, 0.1811, 0.1058, 0.0910, 0.2013], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0307, 0.0288, 0.0269, 0.0353, 0.0316, 0.0251, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:52:45,516 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.403e+02 3.669e+02 4.524e+02 6.083e+02 1.066e+03, threshold=9.047e+02, percent-clipped=1.0 2023-04-27 23:53:12,432 INFO [train.py:904] (0/8) Epoch 3, batch 7250, loss[loss=0.2415, simple_loss=0.3188, pruned_loss=0.08205, over 16773.00 frames. ], tot_loss[loss=0.2737, simple_loss=0.3446, pruned_loss=0.1014, over 3113524.02 frames. ], batch size: 83, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:53:25,922 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27560.0, num_to_drop=1, layers_to_drop={2} 2023-04-27 23:53:31,940 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-27 23:53:36,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9004, 3.9659, 3.3789, 2.7610, 3.2924, 2.6506, 4.3890, 4.5289], device='cuda:0'), covar=tensor([0.2223, 0.0805, 0.1181, 0.1057, 0.1846, 0.1118, 0.0372, 0.0396], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0242, 0.0255, 0.0214, 0.0300, 0.0192, 0.0221, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-27 23:53:47,360 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-27 23:54:26,525 INFO [train.py:904] (0/8) Epoch 3, batch 7300, loss[loss=0.2434, simple_loss=0.3212, pruned_loss=0.08282, over 17015.00 frames. ], tot_loss[loss=0.2716, simple_loss=0.343, pruned_loss=0.1001, over 3138707.35 frames. ], batch size: 53, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:54:35,572 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27606.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:54:38,058 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27608.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:55:17,371 INFO [optim.py:368] (0/8) 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 23:55:40,981 INFO [train.py:904] (0/8) Epoch 3, batch 7350, loss[loss=0.2321, simple_loss=0.3138, pruned_loss=0.07522, over 16603.00 frames. ], tot_loss[loss=0.2715, simple_loss=0.3426, pruned_loss=0.1002, over 3106025.41 frames. ], batch size: 68, lr: 1.97e-02, grad_scale: 8.0 2023-04-27 23:55:46,127 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27654.0, num_to_drop=0, layers_to_drop=set() 2023-04-27 23:56:00,177 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-27 23:56:59,430 INFO [train.py:904] (0/8) Epoch 3, batch 7400, loss[loss=0.3438, simple_loss=0.3793, pruned_loss=0.1542, over 11311.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.3444, pruned_loss=0.102, over 3097257.94 frames. ], batch size: 249, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:57:13,285 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-27 23:57:20,327 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=27714.0, num_to_drop=1, layers_to_drop={1} 2023-04-27 23:57:40,437 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8672, 3.2999, 3.6335, 2.1955, 3.3897, 3.4921, 3.6104, 1.7049], device='cuda:0'), covar=tensor([0.0344, 0.0029, 0.0029, 0.0245, 0.0031, 0.0059, 0.0019, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0052, 0.0057, 0.0111, 0.0052, 0.0061, 0.0057, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:57:55,130 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.894e+02 4.696e+02 5.728e+02 6.981e+02 1.392e+03, threshold=1.146e+03, percent-clipped=2.0 2023-04-27 23:58:18,295 INFO [train.py:904] (0/8) Epoch 3, batch 7450, loss[loss=0.2843, simple_loss=0.3634, pruned_loss=0.1026, over 16745.00 frames. ], tot_loss[loss=0.2757, simple_loss=0.3456, pruned_loss=0.103, over 3113399.15 frames. ], batch size: 124, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:58:58,312 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=27775.0, num_to_drop=1, layers_to_drop={0} 2023-04-27 23:59:39,396 INFO [train.py:904] (0/8) Epoch 3, batch 7500, loss[loss=0.272, simple_loss=0.3439, pruned_loss=0.1001, over 16240.00 frames. ], tot_loss[loss=0.2772, simple_loss=0.3473, pruned_loss=0.1035, over 3101524.48 frames. ], batch size: 165, lr: 1.96e-02, grad_scale: 4.0 2023-04-27 23:59:41,878 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1493, 3.7558, 3.9611, 2.6436, 3.6875, 3.7354, 3.9128, 2.0589], device='cuda:0'), covar=tensor([0.0312, 0.0021, 0.0026, 0.0225, 0.0029, 0.0063, 0.0020, 0.0296], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0051, 0.0056, 0.0108, 0.0051, 0.0061, 0.0056, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-27 23:59:44,141 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=27804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:00:07,724 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8917, 4.0692, 3.7969, 3.9288, 3.6306, 3.7148, 3.8162, 4.0230], device='cuda:0'), covar=tensor([0.0518, 0.0693, 0.0885, 0.0412, 0.0483, 0.0922, 0.0521, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0379, 0.0340, 0.0240, 0.0245, 0.0250, 0.0302, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:00:33,136 INFO [optim.py:368] (0/8) 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 00:00:43,370 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4974, 2.9413, 2.5930, 2.3399, 2.3191, 1.9834, 2.8841, 3.0284], device='cuda:0'), covar=tensor([0.1312, 0.0618, 0.0872, 0.0885, 0.1621, 0.1190, 0.0335, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0242, 0.0256, 0.0216, 0.0296, 0.0192, 0.0218, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:00:55,952 INFO [train.py:904] (0/8) Epoch 3, batch 7550, loss[loss=0.2349, simple_loss=0.3037, pruned_loss=0.08305, over 17072.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.346, pruned_loss=0.1034, over 3084671.20 frames. ], batch size: 53, lr: 1.96e-02, grad_scale: 4.0 2023-04-28 00:00:58,224 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=27852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:01:31,866 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7005, 3.4111, 3.6594, 3.4947, 3.7087, 4.0887, 3.9286, 3.5644], device='cuda:0'), covar=tensor([0.1566, 0.1804, 0.1251, 0.1959, 0.2323, 0.1144, 0.1137, 0.2166], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0311, 0.0297, 0.0271, 0.0361, 0.0318, 0.0257, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 00:02:07,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3881, 4.3605, 4.2208, 3.5025, 4.2529, 1.6061, 3.9543, 4.0795], device='cuda:0'), covar=tensor([0.0054, 0.0046, 0.0067, 0.0291, 0.0050, 0.1396, 0.0060, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0062, 0.0096, 0.0110, 0.0070, 0.0121, 0.0083, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:02:13,370 INFO [train.py:904] (0/8) Epoch 3, batch 7600, loss[loss=0.2675, simple_loss=0.3409, pruned_loss=0.09702, over 16953.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3452, pruned_loss=0.1038, over 3067875.24 frames. ], batch size: 109, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:03:08,900 INFO [optim.py:368] (0/8) 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 00:03:31,749 INFO [train.py:904] (0/8) Epoch 3, batch 7650, loss[loss=0.2693, simple_loss=0.3402, pruned_loss=0.09924, over 16746.00 frames. ], tot_loss[loss=0.2782, simple_loss=0.346, pruned_loss=0.1052, over 3060496.93 frames. ], batch size: 124, lr: 1.96e-02, grad_scale: 8.0 2023-04-28 00:04:48,389 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-28000.pt 2023-04-28 00:04:52,692 INFO [train.py:904] (0/8) Epoch 3, batch 7700, loss[loss=0.26, simple_loss=0.3386, pruned_loss=0.0907, over 16900.00 frames. ], tot_loss[loss=0.2779, simple_loss=0.3457, pruned_loss=0.1051, over 3086756.69 frames. ], batch size: 96, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:05:46,978 INFO [optim.py:368] (0/8) 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 00:06:11,257 INFO [train.py:904] (0/8) Epoch 3, batch 7750, loss[loss=0.288, simple_loss=0.3429, pruned_loss=0.1165, over 11696.00 frames. ], tot_loss[loss=0.2779, simple_loss=0.346, pruned_loss=0.1049, over 3094650.89 frames. ], batch size: 248, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:06:16,625 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4944, 3.5230, 3.9946, 3.9724, 3.9373, 3.5815, 3.6386, 3.6510], device='cuda:0'), covar=tensor([0.0257, 0.0387, 0.0276, 0.0310, 0.0402, 0.0293, 0.0834, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0176, 0.0189, 0.0189, 0.0230, 0.0197, 0.0299, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 00:06:26,138 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 00:06:40,936 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28070.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:07:28,338 INFO [train.py:904] (0/8) Epoch 3, batch 7800, loss[loss=0.2482, simple_loss=0.3257, pruned_loss=0.08536, over 16512.00 frames. ], tot_loss[loss=0.2789, simple_loss=0.3469, pruned_loss=0.1055, over 3106607.70 frames. ], batch size: 75, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:07:33,940 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3926, 3.1030, 2.5953, 2.1946, 2.4028, 2.1379, 3.1736, 3.4943], device='cuda:0'), covar=tensor([0.1880, 0.0687, 0.1141, 0.1186, 0.1571, 0.1147, 0.0407, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0241, 0.0256, 0.0218, 0.0300, 0.0191, 0.0219, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:07:54,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6277, 3.4445, 3.5558, 2.5355, 3.3382, 3.6409, 3.6423, 1.8914], device='cuda:0'), covar=tensor([0.0368, 0.0029, 0.0030, 0.0194, 0.0032, 0.0038, 0.0022, 0.0300], device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0051, 0.0054, 0.0105, 0.0052, 0.0059, 0.0056, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 00:08:22,849 INFO [optim.py:368] (0/8) 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 00:08:24,711 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9807, 4.7405, 4.9546, 5.2904, 5.3539, 4.7044, 5.3964, 5.2415], device='cuda:0'), covar=tensor([0.0615, 0.0626, 0.0975, 0.0323, 0.0313, 0.0356, 0.0298, 0.0347], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0365, 0.0467, 0.0359, 0.0273, 0.0259, 0.0291, 0.0301], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:08:45,066 INFO [train.py:904] (0/8) Epoch 3, batch 7850, loss[loss=0.3203, simple_loss=0.3633, pruned_loss=0.1387, over 11134.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.3481, pruned_loss=0.1053, over 3099020.61 frames. ], batch size: 248, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:09:43,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7130, 4.7145, 4.4880, 3.7915, 4.4462, 1.9638, 4.2936, 4.5283], device='cuda:0'), covar=tensor([0.0049, 0.0037, 0.0056, 0.0276, 0.0051, 0.1295, 0.0057, 0.0079], device='cuda:0'), in_proj_covar=tensor([0.0073, 0.0061, 0.0094, 0.0109, 0.0070, 0.0119, 0.0082, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:10:00,841 INFO [train.py:904] (0/8) Epoch 3, batch 7900, loss[loss=0.2993, simple_loss=0.3496, pruned_loss=0.1245, over 11562.00 frames. ], tot_loss[loss=0.2789, simple_loss=0.3478, pruned_loss=0.105, over 3092483.40 frames. ], batch size: 246, lr: 1.95e-02, grad_scale: 8.0 2023-04-28 00:10:45,196 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:10:55,800 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.692e+02 4.334e+02 4.942e+02 5.999e+02 2.073e+03, threshold=9.884e+02, percent-clipped=3.0 2023-04-28 00:11:18,500 INFO [train.py:904] (0/8) Epoch 3, batch 7950, loss[loss=0.2586, simple_loss=0.3293, pruned_loss=0.09391, over 16211.00 frames. ], tot_loss[loss=0.2781, simple_loss=0.3474, pruned_loss=0.1045, over 3117098.56 frames. ], batch size: 165, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:11:26,026 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28255.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:12:13,431 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28287.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:18,119 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:12:22,992 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1597, 2.9684, 2.7402, 2.0145, 2.5281, 2.0692, 2.7419, 2.9555], device='cuda:0'), covar=tensor([0.0288, 0.0451, 0.0485, 0.1233, 0.0617, 0.0879, 0.0534, 0.0403], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0116, 0.0155, 0.0143, 0.0136, 0.0129, 0.0143, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 00:12:24,647 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-04-28 00:12:29,832 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9383, 3.2488, 2.4728, 4.7458, 4.4645, 4.0868, 1.9514, 3.3026], device='cuda:0'), covar=tensor([0.1294, 0.0471, 0.1165, 0.0056, 0.0169, 0.0246, 0.1165, 0.0592], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0136, 0.0163, 0.0075, 0.0139, 0.0143, 0.0154, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-04-28 00:12:33,520 INFO [train.py:904] (0/8) Epoch 3, batch 8000, loss[loss=0.3844, simple_loss=0.4055, pruned_loss=0.1816, over 11520.00 frames. ], tot_loss[loss=0.2795, simple_loss=0.3479, pruned_loss=0.1056, over 3108061.73 frames. ], batch size: 246, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:12:55,773 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28316.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 00:13:27,064 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.455e+02 4.065e+02 5.202e+02 6.886e+02 1.574e+03, threshold=1.040e+03, percent-clipped=4.0 2023-04-28 00:13:45,825 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:13:49,306 INFO [train.py:904] (0/8) Epoch 3, batch 8050, loss[loss=0.2616, simple_loss=0.3347, pruned_loss=0.09431, over 16385.00 frames. ], tot_loss[loss=0.2792, simple_loss=0.3476, pruned_loss=0.1054, over 3101100.97 frames. ], batch size: 146, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:14:18,187 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28370.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 00:15:05,415 INFO [train.py:904] (0/8) Epoch 3, batch 8100, loss[loss=0.3558, simple_loss=0.3932, pruned_loss=0.1592, over 11494.00 frames. ], tot_loss[loss=0.2779, simple_loss=0.3468, pruned_loss=0.1045, over 3101897.38 frames. ], batch size: 247, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:15:30,574 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28418.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:15:57,046 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.864e+02 4.688e+02 5.875e+02 7.578e+02 1.406e+03, threshold=1.175e+03, percent-clipped=5.0 2023-04-28 00:16:20,015 INFO [train.py:904] (0/8) Epoch 3, batch 8150, loss[loss=0.2964, simple_loss=0.3486, pruned_loss=0.1221, over 11812.00 frames. ], tot_loss[loss=0.2764, simple_loss=0.3447, pruned_loss=0.1041, over 3083556.58 frames. ], batch size: 247, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:17:14,588 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:17:36,831 INFO [train.py:904] (0/8) Epoch 3, batch 8200, loss[loss=0.3276, simple_loss=0.367, pruned_loss=0.1441, over 11129.00 frames. ], tot_loss[loss=0.2727, simple_loss=0.3411, pruned_loss=0.1022, over 3100007.89 frames. ], batch size: 246, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:18:31,949 INFO [optim.py:368] (0/8) 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 00:18:50,499 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28547.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:18:55,854 INFO [train.py:904] (0/8) Epoch 3, batch 8250, loss[loss=0.2764, simple_loss=0.3525, pruned_loss=0.1002, over 16749.00 frames. ], tot_loss[loss=0.2705, simple_loss=0.3405, pruned_loss=0.1002, over 3102323.16 frames. ], batch size: 124, lr: 1.94e-02, grad_scale: 8.0 2023-04-28 00:18:59,669 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:19:51,815 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:20:16,997 INFO [train.py:904] (0/8) Epoch 3, batch 8300, loss[loss=0.2501, simple_loss=0.3127, pruned_loss=0.09368, over 11967.00 frames. ], tot_loss[loss=0.264, simple_loss=0.3363, pruned_loss=0.09586, over 3079280.72 frames. ], batch size: 248, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:20:33,948 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28611.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 00:21:14,649 INFO [optim.py:368] (0/8) 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 00:21:27,054 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:21:39,720 INFO [train.py:904] (0/8) Epoch 3, batch 8350, loss[loss=0.2468, simple_loss=0.3249, pruned_loss=0.0844, over 15379.00 frames. ], tot_loss[loss=0.2586, simple_loss=0.333, pruned_loss=0.09209, over 3064976.27 frames. ], batch size: 191, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:00,538 INFO [train.py:904] (0/8) Epoch 3, batch 8400, loss[loss=0.2389, simple_loss=0.3108, pruned_loss=0.08349, over 12584.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.3302, pruned_loss=0.08945, over 3056158.36 frames. ], batch size: 247, lr: 1.93e-02, grad_scale: 8.0 2023-04-28 00:23:58,231 INFO [optim.py:368] (0/8) 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 00:24:20,123 INFO [train.py:904] (0/8) Epoch 3, batch 8450, loss[loss=0.2819, simple_loss=0.3352, pruned_loss=0.1143, over 12565.00 frames. ], tot_loss[loss=0.2513, simple_loss=0.3278, pruned_loss=0.08737, over 3046359.35 frames. ], batch size: 247, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:25:42,034 INFO [train.py:904] (0/8) Epoch 3, batch 8500, loss[loss=0.2264, simple_loss=0.3039, pruned_loss=0.07442, over 15120.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3227, pruned_loss=0.08378, over 3033948.27 frames. ], batch size: 190, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:26:31,719 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8966, 3.1275, 3.0551, 2.1561, 2.9682, 3.0639, 2.9560, 1.8343], device='cuda:0'), covar=tensor([0.0299, 0.0022, 0.0032, 0.0213, 0.0029, 0.0047, 0.0029, 0.0304], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0052, 0.0055, 0.0109, 0.0051, 0.0061, 0.0057, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 00:26:40,696 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.668e+02 4.634e+02 5.862e+02 2.485e+03, threshold=9.268e+02, percent-clipped=6.0 2023-04-28 00:26:50,011 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:00,180 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=28848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:05,707 INFO [train.py:904] (0/8) Epoch 3, batch 8550, loss[loss=0.2142, simple_loss=0.2847, pruned_loss=0.07181, over 11690.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3198, pruned_loss=0.08258, over 3007847.06 frames. ], batch size: 247, lr: 1.93e-02, grad_scale: 4.0 2023-04-28 00:27:25,984 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=28861.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:27:58,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1134, 1.3769, 1.5040, 2.1010, 2.0823, 2.0101, 1.4114, 2.1831], device='cuda:0'), covar=tensor([0.0054, 0.0227, 0.0184, 0.0124, 0.0092, 0.0114, 0.0238, 0.0047], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0121, 0.0109, 0.0100, 0.0096, 0.0071, 0.0118, 0.0064], device='cuda:0'), out_proj_covar=tensor([1.3328e-04, 1.8736e-04, 1.7334e-04, 1.5716e-04, 1.4778e-04, 1.0695e-04, 1.7944e-04, 9.6015e-05], device='cuda:0') 2023-04-28 00:28:12,902 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:28:46,471 INFO [train.py:904] (0/8) Epoch 3, batch 8600, loss[loss=0.2417, simple_loss=0.3235, pruned_loss=0.0799, over 15288.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.3203, pruned_loss=0.08164, over 2996807.12 frames. ], batch size: 190, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:29:07,497 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28911.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:29:30,915 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=28922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:50,788 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:29:57,556 INFO [optim.py:368] (0/8) 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 00:30:11,159 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=28943.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:30:25,446 INFO [train.py:904] (0/8) Epoch 3, batch 8650, loss[loss=0.2058, simple_loss=0.2906, pruned_loss=0.06048, over 12092.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3176, pruned_loss=0.07909, over 2993641.01 frames. ], batch size: 246, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:30:45,514 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28959.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 00:31:40,450 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9744, 3.6911, 3.6648, 2.7017, 3.2308, 3.4783, 3.3937, 1.7089], device='cuda:0'), covar=tensor([0.0328, 0.0015, 0.0024, 0.0186, 0.0029, 0.0030, 0.0028, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0051, 0.0054, 0.0106, 0.0051, 0.0058, 0.0055, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0003, 0.0001, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 00:31:52,872 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=28991.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:32:12,040 INFO [train.py:904] (0/8) Epoch 3, batch 8700, loss[loss=0.245, simple_loss=0.3253, pruned_loss=0.08231, over 15316.00 frames. ], tot_loss[loss=0.2333, simple_loss=0.3135, pruned_loss=0.07653, over 3019863.73 frames. ], batch size: 190, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:32:15,218 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29002.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:33:18,501 INFO [optim.py:368] (0/8) 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 00:33:46,765 INFO [train.py:904] (0/8) Epoch 3, batch 8750, loss[loss=0.2221, simple_loss=0.3109, pruned_loss=0.0666, over 15204.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3127, pruned_loss=0.07492, over 3036088.51 frames. ], batch size: 190, lr: 1.92e-02, grad_scale: 4.0 2023-04-28 00:33:53,273 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 00:34:18,769 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29063.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:35:38,463 INFO [train.py:904] (0/8) Epoch 3, batch 8800, loss[loss=0.2289, simple_loss=0.3227, pruned_loss=0.06753, over 15488.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3109, pruned_loss=0.07363, over 3029678.04 frames. ], batch size: 191, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:35:49,507 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2822, 3.8143, 3.7677, 1.4228, 4.0215, 3.9738, 3.2066, 2.7380], device='cuda:0'), covar=tensor([0.0808, 0.0120, 0.0168, 0.1475, 0.0046, 0.0040, 0.0245, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0081, 0.0076, 0.0144, 0.0069, 0.0069, 0.0108, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 00:36:18,348 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9615, 2.7854, 2.6585, 1.6590, 2.8722, 2.8582, 2.5398, 2.3214], device='cuda:0'), covar=tensor([0.0743, 0.0156, 0.0176, 0.1221, 0.0095, 0.0086, 0.0315, 0.0435], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0081, 0.0077, 0.0144, 0.0069, 0.0069, 0.0108, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 00:36:52,046 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.420e+02 3.673e+02 4.669e+02 5.868e+02 1.100e+03, threshold=9.339e+02, percent-clipped=8.0 2023-04-28 00:37:03,464 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:17,051 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:37:21,278 INFO [train.py:904] (0/8) Epoch 3, batch 8850, loss[loss=0.2096, simple_loss=0.3093, pruned_loss=0.05499, over 16541.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3132, pruned_loss=0.07256, over 3032612.29 frames. ], batch size: 62, lr: 1.92e-02, grad_scale: 8.0 2023-04-28 00:38:45,267 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:38:57,576 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:39:06,873 INFO [train.py:904] (0/8) Epoch 3, batch 8900, loss[loss=0.2522, simple_loss=0.325, pruned_loss=0.08972, over 15348.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3132, pruned_loss=0.07177, over 3039992.78 frames. ], batch size: 191, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:39:40,219 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:35,244 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:40:36,247 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 3.766e+02 4.506e+02 5.745e+02 1.186e+03, threshold=9.012e+02, percent-clipped=1.0 2023-04-28 00:41:10,895 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9025, 3.6594, 3.2696, 2.0043, 2.7135, 2.1594, 3.2340, 3.7839], device='cuda:0'), covar=tensor([0.0294, 0.0444, 0.0498, 0.1594, 0.0716, 0.1057, 0.0787, 0.0435], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0111, 0.0153, 0.0145, 0.0134, 0.0131, 0.0141, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-28 00:41:11,866 INFO [train.py:904] (0/8) Epoch 3, batch 8950, loss[loss=0.2138, simple_loss=0.2947, pruned_loss=0.06638, over 16797.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3136, pruned_loss=0.07272, over 3051851.05 frames. ], batch size: 124, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:42:53,670 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:43:00,679 INFO [train.py:904] (0/8) Epoch 3, batch 9000, loss[loss=0.2209, simple_loss=0.3037, pruned_loss=0.06904, over 16387.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.309, pruned_loss=0.07013, over 3062635.57 frames. ], batch size: 146, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:43:00,680 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 00:43:11,839 INFO [train.py:938] (0/8) Epoch 3, validation: loss=0.1886, simple_loss=0.2904, pruned_loss=0.04341, over 944034.00 frames. 2023-04-28 00:43:11,840 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-28 00:43:59,195 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1595, 4.0633, 4.0332, 3.4845, 3.9369, 1.6295, 3.7675, 3.9115], device='cuda:0'), covar=tensor([0.0048, 0.0041, 0.0055, 0.0197, 0.0051, 0.1423, 0.0066, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0059, 0.0094, 0.0098, 0.0070, 0.0121, 0.0082, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:44:26,940 INFO [optim.py:368] (0/8) 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 00:44:57,072 INFO [train.py:904] (0/8) Epoch 3, batch 9050, loss[loss=0.2329, simple_loss=0.3046, pruned_loss=0.08059, over 16903.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.3107, pruned_loss=0.07132, over 3064328.22 frames. ], batch size: 109, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:45:12,126 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:46:41,633 INFO [train.py:904] (0/8) Epoch 3, batch 9100, loss[loss=0.209, simple_loss=0.2858, pruned_loss=0.06611, over 12025.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3104, pruned_loss=0.07192, over 3063878.95 frames. ], batch size: 247, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:48:08,508 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.097e+02 4.371e+02 5.293e+02 6.986e+02 1.215e+03, threshold=1.059e+03, percent-clipped=12.0 2023-04-28 00:48:40,651 INFO [train.py:904] (0/8) Epoch 3, batch 9150, loss[loss=0.2215, simple_loss=0.3098, pruned_loss=0.06661, over 15302.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.311, pruned_loss=0.07183, over 3046560.67 frames. ], batch size: 190, lr: 1.91e-02, grad_scale: 8.0 2023-04-28 00:49:32,883 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3671, 1.4553, 1.8784, 2.2503, 2.1897, 2.3236, 1.4348, 2.3270], device='cuda:0'), covar=tensor([0.0056, 0.0184, 0.0108, 0.0095, 0.0076, 0.0070, 0.0173, 0.0049], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0123, 0.0110, 0.0099, 0.0096, 0.0071, 0.0115, 0.0063], device='cuda:0'), out_proj_covar=tensor([1.3439e-04, 1.8935e-04, 1.7291e-04, 1.5463e-04, 1.4665e-04, 1.0595e-04, 1.7519e-04, 9.4004e-05], device='cuda:0') 2023-04-28 00:49:42,418 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8196, 3.3552, 3.4419, 2.2077, 3.2256, 3.2953, 3.4776, 1.9583], device='cuda:0'), covar=tensor([0.0356, 0.0022, 0.0033, 0.0256, 0.0038, 0.0040, 0.0019, 0.0305], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0052, 0.0055, 0.0107, 0.0051, 0.0060, 0.0055, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0001, 0.0002, 0.0001, 0.0003], device='cuda:0') 2023-04-28 00:50:24,883 INFO [train.py:904] (0/8) Epoch 3, batch 9200, loss[loss=0.2087, simple_loss=0.2759, pruned_loss=0.07076, over 12226.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3056, pruned_loss=0.06999, over 3052313.80 frames. ], batch size: 248, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:50:54,915 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29517.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:51:32,425 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.365e+02 4.043e+02 4.899e+02 6.737e+02 1.438e+03, threshold=9.797e+02, percent-clipped=1.0 2023-04-28 00:51:48,228 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9469, 3.8907, 3.8285, 3.4632, 3.7404, 1.8086, 3.5609, 3.7114], device='cuda:0'), covar=tensor([0.0062, 0.0049, 0.0081, 0.0209, 0.0069, 0.1515, 0.0092, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0061, 0.0097, 0.0098, 0.0071, 0.0123, 0.0085, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:52:01,387 INFO [train.py:904] (0/8) Epoch 3, batch 9250, loss[loss=0.1774, simple_loss=0.2714, pruned_loss=0.04172, over 16875.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3052, pruned_loss=0.07011, over 3041772.63 frames. ], batch size: 96, lr: 1.90e-02, grad_scale: 8.0 2023-04-28 00:52:02,460 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29551.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:29,200 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:52:59,360 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8510, 3.5626, 3.8169, 3.7276, 3.9297, 4.1878, 3.9877, 3.7087], device='cuda:0'), covar=tensor([0.1145, 0.1798, 0.1098, 0.1825, 0.2106, 0.1005, 0.0849, 0.2224], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0295, 0.0286, 0.0264, 0.0336, 0.0313, 0.0242, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:53:31,872 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:53:50,326 INFO [train.py:904] (0/8) Epoch 3, batch 9300, loss[loss=0.2088, simple_loss=0.2893, pruned_loss=0.06419, over 16817.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3028, pruned_loss=0.06864, over 3042840.33 frames. ], batch size: 124, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:53:53,253 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4048, 4.6525, 4.4165, 4.3926, 4.0658, 3.9891, 4.1786, 4.6549], device='cuda:0'), covar=tensor([0.0422, 0.0521, 0.0609, 0.0362, 0.0440, 0.0839, 0.0481, 0.0531], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0358, 0.0303, 0.0232, 0.0233, 0.0239, 0.0289, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:54:16,033 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:55:11,310 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.186e+02 3.534e+02 4.207e+02 4.979e+02 1.086e+03, threshold=8.414e+02, percent-clipped=1.0 2023-04-28 00:55:35,085 INFO [train.py:904] (0/8) Epoch 3, batch 9350, loss[loss=0.2228, simple_loss=0.3076, pruned_loss=0.06901, over 16566.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.303, pruned_loss=0.06817, over 3062998.75 frames. ], batch size: 68, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:55:51,111 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29658.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:56:00,146 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:56:40,422 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3526, 4.6289, 4.3749, 4.4444, 4.0559, 4.0204, 4.2598, 4.6240], device='cuda:0'), covar=tensor([0.0476, 0.0582, 0.0739, 0.0363, 0.0475, 0.0872, 0.0498, 0.0585], device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0362, 0.0308, 0.0233, 0.0235, 0.0244, 0.0293, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:57:16,976 INFO [train.py:904] (0/8) Epoch 3, batch 9400, loss[loss=0.2282, simple_loss=0.3183, pruned_loss=0.06904, over 15340.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3034, pruned_loss=0.06786, over 3076986.11 frames. ], batch size: 190, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 00:57:27,904 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29706.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 00:57:44,211 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1347, 4.1088, 3.8612, 3.9143, 3.6298, 4.0534, 3.8029, 3.7262], device='cuda:0'), covar=tensor([0.0273, 0.0198, 0.0194, 0.0129, 0.0521, 0.0175, 0.0457, 0.0297], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0123, 0.0165, 0.0135, 0.0188, 0.0151, 0.0117, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 00:58:01,541 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29723.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 00:58:32,713 INFO [optim.py:368] (0/8) 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 00:58:58,189 INFO [train.py:904] (0/8) Epoch 3, batch 9450, loss[loss=0.2084, simple_loss=0.2961, pruned_loss=0.06039, over 16899.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3056, pruned_loss=0.06861, over 3070310.42 frames. ], batch size: 116, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:37,286 INFO [train.py:904] (0/8) Epoch 3, batch 9500, loss[loss=0.2166, simple_loss=0.3046, pruned_loss=0.06429, over 16681.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3043, pruned_loss=0.06772, over 3081616.47 frames. ], batch size: 134, lr: 1.90e-02, grad_scale: 4.0 2023-04-28 01:00:41,194 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=29802.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:01:36,635 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 01:01:43,155 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6313, 2.8218, 2.3676, 3.9481, 3.6637, 3.9448, 1.5802, 2.7622], device='cuda:0'), covar=tensor([0.1616, 0.0574, 0.1326, 0.0088, 0.0278, 0.0289, 0.1474, 0.0843], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0132, 0.0160, 0.0066, 0.0131, 0.0140, 0.0152, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-04-28 01:01:50,895 INFO [optim.py:368] (0/8) 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 01:02:13,898 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0859, 3.0778, 2.6621, 1.8834, 2.6227, 2.0129, 2.7583, 2.9041], device='cuda:0'), covar=tensor([0.0232, 0.0340, 0.0496, 0.1477, 0.0582, 0.1040, 0.0594, 0.0585], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0109, 0.0150, 0.0143, 0.0133, 0.0129, 0.0138, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 01:02:22,323 INFO [train.py:904] (0/8) Epoch 3, batch 9550, loss[loss=0.239, simple_loss=0.3196, pruned_loss=0.07916, over 16372.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.304, pruned_loss=0.06792, over 3079620.55 frames. ], batch size: 146, lr: 1.89e-02, grad_scale: 4.0 2023-04-28 01:02:49,232 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=29863.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:03:48,944 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=29892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:03,537 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7905, 1.6400, 1.3824, 1.2952, 1.8819, 1.5929, 1.7582, 1.9219], device='cuda:0'), covar=tensor([0.0019, 0.0114, 0.0137, 0.0121, 0.0065, 0.0104, 0.0050, 0.0063], device='cuda:0'), in_proj_covar=tensor([0.0054, 0.0123, 0.0123, 0.0123, 0.0113, 0.0123, 0.0077, 0.0097], device='cuda:0'), out_proj_covar=tensor([6.8831e-05, 1.7101e-04, 1.6483e-04, 1.6814e-04, 1.5784e-04, 1.7126e-04, 1.0362e-04, 1.3452e-04], device='cuda:0') 2023-04-28 01:04:04,169 INFO [train.py:904] (0/8) Epoch 3, batch 9600, loss[loss=0.238, simple_loss=0.3005, pruned_loss=0.0877, over 12216.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3051, pruned_loss=0.06906, over 3077506.08 frames. ], batch size: 246, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:04:16,351 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=29907.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:04:42,504 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8141, 5.4902, 5.4677, 5.4204, 5.3427, 5.9276, 5.4120, 5.1871], device='cuda:0'), covar=tensor([0.0678, 0.1100, 0.1080, 0.1415, 0.1793, 0.0666, 0.1201, 0.1839], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0302, 0.0291, 0.0264, 0.0343, 0.0314, 0.0243, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:04:55,614 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3440, 1.4280, 1.6900, 2.2647, 2.1758, 2.2062, 1.3394, 2.2495], device='cuda:0'), covar=tensor([0.0054, 0.0201, 0.0131, 0.0109, 0.0083, 0.0065, 0.0187, 0.0056], device='cuda:0'), in_proj_covar=tensor([0.0092, 0.0124, 0.0111, 0.0102, 0.0099, 0.0072, 0.0116, 0.0063], device='cuda:0'), out_proj_covar=tensor([1.3725e-04, 1.8959e-04, 1.7307e-04, 1.5864e-04, 1.4989e-04, 1.0627e-04, 1.7632e-04, 9.4291e-05], device='cuda:0') 2023-04-28 01:05:18,578 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.324e+02 3.589e+02 4.321e+02 5.133e+02 1.002e+03, threshold=8.641e+02, percent-clipped=3.0 2023-04-28 01:05:23,704 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=29940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:05:28,566 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 01:05:51,210 INFO [train.py:904] (0/8) Epoch 3, batch 9650, loss[loss=0.2178, simple_loss=0.3043, pruned_loss=0.06566, over 16796.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3072, pruned_loss=0.06944, over 3085591.94 frames. ], batch size: 83, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:06:22,041 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6872, 2.0008, 1.7512, 1.6947, 2.6858, 2.2891, 2.9797, 2.9497], device='cuda:0'), covar=tensor([0.0017, 0.0157, 0.0189, 0.0189, 0.0083, 0.0136, 0.0041, 0.0054], device='cuda:0'), in_proj_covar=tensor([0.0054, 0.0123, 0.0122, 0.0122, 0.0113, 0.0122, 0.0076, 0.0096], device='cuda:0'), out_proj_covar=tensor([6.8694e-05, 1.7069e-04, 1.6361e-04, 1.6639e-04, 1.5820e-04, 1.6912e-04, 1.0222e-04, 1.3357e-04], device='cuda:0') 2023-04-28 01:06:31,103 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5758, 3.7672, 3.1307, 2.2403, 2.6620, 2.2554, 3.8659, 4.1225], device='cuda:0'), covar=tensor([0.2126, 0.0608, 0.1097, 0.1275, 0.1702, 0.1225, 0.0359, 0.0344], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0236, 0.0253, 0.0211, 0.0225, 0.0192, 0.0211, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:07:15,909 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3792, 3.2333, 2.8243, 2.1428, 2.2932, 2.0635, 3.2660, 3.4706], device='cuda:0'), covar=tensor([0.1938, 0.0632, 0.0998, 0.1235, 0.1701, 0.1298, 0.0348, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0240, 0.0258, 0.0215, 0.0228, 0.0195, 0.0214, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:07:37,404 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-30000.pt 2023-04-28 01:07:41,663 INFO [train.py:904] (0/8) Epoch 3, batch 9700, loss[loss=0.2018, simple_loss=0.2954, pruned_loss=0.05408, over 16823.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3055, pruned_loss=0.06827, over 3097124.82 frames. ], batch size: 96, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:07:44,427 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3283, 4.6147, 4.3655, 4.4357, 4.0270, 4.0135, 4.2443, 4.6026], device='cuda:0'), covar=tensor([0.0490, 0.0585, 0.0783, 0.0418, 0.0544, 0.0885, 0.0512, 0.0626], device='cuda:0'), in_proj_covar=tensor([0.0256, 0.0360, 0.0304, 0.0233, 0.0233, 0.0240, 0.0287, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:08:16,002 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30018.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:08:59,979 INFO [optim.py:368] (0/8) 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 01:09:24,300 INFO [train.py:904] (0/8) Epoch 3, batch 9750, loss[loss=0.2062, simple_loss=0.282, pruned_loss=0.06519, over 12503.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3039, pruned_loss=0.06825, over 3086500.91 frames. ], batch size: 248, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:10:58,804 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-28 01:11:02,941 INFO [train.py:904] (0/8) Epoch 3, batch 9800, loss[loss=0.2473, simple_loss=0.3395, pruned_loss=0.0775, over 16332.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3043, pruned_loss=0.06786, over 3085704.69 frames. ], batch size: 146, lr: 1.89e-02, grad_scale: 8.0 2023-04-28 01:11:30,661 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7407, 2.6528, 2.5882, 1.9284, 2.3972, 2.5487, 2.6263, 1.6418], device='cuda:0'), covar=tensor([0.0293, 0.0038, 0.0061, 0.0188, 0.0065, 0.0066, 0.0043, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0053, 0.0057, 0.0108, 0.0054, 0.0061, 0.0055, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 01:11:48,280 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:12:14,773 INFO [optim.py:368] (0/8) 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 01:12:47,471 INFO [train.py:904] (0/8) Epoch 3, batch 9850, loss[loss=0.225, simple_loss=0.3092, pruned_loss=0.07036, over 16785.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3057, pruned_loss=0.0678, over 3068869.87 frames. ], batch size: 124, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:12:48,739 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9831, 3.9558, 3.8208, 3.4345, 3.8668, 1.7748, 3.5922, 3.7613], device='cuda:0'), covar=tensor([0.0053, 0.0044, 0.0078, 0.0167, 0.0049, 0.1419, 0.0069, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0059, 0.0091, 0.0092, 0.0066, 0.0118, 0.0080, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:13:02,428 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:06,762 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:39,082 INFO [train.py:904] (0/8) Epoch 3, batch 9900, loss[loss=0.2232, simple_loss=0.3174, pruned_loss=0.06448, over 15323.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3057, pruned_loss=0.06776, over 3049106.52 frames. ], batch size: 191, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:14:54,219 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30207.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:14:59,655 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:15:09,237 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 01:15:36,682 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1090, 4.0447, 4.6263, 4.5712, 4.5466, 4.1615, 4.2311, 3.9031], device='cuda:0'), covar=tensor([0.0216, 0.0280, 0.0238, 0.0338, 0.0307, 0.0218, 0.0557, 0.0326], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0164, 0.0168, 0.0170, 0.0199, 0.0177, 0.0253, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-28 01:16:05,827 INFO [optim.py:368] (0/8) 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 01:16:35,571 INFO [train.py:904] (0/8) Epoch 3, batch 9950, loss[loss=0.2151, simple_loss=0.3036, pruned_loss=0.06327, over 16540.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3079, pruned_loss=0.06843, over 3045247.66 frames. ], batch size: 68, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:16:46,524 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30255.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:17:23,147 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:18:05,686 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2279, 3.1434, 2.7964, 1.9689, 2.5947, 2.1111, 2.8209, 3.0592], device='cuda:0'), covar=tensor([0.0222, 0.0375, 0.0437, 0.1246, 0.0561, 0.0779, 0.0545, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0106, 0.0150, 0.0142, 0.0131, 0.0126, 0.0133, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 01:18:37,070 INFO [train.py:904] (0/8) Epoch 3, batch 10000, loss[loss=0.211, simple_loss=0.3006, pruned_loss=0.06064, over 16912.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3062, pruned_loss=0.06751, over 3055316.63 frames. ], batch size: 116, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:19:11,813 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30318.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:19:55,686 INFO [optim.py:368] (0/8) 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 01:20:19,872 INFO [train.py:904] (0/8) Epoch 3, batch 10050, loss[loss=0.2642, simple_loss=0.3445, pruned_loss=0.09193, over 11943.00 frames. ], tot_loss[loss=0.22, simple_loss=0.3059, pruned_loss=0.06702, over 3055789.51 frames. ], batch size: 247, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:20:38,420 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:20:50,751 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30366.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:21:54,410 INFO [train.py:904] (0/8) Epoch 3, batch 10100, loss[loss=0.2263, simple_loss=0.3058, pruned_loss=0.0734, over 16937.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3073, pruned_loss=0.06821, over 3052503.82 frames. ], batch size: 109, lr: 1.88e-02, grad_scale: 8.0 2023-04-28 01:22:36,857 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:23:00,424 INFO [optim.py:368] (0/8) 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 01:23:15,094 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-3.pt 2023-04-28 01:23:38,636 INFO [train.py:904] (0/8) Epoch 4, batch 0, loss[loss=0.4207, simple_loss=0.4209, pruned_loss=0.2103, over 16882.00 frames. ], tot_loss[loss=0.4207, simple_loss=0.4209, pruned_loss=0.2103, over 16882.00 frames. ], batch size: 109, lr: 1.75e-02, grad_scale: 8.0 2023-04-28 01:23:38,637 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 01:23:46,516 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17747MB 2023-04-28 01:23:56,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1875, 3.8331, 3.9365, 2.7778, 3.7914, 3.7794, 3.8320, 2.1603], device='cuda:0'), covar=tensor([0.0335, 0.0016, 0.0035, 0.0206, 0.0032, 0.0045, 0.0024, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0052, 0.0055, 0.0109, 0.0054, 0.0060, 0.0055, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 01:23:57,937 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:29,173 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:24:50,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6706, 5.1132, 5.1990, 5.1312, 5.0889, 5.6641, 5.2637, 5.0214], device='cuda:0'), covar=tensor([0.0795, 0.1352, 0.1060, 0.1514, 0.2175, 0.0737, 0.0995, 0.1804], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0322, 0.0298, 0.0276, 0.0359, 0.0322, 0.0257, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:24:56,033 INFO [train.py:904] (0/8) Epoch 4, batch 50, loss[loss=0.2619, simple_loss=0.332, pruned_loss=0.09592, over 17067.00 frames. ], tot_loss[loss=0.2735, simple_loss=0.336, pruned_loss=0.1055, over 754407.68 frames. ], batch size: 50, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:25:02,248 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:25:07,097 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 01:25:49,883 INFO [optim.py:368] (0/8) 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 01:25:57,377 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.19 vs. limit=5.0 2023-04-28 01:26:02,171 INFO [train.py:904] (0/8) Epoch 4, batch 100, loss[loss=0.2541, simple_loss=0.3134, pruned_loss=0.09739, over 16826.00 frames. ], tot_loss[loss=0.2617, simple_loss=0.328, pruned_loss=0.09775, over 1324378.83 frames. ], batch size: 96, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:26:04,197 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 01:26:23,719 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:27:11,705 INFO [train.py:904] (0/8) Epoch 4, batch 150, loss[loss=0.261, simple_loss=0.3116, pruned_loss=0.1052, over 16675.00 frames. ], tot_loss[loss=0.2538, simple_loss=0.3217, pruned_loss=0.093, over 1771069.25 frames. ], batch size: 124, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:27:47,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9869, 5.7759, 5.7067, 5.6361, 5.7210, 6.1362, 5.8306, 5.6396], device='cuda:0'), covar=tensor([0.0669, 0.1067, 0.1052, 0.1426, 0.2098, 0.0677, 0.0729, 0.1724], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0350, 0.0323, 0.0299, 0.0392, 0.0347, 0.0271, 0.0395], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 01:28:04,896 INFO [optim.py:368] (0/8) 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 01:28:19,175 INFO [train.py:904] (0/8) Epoch 4, batch 200, loss[loss=0.2096, simple_loss=0.2947, pruned_loss=0.06222, over 17277.00 frames. ], tot_loss[loss=0.2525, simple_loss=0.3206, pruned_loss=0.09224, over 2107777.94 frames. ], batch size: 52, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:28:24,023 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30655.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:26,989 INFO [train.py:904] (0/8) Epoch 4, batch 250, loss[loss=0.21, simple_loss=0.2886, pruned_loss=0.06574, over 17097.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3157, pruned_loss=0.09023, over 2376121.50 frames. ], batch size: 47, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:29:48,577 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=30716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:29:48,749 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30716.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:30:21,643 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 4, batch 300, loss[loss=0.2576, simple_loss=0.3137, pruned_loss=0.1008, over 16936.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3138, pruned_loss=0.08894, over 2582529.43 frames. ], batch size: 109, lr: 1.75e-02, grad_scale: 1.0 2023-04-28 01:31:16,614 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30781.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:31:31,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4948, 5.9151, 5.6428, 5.6957, 5.0612, 4.6787, 5.2872, 5.9788], device='cuda:0'), covar=tensor([0.0581, 0.0607, 0.0671, 0.0403, 0.0521, 0.0597, 0.0549, 0.0591], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0424, 0.0364, 0.0272, 0.0274, 0.0275, 0.0335, 0.0294], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:31:43,561 INFO [train.py:904] (0/8) Epoch 4, batch 350, loss[loss=0.2641, simple_loss=0.314, pruned_loss=0.1071, over 16433.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3104, pruned_loss=0.08627, over 2755015.62 frames. ], batch size: 165, lr: 1.74e-02, grad_scale: 1.0 2023-04-28 01:32:20,670 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30829.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:32:23,845 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8438, 4.9087, 5.4943, 5.4960, 5.4225, 4.9247, 4.9040, 4.7406], device='cuda:0'), covar=tensor([0.0196, 0.0191, 0.0211, 0.0245, 0.0288, 0.0235, 0.0640, 0.0280], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0198, 0.0208, 0.0204, 0.0247, 0.0219, 0.0309, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 01:32:38,674 INFO [optim.py:368] (0/8) 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 01:32:51,072 INFO [train.py:904] (0/8) Epoch 4, batch 400, loss[loss=0.2669, simple_loss=0.3179, pruned_loss=0.1079, over 16830.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3076, pruned_loss=0.08498, over 2881338.72 frames. ], batch size: 102, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:33:11,824 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=30865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:01,540 INFO [train.py:904] (0/8) Epoch 4, batch 450, loss[loss=0.2427, simple_loss=0.3025, pruned_loss=0.09147, over 16266.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3052, pruned_loss=0.0828, over 2975022.28 frames. ], batch size: 165, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:34:05,070 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=30913.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:28,247 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 01:34:35,345 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:41,612 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 01:34:45,853 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=30934.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:34:56,474 INFO [optim.py:368] (0/8) 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 01:35:09,311 INFO [train.py:904] (0/8) Epoch 4, batch 500, loss[loss=0.2093, simple_loss=0.2946, pruned_loss=0.06197, over 17049.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.3027, pruned_loss=0.08098, over 3051866.18 frames. ], batch size: 53, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:35:28,268 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:35:45,651 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 01:35:58,375 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:09,065 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=30995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:17,833 INFO [train.py:904] (0/8) Epoch 4, batch 550, loss[loss=0.2015, simple_loss=0.276, pruned_loss=0.0635, over 17017.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3023, pruned_loss=0.08088, over 3108888.04 frames. ], batch size: 41, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:36:33,162 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31011.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:36:33,327 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31011.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:36:39,307 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:36:46,820 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3224, 4.6176, 2.3674, 4.8963, 3.1309, 4.9223, 2.4522, 3.5515], device='cuda:0'), covar=tensor([0.0067, 0.0179, 0.1382, 0.0022, 0.0610, 0.0189, 0.1185, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0142, 0.0173, 0.0082, 0.0160, 0.0169, 0.0180, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 01:37:13,485 INFO [optim.py:368] (0/8) 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 01:37:28,711 INFO [train.py:904] (0/8) Epoch 4, batch 600, loss[loss=0.2363, simple_loss=0.2914, pruned_loss=0.09055, over 16827.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3016, pruned_loss=0.08095, over 3144605.53 frames. ], batch size: 83, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:37:46,203 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:37:58,235 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31072.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:38:36,817 INFO [train.py:904] (0/8) Epoch 4, batch 650, loss[loss=0.2549, simple_loss=0.3103, pruned_loss=0.0997, over 16882.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3006, pruned_loss=0.08054, over 3185074.51 frames. ], batch size: 116, lr: 1.74e-02, grad_scale: 2.0 2023-04-28 01:39:30,540 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 3.618e+02 4.311e+02 5.742e+02 1.399e+03, threshold=8.622e+02, percent-clipped=6.0 2023-04-28 01:39:45,331 INFO [train.py:904] (0/8) Epoch 4, batch 700, loss[loss=0.2277, simple_loss=0.2975, pruned_loss=0.07892, over 16711.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.2995, pruned_loss=0.07905, over 3216179.55 frames. ], batch size: 83, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:40:53,747 INFO [train.py:904] (0/8) Epoch 4, batch 750, loss[loss=0.2279, simple_loss=0.3115, pruned_loss=0.0721, over 17148.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.2997, pruned_loss=0.07865, over 3238126.42 frames. ], batch size: 49, lr: 1.73e-02, grad_scale: 2.0 2023-04-28 01:41:48,309 INFO [optim.py:368] (0/8) 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 01:42:01,633 INFO [train.py:904] (0/8) Epoch 4, batch 800, loss[loss=0.2573, simple_loss=0.3134, pruned_loss=0.1006, over 16899.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3001, pruned_loss=0.07943, over 3251652.53 frames. ], batch size: 116, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:42:14,290 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:42:44,502 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:43:11,603 INFO [train.py:904] (0/8) Epoch 4, batch 850, loss[loss=0.1949, simple_loss=0.2851, pruned_loss=0.05238, over 17146.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3, pruned_loss=0.07885, over 3272127.79 frames. ], batch size: 48, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:43:24,689 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31311.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:07,365 INFO [optim.py:368] (0/8) 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 01:44:19,711 INFO [train.py:904] (0/8) Epoch 4, batch 900, loss[loss=0.2287, simple_loss=0.2911, pruned_loss=0.08311, over 16857.00 frames. ], tot_loss[loss=0.227, simple_loss=0.2983, pruned_loss=0.07787, over 3285832.89 frames. ], batch size: 83, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:44:32,188 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:44:42,745 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31367.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:45:31,329 INFO [train.py:904] (0/8) Epoch 4, batch 950, loss[loss=0.261, simple_loss=0.3178, pruned_loss=0.1021, over 12210.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.2985, pruned_loss=0.0779, over 3283454.67 frames. ], batch size: 246, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:45:41,855 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31409.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:46:26,069 INFO [optim.py:368] (0/8) 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 01:46:38,683 INFO [train.py:904] (0/8) Epoch 4, batch 1000, loss[loss=0.1975, simple_loss=0.2692, pruned_loss=0.06295, over 16315.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2971, pruned_loss=0.07758, over 3291475.97 frames. ], batch size: 36, lr: 1.73e-02, grad_scale: 4.0 2023-04-28 01:46:58,801 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2534, 1.9069, 1.4933, 1.6714, 2.1270, 2.0479, 2.0575, 2.3165], device='cuda:0'), covar=tensor([0.0042, 0.0129, 0.0163, 0.0161, 0.0071, 0.0126, 0.0081, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0135, 0.0133, 0.0132, 0.0125, 0.0135, 0.0097, 0.0113], device='cuda:0'), out_proj_covar=tensor([8.7690e-05, 1.8532e-04, 1.7520e-04, 1.7713e-04, 1.7276e-04, 1.8523e-04, 1.3073e-04, 1.5572e-04], device='cuda:0') 2023-04-28 01:47:05,416 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 01:47:06,292 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:47:34,931 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6616, 4.9975, 4.6035, 4.7837, 4.3297, 4.3336, 4.5521, 5.0303], device='cuda:0'), covar=tensor([0.0562, 0.0655, 0.0908, 0.0429, 0.0649, 0.0739, 0.0633, 0.0715], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0441, 0.0373, 0.0274, 0.0281, 0.0279, 0.0345, 0.0305], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:47:36,213 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31491.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:47:36,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2591, 2.1908, 1.8403, 2.0928, 2.7836, 2.4862, 3.6604, 3.0781], device='cuda:0'), covar=tensor([0.0030, 0.0152, 0.0185, 0.0163, 0.0095, 0.0141, 0.0045, 0.0077], device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0135, 0.0134, 0.0132, 0.0126, 0.0135, 0.0097, 0.0113], device='cuda:0'), out_proj_covar=tensor([8.8749e-05, 1.8516e-04, 1.7622e-04, 1.7646e-04, 1.7425e-04, 1.8503e-04, 1.3092e-04, 1.5531e-04], device='cuda:0') 2023-04-28 01:47:48,998 INFO [train.py:904] (0/8) Epoch 4, batch 1050, loss[loss=0.2355, simple_loss=0.2936, pruned_loss=0.08868, over 16858.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.2961, pruned_loss=0.07674, over 3275327.41 frames. ], batch size: 109, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:48:45,342 INFO [optim.py:368] (0/8) 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 01:49:00,852 INFO [train.py:904] (0/8) Epoch 4, batch 1100, loss[loss=0.2595, simple_loss=0.3436, pruned_loss=0.08773, over 17258.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.2949, pruned_loss=0.07614, over 3269532.47 frames. ], batch size: 52, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:49:02,431 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31552.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:49:11,027 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31559.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:41,921 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:49:49,850 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2361, 4.1389, 4.1406, 3.6813, 4.1012, 1.8429, 3.9235, 4.0213], device='cuda:0'), covar=tensor([0.0058, 0.0052, 0.0075, 0.0237, 0.0059, 0.1341, 0.0075, 0.0106], device='cuda:0'), in_proj_covar=tensor([0.0084, 0.0075, 0.0113, 0.0123, 0.0083, 0.0130, 0.0100, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 01:49:55,705 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:09,616 INFO [train.py:904] (0/8) Epoch 4, batch 1150, loss[loss=0.2277, simple_loss=0.2926, pruned_loss=0.08142, over 16423.00 frames. ], tot_loss[loss=0.2239, simple_loss=0.2953, pruned_loss=0.07626, over 3282383.12 frames. ], batch size: 146, lr: 1.72e-02, grad_scale: 4.0 2023-04-28 01:50:18,983 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:32,112 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 01:50:48,423 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:50:56,618 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 01:51:00,558 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31638.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:51:04,259 INFO [optim.py:368] (0/8) 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 01:51:19,075 INFO [train.py:904] (0/8) Epoch 4, batch 1200, loss[loss=0.2217, simple_loss=0.3006, pruned_loss=0.07138, over 17063.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2946, pruned_loss=0.07571, over 3288131.40 frames. ], batch size: 55, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:51:41,018 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=31667.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 01:51:43,175 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 01:52:27,396 INFO [train.py:904] (0/8) Epoch 4, batch 1250, loss[loss=0.2535, simple_loss=0.3103, pruned_loss=0.09841, over 16442.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2952, pruned_loss=0.07621, over 3292435.23 frames. ], batch size: 146, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:52:31,886 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31704.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:52:47,964 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=31715.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 01:53:22,525 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.448e+02 3.415e+02 4.132e+02 5.466e+02 7.512e+02, threshold=8.265e+02, percent-clipped=0.0 2023-04-28 01:53:35,331 INFO [train.py:904] (0/8) Epoch 4, batch 1300, loss[loss=0.1844, simple_loss=0.2569, pruned_loss=0.05599, over 16997.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.2937, pruned_loss=0.0757, over 3299280.88 frames. ], batch size: 41, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:53:54,664 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:53:54,888 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:26,715 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:54:44,574 INFO [train.py:904] (0/8) Epoch 4, batch 1350, loss[loss=0.2414, simple_loss=0.2945, pruned_loss=0.09416, over 16816.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2944, pruned_loss=0.07575, over 3299452.69 frames. ], batch size: 90, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:03,486 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 01:55:14,587 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:39,668 INFO [optim.py:368] (0/8) 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 01:55:47,531 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=31847.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 01:55:50,781 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:55:52,772 INFO [train.py:904] (0/8) Epoch 4, batch 1400, loss[loss=0.2269, simple_loss=0.3079, pruned_loss=0.07294, over 17054.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.2944, pruned_loss=0.07564, over 3311462.41 frames. ], batch size: 53, lr: 1.72e-02, grad_scale: 8.0 2023-04-28 01:55:54,933 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3791, 4.0695, 3.3307, 1.9055, 2.7809, 2.4611, 3.6090, 3.9088], device='cuda:0'), covar=tensor([0.0223, 0.0379, 0.0564, 0.1490, 0.0711, 0.0878, 0.0538, 0.0565], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0127, 0.0152, 0.0142, 0.0133, 0.0127, 0.0140, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 01:55:59,383 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2283, 5.1422, 5.0249, 3.6801, 5.0867, 1.9119, 4.7581, 5.0051], device='cuda:0'), covar=tensor([0.0084, 0.0073, 0.0098, 0.0519, 0.0073, 0.1835, 0.0101, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0077, 0.0115, 0.0127, 0.0085, 0.0132, 0.0102, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 01:56:32,112 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31878.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:40,541 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:56:46,985 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 01:57:02,753 INFO [train.py:904] (0/8) Epoch 4, batch 1450, loss[loss=0.2435, simple_loss=0.2969, pruned_loss=0.09504, over 16755.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2932, pruned_loss=0.07463, over 3309729.24 frames. ], batch size: 89, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:57:22,417 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:57:23,552 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4157, 4.7591, 4.4491, 4.5651, 4.2043, 4.1720, 4.2983, 4.7088], device='cuda:0'), covar=tensor([0.0581, 0.0672, 0.0877, 0.0405, 0.0583, 0.0951, 0.0611, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0451, 0.0389, 0.0282, 0.0289, 0.0286, 0.0357, 0.0312], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 01:57:56,598 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:00,354 INFO [optim.py:368] (0/8) 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 01:58:12,695 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=31950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:58:13,465 INFO [train.py:904] (0/8) Epoch 4, batch 1500, loss[loss=0.1924, simple_loss=0.28, pruned_loss=0.05246, over 17172.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.2929, pruned_loss=0.07441, over 3314024.20 frames. ], batch size: 46, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:58:47,444 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=31976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 01:59:22,209 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-32000.pt 2023-04-28 01:59:26,629 INFO [train.py:904] (0/8) Epoch 4, batch 1550, loss[loss=0.2355, simple_loss=0.2951, pruned_loss=0.08796, over 16882.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2961, pruned_loss=0.07642, over 3308993.11 frames. ], batch size: 96, lr: 1.71e-02, grad_scale: 4.0 2023-04-28 01:59:40,625 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=32011.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:07,373 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 02:00:21,780 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.322e+02 3.491e+02 4.515e+02 5.293e+02 8.780e+02, threshold=9.030e+02, percent-clipped=2.0 2023-04-28 02:00:34,388 INFO [train.py:904] (0/8) Epoch 4, batch 1600, loss[loss=0.2006, simple_loss=0.2811, pruned_loss=0.06001, over 15921.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.2981, pruned_loss=0.07707, over 3308783.02 frames. ], batch size: 35, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:00:46,072 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32060.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:00:52,688 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:01:41,700 INFO [train.py:904] (0/8) Epoch 4, batch 1650, loss[loss=0.1906, simple_loss=0.2738, pruned_loss=0.05374, over 17217.00 frames. ], tot_loss[loss=0.2277, simple_loss=0.2996, pruned_loss=0.07789, over 3310463.45 frames. ], batch size: 45, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:01:58,773 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32113.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:02,791 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-28 02:02:23,175 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3351, 2.1911, 2.4786, 2.9609, 3.0144, 3.5841, 2.5078, 3.4385], device='cuda:0'), covar=tensor([0.0044, 0.0157, 0.0119, 0.0100, 0.0075, 0.0043, 0.0139, 0.0042], device='cuda:0'), in_proj_covar=tensor([0.0104, 0.0131, 0.0118, 0.0115, 0.0109, 0.0083, 0.0126, 0.0072], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 02:02:37,466 INFO [optim.py:368] (0/8) 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 02:02:40,718 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:02:45,015 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32147.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 02:02:50,066 INFO [train.py:904] (0/8) Epoch 4, batch 1700, loss[loss=0.2235, simple_loss=0.3035, pruned_loss=0.07179, over 17183.00 frames. ], tot_loss[loss=0.2282, simple_loss=0.301, pruned_loss=0.07772, over 3320136.05 frames. ], batch size: 46, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:03:06,295 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7015, 4.6230, 4.0361, 2.0836, 3.0338, 2.6662, 4.0324, 4.3154], device='cuda:0'), covar=tensor([0.0214, 0.0386, 0.0360, 0.1422, 0.0678, 0.0871, 0.0572, 0.0641], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0129, 0.0152, 0.0142, 0.0135, 0.0128, 0.0139, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 02:03:29,723 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:03:53,203 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32195.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 02:04:01,132 INFO [train.py:904] (0/8) Epoch 4, batch 1750, loss[loss=0.1987, simple_loss=0.2824, pruned_loss=0.05749, over 15994.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3023, pruned_loss=0.0781, over 3306517.30 frames. ], batch size: 35, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:04:42,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9231, 3.5158, 2.9649, 5.1115, 4.8985, 4.3618, 2.3007, 3.1405], device='cuda:0'), covar=tensor([0.1254, 0.0454, 0.0996, 0.0063, 0.0233, 0.0319, 0.1078, 0.0704], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0136, 0.0161, 0.0075, 0.0167, 0.0155, 0.0152, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 02:04:47,682 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:04:58,765 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.044e+02 3.415e+02 4.164e+02 5.127e+02 8.958e+02, threshold=8.329e+02, percent-clipped=1.0 2023-04-28 02:05:11,466 INFO [train.py:904] (0/8) Epoch 4, batch 1800, loss[loss=0.2072, simple_loss=0.2878, pruned_loss=0.06332, over 17212.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.3043, pruned_loss=0.07892, over 3296860.90 frames. ], batch size: 46, lr: 1.71e-02, grad_scale: 8.0 2023-04-28 02:05:39,585 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32271.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:18,945 INFO [train.py:904] (0/8) Epoch 4, batch 1850, loss[loss=0.2208, simple_loss=0.2903, pruned_loss=0.07566, over 16505.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.305, pruned_loss=0.07876, over 3297824.27 frames. ], batch size: 146, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:06:26,554 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=32306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:06:26,791 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7467, 2.5192, 1.9682, 2.4725, 2.9708, 2.8702, 3.9596, 3.3812], device='cuda:0'), covar=tensor([0.0025, 0.0158, 0.0206, 0.0171, 0.0099, 0.0144, 0.0063, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0138, 0.0137, 0.0134, 0.0130, 0.0139, 0.0103, 0.0117], device='cuda:0'), out_proj_covar=tensor([9.6164e-05, 1.8620e-04, 1.7907e-04, 1.7791e-04, 1.7764e-04, 1.8930e-04, 1.3927e-04, 1.6004e-04], device='cuda:0') 2023-04-28 02:07:18,153 INFO [optim.py:368] (0/8) 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 02:07:29,365 INFO [train.py:904] (0/8) Epoch 4, batch 1900, loss[loss=0.2597, simple_loss=0.3255, pruned_loss=0.09695, over 16311.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.304, pruned_loss=0.07833, over 3300462.14 frames. ], batch size: 165, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:07:43,138 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:08:41,350 INFO [train.py:904] (0/8) Epoch 4, batch 1950, loss[loss=0.1979, simple_loss=0.2775, pruned_loss=0.05913, over 16802.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3037, pruned_loss=0.07732, over 3304818.96 frames. ], batch size: 42, lr: 1.70e-02, grad_scale: 4.0 2023-04-28 02:08:45,492 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9172, 5.2582, 4.8881, 5.0041, 4.6208, 4.4910, 4.7692, 5.2739], device='cuda:0'), covar=tensor([0.0566, 0.0542, 0.0848, 0.0404, 0.0597, 0.0741, 0.0583, 0.0618], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0453, 0.0387, 0.0286, 0.0289, 0.0287, 0.0358, 0.0316], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:08:50,797 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:37,982 INFO [optim.py:368] (0/8) 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 02:09:39,312 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:09:47,994 INFO [train.py:904] (0/8) Epoch 4, batch 2000, loss[loss=0.2114, simple_loss=0.2939, pruned_loss=0.06446, over 17192.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3035, pruned_loss=0.07746, over 3302532.98 frames. ], batch size: 46, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:10:00,750 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9815, 4.9976, 4.7953, 4.2654, 4.7732, 1.9617, 4.6315, 4.9488], device='cuda:0'), covar=tensor([0.0060, 0.0058, 0.0079, 0.0325, 0.0071, 0.1373, 0.0093, 0.0105], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0078, 0.0116, 0.0126, 0.0085, 0.0128, 0.0104, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:10:09,588 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9945, 3.7425, 3.0645, 1.8849, 2.5873, 2.0553, 3.5101, 3.5446], device='cuda:0'), covar=tensor([0.0223, 0.0400, 0.0521, 0.1370, 0.0715, 0.0947, 0.0454, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0132, 0.0155, 0.0142, 0.0136, 0.0127, 0.0141, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 02:10:28,150 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:10:58,288 INFO [train.py:904] (0/8) Epoch 4, batch 2050, loss[loss=0.2049, simple_loss=0.2839, pruned_loss=0.06293, over 17236.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.303, pruned_loss=0.07759, over 3315510.79 frames. ], batch size: 45, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:11:34,839 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:44,340 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32534.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:11:46,880 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5233, 4.4200, 4.0893, 1.8141, 3.2161, 2.2311, 4.0027, 4.0534], device='cuda:0'), covar=tensor([0.0237, 0.0438, 0.0384, 0.1619, 0.0674, 0.1007, 0.0543, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0131, 0.0155, 0.0143, 0.0135, 0.0128, 0.0140, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 02:11:57,880 INFO [optim.py:368] (0/8) 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 02:12:07,696 INFO [train.py:904] (0/8) Epoch 4, batch 2100, loss[loss=0.2423, simple_loss=0.3218, pruned_loss=0.08136, over 16638.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3065, pruned_loss=0.08044, over 3306847.76 frames. ], batch size: 57, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:12:36,408 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32571.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:51,002 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32582.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:12:59,087 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.95 vs. limit=5.0 2023-04-28 02:13:16,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2339, 4.9853, 5.1846, 5.5416, 5.6232, 4.7422, 5.5181, 5.5664], device='cuda:0'), covar=tensor([0.0582, 0.0571, 0.1046, 0.0309, 0.0287, 0.0409, 0.0311, 0.0228], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0431, 0.0581, 0.0452, 0.0333, 0.0327, 0.0351, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:13:17,664 INFO [train.py:904] (0/8) Epoch 4, batch 2150, loss[loss=0.1954, simple_loss=0.274, pruned_loss=0.05839, over 16769.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.307, pruned_loss=0.08064, over 3306080.33 frames. ], batch size: 39, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:13:24,598 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=32606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:13:28,673 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1311, 3.8747, 3.6890, 4.2990, 4.3185, 3.9099, 4.0271, 4.3271], device='cuda:0'), covar=tensor([0.0653, 0.0753, 0.1878, 0.0610, 0.0658, 0.1086, 0.1296, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0433, 0.0584, 0.0455, 0.0336, 0.0330, 0.0353, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:13:42,054 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32619.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:14:15,065 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.513e+02 3.522e+02 4.278e+02 5.461e+02 9.146e+02, threshold=8.556e+02, percent-clipped=3.0 2023-04-28 02:14:26,823 INFO [train.py:904] (0/8) Epoch 4, batch 2200, loss[loss=0.283, simple_loss=0.3336, pruned_loss=0.1162, over 16644.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3081, pruned_loss=0.08172, over 3302421.05 frames. ], batch size: 134, lr: 1.70e-02, grad_scale: 8.0 2023-04-28 02:14:30,432 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=32654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:15:21,395 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1085, 5.4221, 5.0699, 5.2946, 4.8132, 4.5596, 4.9789, 5.5157], device='cuda:0'), covar=tensor([0.0524, 0.0623, 0.0839, 0.0337, 0.0541, 0.0673, 0.0546, 0.0631], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0456, 0.0388, 0.0287, 0.0291, 0.0287, 0.0356, 0.0315], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:15:24,494 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9160, 4.9700, 5.5762, 5.5314, 5.5038, 4.9592, 5.0437, 4.7302], device='cuda:0'), covar=tensor([0.0236, 0.0279, 0.0232, 0.0308, 0.0332, 0.0280, 0.0764, 0.0372], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0219, 0.0226, 0.0227, 0.0271, 0.0241, 0.0345, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 02:15:34,040 INFO [train.py:904] (0/8) Epoch 4, batch 2250, loss[loss=0.2478, simple_loss=0.3154, pruned_loss=0.09007, over 16233.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3082, pruned_loss=0.08122, over 3312171.02 frames. ], batch size: 165, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:16:32,022 INFO [optim.py:368] (0/8) 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 02:16:42,224 INFO [train.py:904] (0/8) Epoch 4, batch 2300, loss[loss=0.244, simple_loss=0.3095, pruned_loss=0.08928, over 16725.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3061, pruned_loss=0.07932, over 3317820.11 frames. ], batch size: 134, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:17:12,312 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9868, 5.0250, 5.6516, 5.5972, 5.5322, 5.1231, 5.1788, 4.9401], device='cuda:0'), covar=tensor([0.0211, 0.0255, 0.0207, 0.0293, 0.0342, 0.0236, 0.0604, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0219, 0.0226, 0.0227, 0.0273, 0.0241, 0.0344, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 02:17:31,192 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6333, 2.5559, 2.2580, 2.2428, 2.9928, 2.8322, 3.9231, 3.3082], device='cuda:0'), covar=tensor([0.0028, 0.0145, 0.0169, 0.0184, 0.0085, 0.0135, 0.0045, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0077, 0.0137, 0.0136, 0.0136, 0.0131, 0.0140, 0.0104, 0.0119], device='cuda:0'), out_proj_covar=tensor([9.9414e-05, 1.8452e-04, 1.7744e-04, 1.8046e-04, 1.7784e-04, 1.8985e-04, 1.3939e-04, 1.6211e-04], device='cuda:0') 2023-04-28 02:17:51,380 INFO [train.py:904] (0/8) Epoch 4, batch 2350, loss[loss=0.2958, simple_loss=0.3483, pruned_loss=0.1217, over 16277.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3061, pruned_loss=0.07984, over 3309559.16 frames. ], batch size: 165, lr: 1.69e-02, grad_scale: 4.0 2023-04-28 02:18:39,259 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 02:18:48,986 INFO [optim.py:368] (0/8) 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 02:18:57,205 INFO [train.py:904] (0/8) Epoch 4, batch 2400, loss[loss=0.2571, simple_loss=0.3274, pruned_loss=0.09344, over 16456.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3074, pruned_loss=0.07976, over 3296276.15 frames. ], batch size: 75, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:19:26,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7565, 2.5257, 2.5714, 1.9769, 2.6008, 2.6703, 2.7047, 1.7006], device='cuda:0'), covar=tensor([0.0254, 0.0058, 0.0035, 0.0177, 0.0044, 0.0044, 0.0032, 0.0237], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0057, 0.0058, 0.0108, 0.0059, 0.0065, 0.0059, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:20:06,099 INFO [train.py:904] (0/8) Epoch 4, batch 2450, loss[loss=0.2077, simple_loss=0.2941, pruned_loss=0.06065, over 17078.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3081, pruned_loss=0.07963, over 3300802.97 frames. ], batch size: 47, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:03,695 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.564e+02 3.644e+02 4.364e+02 5.421e+02 9.049e+02, threshold=8.728e+02, percent-clipped=5.0 2023-04-28 02:21:13,298 INFO [train.py:904] (0/8) Epoch 4, batch 2500, loss[loss=0.2716, simple_loss=0.3549, pruned_loss=0.09418, over 17096.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3077, pruned_loss=0.07872, over 3307694.65 frames. ], batch size: 53, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:21:20,926 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4008, 2.8552, 2.6643, 4.9675, 2.1364, 4.3480, 2.6659, 2.8173], device='cuda:0'), covar=tensor([0.0345, 0.1161, 0.0639, 0.0168, 0.2385, 0.0382, 0.1291, 0.1966], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0272, 0.0224, 0.0285, 0.0336, 0.0254, 0.0251, 0.0340], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:21:34,465 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=32966.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:22:21,148 INFO [train.py:904] (0/8) Epoch 4, batch 2550, loss[loss=0.2718, simple_loss=0.3187, pruned_loss=0.1125, over 16923.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3078, pruned_loss=0.07919, over 3297588.03 frames. ], batch size: 109, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:22:53,415 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3406, 2.9895, 2.9039, 4.8676, 2.1804, 4.1863, 2.5994, 2.6740], device='cuda:0'), covar=tensor([0.0351, 0.1031, 0.0593, 0.0191, 0.2234, 0.0418, 0.1238, 0.1930], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0274, 0.0226, 0.0288, 0.0337, 0.0256, 0.0252, 0.0340], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:22:57,672 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33027.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:23:20,060 INFO [optim.py:368] (0/8) 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 02:23:30,982 INFO [train.py:904] (0/8) Epoch 4, batch 2600, loss[loss=0.2621, simple_loss=0.3269, pruned_loss=0.09869, over 16749.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3074, pruned_loss=0.07882, over 3304329.59 frames. ], batch size: 124, lr: 1.69e-02, grad_scale: 8.0 2023-04-28 02:23:40,656 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0633, 4.3624, 2.1760, 4.7298, 2.6740, 4.7034, 2.2128, 3.0828], device='cuda:0'), covar=tensor([0.0112, 0.0202, 0.1280, 0.0022, 0.0776, 0.0219, 0.1412, 0.0602], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0148, 0.0170, 0.0084, 0.0158, 0.0179, 0.0182, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 02:23:55,328 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33069.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:24:13,147 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 02:24:31,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1306, 4.5620, 4.4859, 1.8525, 4.7556, 4.8185, 3.3110, 3.4858], device='cuda:0'), covar=tensor([0.0621, 0.0138, 0.0196, 0.1262, 0.0059, 0.0044, 0.0307, 0.0368], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0085, 0.0083, 0.0145, 0.0075, 0.0078, 0.0116, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 02:24:39,074 INFO [train.py:904] (0/8) Epoch 4, batch 2650, loss[loss=0.2616, simple_loss=0.326, pruned_loss=0.09858, over 16599.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3073, pruned_loss=0.07831, over 3306829.69 frames. ], batch size: 68, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:19,548 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:25:34,603 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5591, 3.3244, 2.6219, 2.2073, 2.5662, 2.0052, 3.4311, 3.6021], device='cuda:0'), covar=tensor([0.1772, 0.0632, 0.1128, 0.1376, 0.1845, 0.1378, 0.0420, 0.0540], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0252, 0.0262, 0.0231, 0.0302, 0.0196, 0.0232, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:25:38,562 INFO [optim.py:368] (0/8) 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 02:25:46,697 INFO [train.py:904] (0/8) Epoch 4, batch 2700, loss[loss=0.2171, simple_loss=0.3091, pruned_loss=0.06253, over 17134.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.3069, pruned_loss=0.07691, over 3314111.00 frames. ], batch size: 48, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:25:54,256 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33155.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:26:56,822 INFO [train.py:904] (0/8) Epoch 4, batch 2750, loss[loss=0.24, simple_loss=0.3291, pruned_loss=0.07538, over 17118.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3072, pruned_loss=0.0773, over 3314458.47 frames. ], batch size: 48, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:27:16,351 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33216.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:27:54,083 INFO [optim.py:368] (0/8) 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 02:28:04,366 INFO [train.py:904] (0/8) Epoch 4, batch 2800, loss[loss=0.2087, simple_loss=0.2818, pruned_loss=0.06783, over 16764.00 frames. ], tot_loss[loss=0.2296, simple_loss=0.3063, pruned_loss=0.07638, over 3315657.35 frames. ], batch size: 39, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:28:05,988 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0474, 4.1282, 4.4872, 4.5125, 4.4849, 4.1225, 4.1436, 4.0806], device='cuda:0'), covar=tensor([0.0284, 0.0289, 0.0279, 0.0344, 0.0369, 0.0294, 0.0677, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0219, 0.0225, 0.0229, 0.0277, 0.0236, 0.0344, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 02:28:09,927 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 02:28:51,544 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 02:29:06,731 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-04-28 02:29:07,437 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7478, 4.9507, 4.9709, 5.0371, 4.8847, 5.4984, 5.1507, 4.8572], device='cuda:0'), covar=tensor([0.0924, 0.1131, 0.1289, 0.1325, 0.2274, 0.0743, 0.0931, 0.1903], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0373, 0.0351, 0.0315, 0.0422, 0.0373, 0.0295, 0.0425], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:29:10,617 INFO [train.py:904] (0/8) Epoch 4, batch 2850, loss[loss=0.2424, simple_loss=0.3244, pruned_loss=0.08015, over 17046.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3066, pruned_loss=0.07696, over 3309762.72 frames. ], batch size: 53, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:29:12,824 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 02:29:24,490 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0287, 4.8639, 4.8532, 4.1722, 4.7966, 2.0516, 4.6076, 4.9107], device='cuda:0'), covar=tensor([0.0046, 0.0044, 0.0064, 0.0278, 0.0053, 0.1244, 0.0073, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0078, 0.0118, 0.0127, 0.0087, 0.0127, 0.0104, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:29:39,604 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:29:59,803 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 02:30:09,814 INFO [optim.py:368] (0/8) 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 02:30:20,486 INFO [train.py:904] (0/8) Epoch 4, batch 2900, loss[loss=0.2264, simple_loss=0.3076, pruned_loss=0.0726, over 17065.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3062, pruned_loss=0.07809, over 3301555.57 frames. ], batch size: 55, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:30:46,051 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 02:31:24,859 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2826, 4.0166, 4.2981, 4.5653, 4.6160, 4.1568, 4.4155, 4.5383], device='cuda:0'), covar=tensor([0.0729, 0.0714, 0.1182, 0.0456, 0.0432, 0.0880, 0.0894, 0.0445], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0427, 0.0576, 0.0455, 0.0338, 0.0323, 0.0352, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:31:29,014 INFO [train.py:904] (0/8) Epoch 4, batch 2950, loss[loss=0.2136, simple_loss=0.2847, pruned_loss=0.07124, over 15903.00 frames. ], tot_loss[loss=0.232, simple_loss=0.3056, pruned_loss=0.07921, over 3293901.97 frames. ], batch size: 35, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:01,337 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:32:14,845 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9791, 5.5197, 5.5402, 5.6153, 5.4809, 5.9978, 5.7850, 5.5989], device='cuda:0'), covar=tensor([0.0765, 0.1441, 0.1314, 0.1811, 0.2641, 0.0935, 0.0967, 0.2271], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0373, 0.0352, 0.0319, 0.0422, 0.0375, 0.0297, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:32:27,163 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 4, batch 3000, loss[loss=0.196, simple_loss=0.282, pruned_loss=0.05506, over 17216.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3052, pruned_loss=0.07897, over 3309961.72 frames. ], batch size: 46, lr: 1.68e-02, grad_scale: 8.0 2023-04-28 02:32:35,895 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 02:32:45,552 INFO [train.py:938] (0/8) Epoch 4, validation: loss=0.1627, simple_loss=0.2694, pruned_loss=0.02796, over 944034.00 frames. 2023-04-28 02:32:45,552 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-28 02:32:47,230 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1424, 5.0485, 4.9353, 4.3285, 4.9676, 2.0672, 4.7255, 5.1329], device='cuda:0'), covar=tensor([0.0050, 0.0045, 0.0071, 0.0317, 0.0048, 0.1246, 0.0080, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0079, 0.0118, 0.0127, 0.0088, 0.0126, 0.0104, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:33:12,877 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:33:41,936 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8285, 4.2001, 3.3379, 2.4650, 3.0502, 2.3504, 4.4515, 4.4596], device='cuda:0'), covar=tensor([0.2205, 0.0581, 0.1078, 0.1305, 0.2265, 0.1368, 0.0305, 0.0437], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0251, 0.0259, 0.0228, 0.0301, 0.0195, 0.0227, 0.0246], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:33:54,602 INFO [train.py:904] (0/8) Epoch 4, batch 3050, loss[loss=0.1942, simple_loss=0.2734, pruned_loss=0.05751, over 16829.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3037, pruned_loss=0.07768, over 3322141.22 frames. ], batch size: 42, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:34:07,861 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33511.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:26,483 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7245, 2.6981, 2.5531, 4.3925, 2.0151, 3.8581, 2.4382, 2.3101], device='cuda:0'), covar=tensor([0.0431, 0.1154, 0.0641, 0.0188, 0.2222, 0.0429, 0.1248, 0.2058], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0272, 0.0227, 0.0289, 0.0337, 0.0257, 0.0253, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:34:38,024 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33533.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:34:54,166 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.952e+02 3.585e+02 4.279e+02 5.215e+02 1.682e+03, threshold=8.559e+02, percent-clipped=1.0 2023-04-28 02:35:02,725 INFO [train.py:904] (0/8) Epoch 4, batch 3100, loss[loss=0.1912, simple_loss=0.2687, pruned_loss=0.05679, over 15879.00 frames. ], tot_loss[loss=0.229, simple_loss=0.303, pruned_loss=0.07752, over 3317637.67 frames. ], batch size: 35, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:35:16,601 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2724, 4.0259, 4.2026, 4.5150, 4.5990, 4.2052, 4.3230, 4.5774], device='cuda:0'), covar=tensor([0.0760, 0.0681, 0.1264, 0.0545, 0.0474, 0.0683, 0.1388, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0430, 0.0578, 0.0454, 0.0338, 0.0318, 0.0352, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:36:10,691 INFO [train.py:904] (0/8) Epoch 4, batch 3150, loss[loss=0.2425, simple_loss=0.323, pruned_loss=0.08104, over 17145.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3023, pruned_loss=0.07788, over 3319862.36 frames. ], batch size: 49, lr: 1.67e-02, grad_scale: 4.0 2023-04-28 02:36:39,495 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:11,330 INFO [optim.py:368] (0/8) 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 02:37:18,479 INFO [train.py:904] (0/8) Epoch 4, batch 3200, loss[loss=0.2367, simple_loss=0.3049, pruned_loss=0.08422, over 16873.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3013, pruned_loss=0.07681, over 3318084.66 frames. ], batch size: 116, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:37:44,855 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:37:53,909 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:38:03,897 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8187, 3.8683, 2.8998, 2.4409, 2.9007, 2.0432, 3.7781, 4.0321], device='cuda:0'), covar=tensor([0.1654, 0.0438, 0.1136, 0.1267, 0.1972, 0.1552, 0.0408, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0248, 0.0259, 0.0230, 0.0304, 0.0194, 0.0225, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:38:25,340 INFO [train.py:904] (0/8) Epoch 4, batch 3250, loss[loss=0.212, simple_loss=0.298, pruned_loss=0.06299, over 17157.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3012, pruned_loss=0.0773, over 3313167.90 frames. ], batch size: 46, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:38:58,313 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33725.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:39:16,407 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=33738.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:39:26,583 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.718e+02 3.624e+02 4.331e+02 5.289e+02 1.384e+03, threshold=8.662e+02, percent-clipped=5.0 2023-04-28 02:39:36,477 INFO [train.py:904] (0/8) Epoch 4, batch 3300, loss[loss=0.1911, simple_loss=0.2812, pruned_loss=0.05051, over 17222.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3021, pruned_loss=0.07772, over 3318301.87 frames. ], batch size: 45, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:07,412 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:40:38,438 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8213, 3.1668, 2.3131, 4.4319, 4.1760, 3.9942, 1.5465, 2.8781], device='cuda:0'), covar=tensor([0.1186, 0.0406, 0.1125, 0.0064, 0.0203, 0.0326, 0.1152, 0.0710], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0138, 0.0163, 0.0080, 0.0170, 0.0158, 0.0153, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 02:40:45,519 INFO [train.py:904] (0/8) Epoch 4, batch 3350, loss[loss=0.2033, simple_loss=0.2775, pruned_loss=0.06452, over 16957.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3022, pruned_loss=0.07735, over 3311053.41 frames. ], batch size: 41, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:40:58,968 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=33811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:21,442 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=33828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:41:44,268 INFO [optim.py:368] (0/8) 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 02:41:52,912 INFO [train.py:904] (0/8) Epoch 4, batch 3400, loss[loss=0.2273, simple_loss=0.292, pruned_loss=0.08134, over 16838.00 frames. ], tot_loss[loss=0.2272, simple_loss=0.3016, pruned_loss=0.07636, over 3319117.99 frames. ], batch size: 102, lr: 1.67e-02, grad_scale: 8.0 2023-04-28 02:42:04,789 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=33859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:43:03,529 INFO [train.py:904] (0/8) Epoch 4, batch 3450, loss[loss=0.1913, simple_loss=0.2776, pruned_loss=0.05245, over 17159.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3002, pruned_loss=0.07568, over 3313835.84 frames. ], batch size: 46, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:43:29,195 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 02:43:36,538 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0094, 4.6645, 4.9206, 5.2842, 5.4311, 4.6991, 5.3835, 5.3356], device='cuda:0'), covar=tensor([0.0819, 0.0794, 0.1544, 0.0495, 0.0374, 0.0558, 0.0353, 0.0383], device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0441, 0.0597, 0.0462, 0.0344, 0.0326, 0.0358, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 02:43:57,764 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=33939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:44:05,159 INFO [optim.py:368] (0/8) 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 02:44:13,353 INFO [train.py:904] (0/8) Epoch 4, batch 3500, loss[loss=0.2294, simple_loss=0.2981, pruned_loss=0.08032, over 16802.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3003, pruned_loss=0.07568, over 3313205.93 frames. ], batch size: 102, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:45:22,253 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7006, 4.3656, 4.3710, 3.2895, 4.1633, 4.4805, 4.2523, 2.7484], device='cuda:0'), covar=tensor([0.0275, 0.0018, 0.0028, 0.0192, 0.0027, 0.0030, 0.0023, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0053, 0.0057, 0.0109, 0.0059, 0.0065, 0.0058, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:45:23,331 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-34000.pt 2023-04-28 02:45:26,903 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34000.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:45:27,575 INFO [train.py:904] (0/8) Epoch 4, batch 3550, loss[loss=0.2626, simple_loss=0.3213, pruned_loss=0.1019, over 11968.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3, pruned_loss=0.07508, over 3308643.30 frames. ], batch size: 246, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:46:03,928 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 02:46:11,013 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34033.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 02:46:27,205 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34044.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:46:27,927 INFO [optim.py:368] (0/8) 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 02:46:29,639 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0943, 4.0899, 2.5895, 5.3412, 5.1036, 4.4424, 2.1021, 3.2415], device='cuda:0'), covar=tensor([0.1051, 0.0301, 0.1043, 0.0065, 0.0186, 0.0267, 0.0970, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0135, 0.0159, 0.0078, 0.0170, 0.0156, 0.0152, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 02:46:35,601 INFO [train.py:904] (0/8) Epoch 4, batch 3600, loss[loss=0.2199, simple_loss=0.294, pruned_loss=0.07285, over 16462.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.2988, pruned_loss=0.07443, over 3315615.96 frames. ], batch size: 68, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:42,367 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.05 vs. limit=5.0 2023-04-28 02:47:48,229 INFO [train.py:904] (0/8) Epoch 4, batch 3650, loss[loss=0.2202, simple_loss=0.2862, pruned_loss=0.07712, over 16885.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.2975, pruned_loss=0.07449, over 3314404.81 frames. ], batch size: 96, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:47:54,717 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34105.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:29,169 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:48:31,450 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9508, 4.3008, 4.4171, 3.5014, 4.2543, 4.4482, 4.1701, 2.4975], device='cuda:0'), covar=tensor([0.0227, 0.0021, 0.0027, 0.0151, 0.0023, 0.0042, 0.0020, 0.0242], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0054, 0.0058, 0.0109, 0.0059, 0.0064, 0.0059, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:48:53,741 INFO [optim.py:368] (0/8) 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 02:48:54,291 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:02,649 INFO [train.py:904] (0/8) Epoch 4, batch 3700, loss[loss=0.2224, simple_loss=0.2816, pruned_loss=0.08162, over 11404.00 frames. ], tot_loss[loss=0.224, simple_loss=0.296, pruned_loss=0.07599, over 3294346.77 frames. ], batch size: 248, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:49:41,417 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34176.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:49:42,936 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 02:50:16,877 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8977, 3.2123, 2.3595, 4.5555, 4.2961, 4.2023, 1.6117, 2.9735], device='cuda:0'), covar=tensor([0.1207, 0.0425, 0.1153, 0.0060, 0.0250, 0.0256, 0.1199, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0137, 0.0162, 0.0078, 0.0168, 0.0157, 0.0154, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 02:50:17,552 INFO [train.py:904] (0/8) Epoch 4, batch 3750, loss[loss=0.2504, simple_loss=0.3059, pruned_loss=0.0974, over 16899.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.2964, pruned_loss=0.07838, over 3300522.91 frames. ], batch size: 109, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:50:24,745 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:51:21,425 INFO [optim.py:368] (0/8) 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 02:51:29,937 INFO [train.py:904] (0/8) Epoch 4, batch 3800, loss[loss=0.2215, simple_loss=0.3054, pruned_loss=0.06878, over 17147.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.2974, pruned_loss=0.0796, over 3284352.48 frames. ], batch size: 49, lr: 1.66e-02, grad_scale: 8.0 2023-04-28 02:51:58,587 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7984, 4.6821, 4.6695, 4.1898, 4.6323, 2.2216, 4.4019, 4.6832], device='cuda:0'), covar=tensor([0.0055, 0.0046, 0.0070, 0.0229, 0.0056, 0.1154, 0.0080, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0079, 0.0118, 0.0127, 0.0088, 0.0128, 0.0105, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:52:35,708 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34295.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:52:44,504 INFO [train.py:904] (0/8) Epoch 4, batch 3850, loss[loss=0.2071, simple_loss=0.2742, pruned_loss=0.07004, over 16835.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.2966, pruned_loss=0.0798, over 3284768.37 frames. ], batch size: 102, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:53:31,118 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 02:53:57,009 INFO [train.py:904] (0/8) Epoch 4, batch 3900, loss[loss=0.232, simple_loss=0.292, pruned_loss=0.08601, over 16663.00 frames. ], tot_loss[loss=0.2275, simple_loss=0.2956, pruned_loss=0.07973, over 3282849.91 frames. ], batch size: 134, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:54:40,915 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:54:44,875 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 02:55:09,056 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 4, batch 3950, loss[loss=0.2385, simple_loss=0.2961, pruned_loss=0.09042, over 16433.00 frames. ], tot_loss[loss=0.2278, simple_loss=0.2951, pruned_loss=0.08028, over 3291977.83 frames. ], batch size: 146, lr: 1.65e-02, grad_scale: 4.0 2023-04-28 02:56:16,761 INFO [optim.py:368] (0/8) 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 02:56:24,493 INFO [train.py:904] (0/8) Epoch 4, batch 4000, loss[loss=0.2406, simple_loss=0.3184, pruned_loss=0.08136, over 15439.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.2945, pruned_loss=0.08037, over 3286099.92 frames. ], batch size: 190, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:56:31,517 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4483, 4.4260, 4.3010, 4.3248, 3.9490, 4.3310, 4.2209, 4.0770], device='cuda:0'), covar=tensor([0.0403, 0.0238, 0.0205, 0.0164, 0.0801, 0.0255, 0.0365, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0164, 0.0208, 0.0174, 0.0242, 0.0196, 0.0149, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:56:59,089 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4529, 1.7667, 2.4865, 3.2719, 2.9820, 3.5969, 1.7911, 3.4812], device='cuda:0'), covar=tensor([0.0038, 0.0224, 0.0136, 0.0084, 0.0083, 0.0047, 0.0208, 0.0022], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0136, 0.0122, 0.0117, 0.0115, 0.0088, 0.0131, 0.0074], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 02:57:04,140 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9223, 3.2897, 3.3494, 2.2518, 3.2006, 3.2767, 3.3176, 1.6191], device='cuda:0'), covar=tensor([0.0309, 0.0034, 0.0021, 0.0203, 0.0034, 0.0038, 0.0022, 0.0319], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0052, 0.0055, 0.0107, 0.0059, 0.0062, 0.0058, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:57:36,439 INFO [train.py:904] (0/8) Epoch 4, batch 4050, loss[loss=0.1808, simple_loss=0.2658, pruned_loss=0.0479, over 16808.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2932, pruned_loss=0.07786, over 3286523.66 frames. ], batch size: 102, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:57:36,755 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:58:41,516 INFO [optim.py:368] (0/8) 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 02:58:48,896 INFO [train.py:904] (0/8) Epoch 4, batch 4100, loss[loss=0.2423, simple_loss=0.3204, pruned_loss=0.08207, over 16523.00 frames. ], tot_loss[loss=0.223, simple_loss=0.2935, pruned_loss=0.07622, over 3283939.57 frames. ], batch size: 62, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 02:59:28,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9862, 2.6766, 2.6945, 2.0068, 2.5827, 2.6346, 2.6828, 1.6501], device='cuda:0'), covar=tensor([0.0204, 0.0035, 0.0028, 0.0167, 0.0041, 0.0040, 0.0032, 0.0255], device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0050, 0.0053, 0.0105, 0.0057, 0.0060, 0.0056, 0.0100], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0001, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 02:59:53,995 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 02:59:56,607 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 4, batch 4150, loss[loss=0.2389, simple_loss=0.3262, pruned_loss=0.07585, over 16840.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3023, pruned_loss=0.08017, over 3258060.29 frames. ], batch size: 102, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:00:36,050 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6168, 2.7776, 2.2653, 4.3540, 3.9820, 3.8780, 1.5331, 2.9179], device='cuda:0'), covar=tensor([0.1305, 0.0525, 0.1142, 0.0055, 0.0153, 0.0281, 0.1277, 0.0697], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0135, 0.0159, 0.0075, 0.0156, 0.0152, 0.0151, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 03:00:38,907 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34622.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:00:45,003 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34626.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:10,645 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34643.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:01:15,087 INFO [optim.py:368] (0/8) 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 03:01:22,994 INFO [train.py:904] (0/8) Epoch 4, batch 4200, loss[loss=0.2706, simple_loss=0.3485, pruned_loss=0.09636, over 15482.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3107, pruned_loss=0.08341, over 3215723.94 frames. ], batch size: 191, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:01:28,041 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:13,244 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34683.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:02:19,322 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34687.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:39,191 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34700.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:02:40,353 INFO [train.py:904] (0/8) Epoch 4, batch 4250, loss[loss=0.225, simple_loss=0.3021, pruned_loss=0.07395, over 12176.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.314, pruned_loss=0.08368, over 3194138.99 frames. ], batch size: 246, lr: 1.65e-02, grad_scale: 8.0 2023-04-28 03:02:51,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2312, 4.6181, 2.0536, 4.9038, 3.0214, 4.8810, 2.3358, 3.0660], device='cuda:0'), covar=tensor([0.0095, 0.0113, 0.1576, 0.0016, 0.0596, 0.0120, 0.1326, 0.0543], device='cuda:0'), in_proj_covar=tensor([0.0104, 0.0143, 0.0172, 0.0081, 0.0157, 0.0170, 0.0179, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 03:02:53,636 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=34710.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:03:47,979 INFO [optim.py:368] (0/8) 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 03:03:51,552 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:03:56,222 INFO [train.py:904] (0/8) Epoch 4, batch 4300, loss[loss=0.2787, simple_loss=0.3524, pruned_loss=0.1024, over 16543.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3146, pruned_loss=0.08232, over 3197487.69 frames. ], batch size: 62, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:04:02,958 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 03:04:27,297 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=34771.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:04:59,641 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6882, 4.9460, 4.9655, 5.0601, 4.9437, 5.5156, 5.2064, 4.8812], device='cuda:0'), covar=tensor([0.0766, 0.1295, 0.1070, 0.1103, 0.1903, 0.0624, 0.0886, 0.1860], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0340, 0.0318, 0.0290, 0.0383, 0.0348, 0.0275, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:05:10,253 INFO [train.py:904] (0/8) Epoch 4, batch 4350, loss[loss=0.2608, simple_loss=0.3337, pruned_loss=0.0939, over 16611.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3183, pruned_loss=0.08331, over 3214170.49 frames. ], batch size: 57, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:05:10,568 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=34801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:05:13,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2897, 2.2230, 1.7937, 2.1626, 2.6845, 2.5250, 3.3447, 3.0233], device='cuda:0'), covar=tensor([0.0016, 0.0162, 0.0207, 0.0164, 0.0093, 0.0134, 0.0025, 0.0069], device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0136, 0.0137, 0.0133, 0.0130, 0.0138, 0.0099, 0.0115], device='cuda:0'), out_proj_covar=tensor([9.2544e-05, 1.8053e-04, 1.7645e-04, 1.7364e-04, 1.7379e-04, 1.8507e-04, 1.3054e-04, 1.5578e-04], device='cuda:0') 2023-04-28 03:06:15,521 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.361e+02 3.334e+02 4.019e+02 4.887e+02 8.589e+02, threshold=8.038e+02, percent-clipped=2.0 2023-04-28 03:06:19,652 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=34849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:06:22,175 INFO [train.py:904] (0/8) Epoch 4, batch 4400, loss[loss=0.2697, simple_loss=0.3483, pruned_loss=0.09558, over 15360.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3207, pruned_loss=0.08482, over 3208737.34 frames. ], batch size: 190, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:06:29,364 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8526, 4.8513, 4.6299, 4.6439, 4.2520, 4.7676, 4.5787, 4.3943], device='cuda:0'), covar=tensor([0.0274, 0.0100, 0.0157, 0.0114, 0.0652, 0.0134, 0.0201, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0149, 0.0192, 0.0159, 0.0223, 0.0175, 0.0137, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:07:25,759 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5886, 4.2382, 4.1581, 4.8137, 4.8699, 4.3078, 4.8520, 4.8871], device='cuda:0'), covar=tensor([0.0636, 0.0705, 0.1649, 0.0481, 0.0499, 0.0573, 0.0547, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0387, 0.0509, 0.0404, 0.0300, 0.0292, 0.0314, 0.0326], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:07:32,110 INFO [train.py:904] (0/8) Epoch 4, batch 4450, loss[loss=0.2619, simple_loss=0.3328, pruned_loss=0.09556, over 17130.00 frames. ], tot_loss[loss=0.2474, simple_loss=0.3239, pruned_loss=0.0854, over 3204829.56 frames. ], batch size: 47, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:08:36,347 INFO [optim.py:368] (0/8) 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 03:08:40,983 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:08:43,011 INFO [train.py:904] (0/8) Epoch 4, batch 4500, loss[loss=0.2611, simple_loss=0.3368, pruned_loss=0.0927, over 17126.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.3231, pruned_loss=0.08474, over 3219676.03 frames. ], batch size: 49, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:08:58,926 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0673, 1.6775, 1.4692, 1.5808, 1.8613, 1.7415, 1.6902, 1.9846], device='cuda:0'), covar=tensor([0.0025, 0.0089, 0.0137, 0.0119, 0.0067, 0.0104, 0.0045, 0.0070], device='cuda:0'), in_proj_covar=tensor([0.0070, 0.0137, 0.0138, 0.0135, 0.0130, 0.0141, 0.0100, 0.0117], device='cuda:0'), out_proj_covar=tensor([9.1884e-05, 1.8148e-04, 1.7656e-04, 1.7594e-04, 1.7453e-04, 1.8777e-04, 1.3197e-04, 1.5779e-04], device='cuda:0') 2023-04-28 03:09:23,021 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34978.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:09:28,656 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=34982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:09:33,619 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4907, 2.3709, 2.2243, 4.1744, 1.8169, 3.5009, 2.2659, 2.3487], device='cuda:0'), covar=tensor([0.0497, 0.1408, 0.0783, 0.0270, 0.2660, 0.0476, 0.1302, 0.1970], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0281, 0.0232, 0.0293, 0.0348, 0.0254, 0.0254, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:09:54,756 INFO [train.py:904] (0/8) Epoch 4, batch 4550, loss[loss=0.2403, simple_loss=0.3154, pruned_loss=0.08255, over 16134.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3233, pruned_loss=0.08481, over 3231501.45 frames. ], batch size: 35, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:10:31,494 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0811, 4.5547, 4.4803, 3.0036, 3.9632, 4.3794, 4.1943, 2.0588], device='cuda:0'), covar=tensor([0.0331, 0.0007, 0.0010, 0.0192, 0.0027, 0.0027, 0.0016, 0.0299], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0050, 0.0054, 0.0108, 0.0057, 0.0062, 0.0057, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:10:37,131 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-28 03:10:57,623 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 2.863e+02 3.417e+02 4.239e+02 9.930e+02, threshold=6.834e+02, percent-clipped=3.0 2023-04-28 03:11:04,052 INFO [train.py:904] (0/8) Epoch 4, batch 4600, loss[loss=0.2212, simple_loss=0.3093, pruned_loss=0.06653, over 16732.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3235, pruned_loss=0.08444, over 3225114.11 frames. ], batch size: 89, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:11:25,531 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35066.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:12:01,932 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:12:15,397 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-28 03:12:15,725 INFO [train.py:904] (0/8) Epoch 4, batch 4650, loss[loss=0.2359, simple_loss=0.3147, pruned_loss=0.07853, over 16461.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3216, pruned_loss=0.08369, over 3222113.36 frames. ], batch size: 68, lr: 1.64e-02, grad_scale: 8.0 2023-04-28 03:13:17,490 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 03:13:20,504 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.712e+02 3.466e+02 4.003e+02 6.879e+02, threshold=6.932e+02, percent-clipped=1.0 2023-04-28 03:13:28,257 INFO [train.py:904] (0/8) Epoch 4, batch 4700, loss[loss=0.2346, simple_loss=0.3181, pruned_loss=0.07552, over 16409.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3189, pruned_loss=0.08223, over 3222422.17 frames. ], batch size: 146, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:13:32,513 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35153.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:14:41,629 INFO [train.py:904] (0/8) Epoch 4, batch 4750, loss[loss=0.2214, simple_loss=0.303, pruned_loss=0.0699, over 16731.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.3164, pruned_loss=0.08145, over 3193730.83 frames. ], batch size: 124, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:15:45,876 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.092e+02 2.820e+02 3.474e+02 4.196e+02 7.562e+02, threshold=6.948e+02, percent-clipped=1.0 2023-04-28 03:15:51,523 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:15:53,759 INFO [train.py:904] (0/8) Epoch 4, batch 4800, loss[loss=0.2571, simple_loss=0.3217, pruned_loss=0.09626, over 11750.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3131, pruned_loss=0.07936, over 3190237.29 frames. ], batch size: 246, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:16:28,430 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 03:16:32,913 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35278.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:16:38,123 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:16:59,984 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:05,719 INFO [train.py:904] (0/8) Epoch 4, batch 4850, loss[loss=0.2237, simple_loss=0.3172, pruned_loss=0.06508, over 16706.00 frames. ], tot_loss[loss=0.236, simple_loss=0.3143, pruned_loss=0.07886, over 3193118.75 frames. ], batch size: 124, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:17:16,222 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0297, 3.9311, 4.4203, 4.4473, 4.4490, 4.0900, 4.0777, 3.9824], device='cuda:0'), covar=tensor([0.0211, 0.0307, 0.0303, 0.0341, 0.0300, 0.0219, 0.0621, 0.0321], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0202, 0.0208, 0.0209, 0.0250, 0.0218, 0.0314, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 03:17:41,751 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:17:47,993 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:18:04,668 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-04-28 03:18:11,897 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 03:18:19,090 INFO [train.py:904] (0/8) Epoch 4, batch 4900, loss[loss=0.2265, simple_loss=0.3036, pruned_loss=0.0747, over 16654.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3125, pruned_loss=0.0768, over 3204419.97 frames. ], batch size: 76, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:18:42,167 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35366.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:18:53,739 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:19:03,911 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0319, 5.6021, 5.7577, 5.4599, 5.6948, 6.0849, 5.7919, 5.5234], device='cuda:0'), covar=tensor([0.0556, 0.1199, 0.1434, 0.1658, 0.2012, 0.0945, 0.0842, 0.2086], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0331, 0.0314, 0.0294, 0.0384, 0.0342, 0.0274, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:19:32,924 INFO [train.py:904] (0/8) Epoch 4, batch 4950, loss[loss=0.2547, simple_loss=0.329, pruned_loss=0.09024, over 12045.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3128, pruned_loss=0.07663, over 3214150.66 frames. ], batch size: 246, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:19:41,845 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35406.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:19:52,863 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35414.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:19:55,105 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 03:19:58,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6550, 1.2879, 1.5338, 1.7141, 1.8870, 1.9412, 1.3096, 1.7922], device='cuda:0'), covar=tensor([0.0071, 0.0162, 0.0079, 0.0108, 0.0077, 0.0045, 0.0146, 0.0029], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0132, 0.0118, 0.0117, 0.0117, 0.0082, 0.0130, 0.0075], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 03:20:22,774 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35435.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:37,435 INFO [optim.py:368] (0/8) 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 03:20:41,137 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35448.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:20:45,623 INFO [train.py:904] (0/8) Epoch 4, batch 5000, loss[loss=0.2265, simple_loss=0.3104, pruned_loss=0.07132, over 16424.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.315, pruned_loss=0.07714, over 3213466.52 frames. ], batch size: 146, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:21:57,701 INFO [train.py:904] (0/8) Epoch 4, batch 5050, loss[loss=0.2227, simple_loss=0.3113, pruned_loss=0.06702, over 15342.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3145, pruned_loss=0.07668, over 3215632.69 frames. ], batch size: 190, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:22:15,242 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 03:22:39,987 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35530.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:23:03,496 INFO [optim.py:368] (0/8) 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 03:23:10,558 INFO [train.py:904] (0/8) Epoch 4, batch 5100, loss[loss=0.2093, simple_loss=0.2918, pruned_loss=0.0634, over 16698.00 frames. ], tot_loss[loss=0.2325, simple_loss=0.3129, pruned_loss=0.07603, over 3220046.27 frames. ], batch size: 124, lr: 1.63e-02, grad_scale: 8.0 2023-04-28 03:24:08,132 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35591.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:24:22,837 INFO [train.py:904] (0/8) Epoch 4, batch 5150, loss[loss=0.206, simple_loss=0.3012, pruned_loss=0.05538, over 16843.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3132, pruned_loss=0.0755, over 3213247.34 frames. ], batch size: 102, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:25:28,482 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4825, 2.4518, 2.0124, 2.1849, 2.9451, 2.7335, 3.5204, 3.2209], device='cuda:0'), covar=tensor([0.0019, 0.0161, 0.0191, 0.0200, 0.0078, 0.0142, 0.0043, 0.0086], device='cuda:0'), in_proj_covar=tensor([0.0068, 0.0140, 0.0140, 0.0139, 0.0131, 0.0147, 0.0100, 0.0118], device='cuda:0'), out_proj_covar=tensor([8.9035e-05, 1.8532e-04, 1.7822e-04, 1.8096e-04, 1.7426e-04, 1.9514e-04, 1.3061e-04, 1.5897e-04], device='cuda:0') 2023-04-28 03:25:29,065 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 4, batch 5200, loss[loss=0.2009, simple_loss=0.2746, pruned_loss=0.06367, over 17267.00 frames. ], tot_loss[loss=0.2302, simple_loss=0.3114, pruned_loss=0.07452, over 3212375.77 frames. ], batch size: 52, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:25:43,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5234, 3.6842, 2.9237, 2.1822, 2.5990, 2.1993, 3.7070, 3.9453], device='cuda:0'), covar=tensor([0.2085, 0.0683, 0.1174, 0.1520, 0.1787, 0.1277, 0.0433, 0.0431], device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0239, 0.0252, 0.0227, 0.0290, 0.0191, 0.0224, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:26:46,342 INFO [train.py:904] (0/8) Epoch 4, batch 5250, loss[loss=0.301, simple_loss=0.3485, pruned_loss=0.1268, over 12265.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3092, pruned_loss=0.07439, over 3215390.45 frames. ], batch size: 246, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:26:46,811 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35701.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 03:27:19,556 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9572, 4.1298, 1.4932, 4.3809, 2.6742, 4.3958, 1.8066, 2.9678], device='cuda:0'), covar=tensor([0.0078, 0.0147, 0.1668, 0.0022, 0.0588, 0.0176, 0.1390, 0.0468], device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0144, 0.0177, 0.0079, 0.0161, 0.0171, 0.0182, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 03:27:25,922 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5342, 4.3695, 4.3327, 3.0077, 4.0700, 4.2354, 3.9778, 2.2642], device='cuda:0'), covar=tensor([0.0270, 0.0013, 0.0016, 0.0188, 0.0029, 0.0037, 0.0027, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0051, 0.0056, 0.0112, 0.0059, 0.0065, 0.0059, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:27:28,559 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35730.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:27:52,068 INFO [optim.py:368] (0/8) 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 03:27:55,688 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=35748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:28:00,148 INFO [train.py:904] (0/8) Epoch 4, batch 5300, loss[loss=0.2173, simple_loss=0.2957, pruned_loss=0.0694, over 16293.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3044, pruned_loss=0.07269, over 3216383.23 frames. ], batch size: 165, lr: 1.62e-02, grad_scale: 8.0 2023-04-28 03:28:17,098 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6526, 2.7604, 2.2495, 3.9291, 3.4951, 3.6221, 1.5651, 2.7365], device='cuda:0'), covar=tensor([0.1293, 0.0467, 0.1103, 0.0064, 0.0171, 0.0338, 0.1266, 0.0722], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0136, 0.0165, 0.0075, 0.0151, 0.0156, 0.0156, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 03:28:28,861 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35771.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:02,751 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=35796.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:11,343 INFO [train.py:904] (0/8) Epoch 4, batch 5350, loss[loss=0.2473, simple_loss=0.329, pruned_loss=0.08284, over 16300.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3024, pruned_loss=0.07155, over 3216337.82 frames. ], batch size: 146, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:29:24,890 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=35811.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:29:56,931 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:30:15,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 03:30:17,966 INFO [optim.py:368] (0/8) 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 03:30:23,346 INFO [train.py:904] (0/8) Epoch 4, batch 5400, loss[loss=0.2363, simple_loss=0.321, pruned_loss=0.07575, over 16913.00 frames. ], tot_loss[loss=0.2257, simple_loss=0.3054, pruned_loss=0.07296, over 3208782.06 frames. ], batch size: 109, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:30:53,680 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=35872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:13,846 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=35886.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:31:21,624 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5788, 3.6500, 2.9077, 2.3256, 2.7951, 2.1307, 3.8160, 3.9314], device='cuda:0'), covar=tensor([0.1930, 0.0561, 0.1015, 0.1229, 0.1646, 0.1207, 0.0337, 0.0403], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0240, 0.0253, 0.0227, 0.0296, 0.0192, 0.0227, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:31:38,600 INFO [train.py:904] (0/8) Epoch 4, batch 5450, loss[loss=0.2656, simple_loss=0.3404, pruned_loss=0.09542, over 16771.00 frames. ], tot_loss[loss=0.2303, simple_loss=0.3096, pruned_loss=0.07549, over 3206129.91 frames. ], batch size: 124, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:32:50,258 INFO [optim.py:368] (0/8) 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 03:32:56,445 INFO [train.py:904] (0/8) Epoch 4, batch 5500, loss[loss=0.2753, simple_loss=0.3492, pruned_loss=0.1007, over 17102.00 frames. ], tot_loss[loss=0.2433, simple_loss=0.32, pruned_loss=0.08329, over 3175999.52 frames. ], batch size: 49, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:34:12,336 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-36000.pt 2023-04-28 03:34:16,686 INFO [train.py:904] (0/8) Epoch 4, batch 5550, loss[loss=0.3748, simple_loss=0.3967, pruned_loss=0.1765, over 11166.00 frames. ], tot_loss[loss=0.2552, simple_loss=0.3289, pruned_loss=0.09077, over 3137486.04 frames. ], batch size: 248, lr: 1.62e-02, grad_scale: 4.0 2023-04-28 03:34:17,098 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36001.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 03:34:33,805 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-28 03:35:02,273 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2265, 3.1164, 3.2501, 3.4235, 3.4028, 3.1884, 3.3292, 3.4325], device='cuda:0'), covar=tensor([0.0624, 0.0606, 0.1001, 0.0438, 0.0522, 0.1407, 0.0743, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0402, 0.0522, 0.0401, 0.0305, 0.0293, 0.0319, 0.0333], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:35:02,287 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:35:13,819 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5494, 1.8168, 1.4863, 1.6382, 2.2056, 2.0885, 2.1289, 2.3130], device='cuda:0'), covar=tensor([0.0020, 0.0126, 0.0163, 0.0143, 0.0067, 0.0106, 0.0058, 0.0076], device='cuda:0'), in_proj_covar=tensor([0.0067, 0.0139, 0.0139, 0.0137, 0.0131, 0.0142, 0.0100, 0.0118], device='cuda:0'), out_proj_covar=tensor([8.6802e-05, 1.8240e-04, 1.7682e-04, 1.7673e-04, 1.7369e-04, 1.8720e-04, 1.2936e-04, 1.5791e-04], device='cuda:0') 2023-04-28 03:35:29,648 INFO [optim.py:368] (0/8) 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 03:35:33,007 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36049.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:35:35,581 INFO [train.py:904] (0/8) Epoch 4, batch 5600, loss[loss=0.2703, simple_loss=0.3393, pruned_loss=0.1007, over 16624.00 frames. ], tot_loss[loss=0.2653, simple_loss=0.3358, pruned_loss=0.09739, over 3099666.74 frames. ], batch size: 62, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:36:16,397 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 03:36:20,419 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:36:57,488 INFO [train.py:904] (0/8) Epoch 4, batch 5650, loss[loss=0.2741, simple_loss=0.3419, pruned_loss=0.1032, over 16947.00 frames. ], tot_loss[loss=0.2748, simple_loss=0.3426, pruned_loss=0.1035, over 3067925.14 frames. ], batch size: 109, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:37:40,670 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36127.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:38:09,921 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.856e+02 5.092e+02 6.431e+02 7.893e+02 1.649e+03, threshold=1.286e+03, percent-clipped=5.0 2023-04-28 03:38:17,767 INFO [train.py:904] (0/8) Epoch 4, batch 5700, loss[loss=0.2741, simple_loss=0.3526, pruned_loss=0.09776, over 16888.00 frames. ], tot_loss[loss=0.2761, simple_loss=0.3439, pruned_loss=0.1042, over 3076796.18 frames. ], batch size: 116, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:38:42,669 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:05,044 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36181.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:13,393 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:39:36,196 INFO [train.py:904] (0/8) Epoch 4, batch 5750, loss[loss=0.2377, simple_loss=0.3213, pruned_loss=0.07705, over 16895.00 frames. ], tot_loss[loss=0.2814, simple_loss=0.3482, pruned_loss=0.1073, over 3046214.43 frames. ], batch size: 90, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:39:57,448 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 03:40:29,640 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:42,512 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:40:43,655 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8002, 2.6243, 2.5649, 1.9344, 2.5168, 2.4859, 2.5137, 1.7890], device='cuda:0'), covar=tensor([0.0260, 0.0034, 0.0049, 0.0193, 0.0055, 0.0064, 0.0040, 0.0254], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0049, 0.0054, 0.0108, 0.0057, 0.0062, 0.0057, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:40:47,234 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 03:40:49,465 INFO [optim.py:368] (0/8) 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 03:40:55,821 INFO [train.py:904] (0/8) Epoch 4, batch 5800, loss[loss=0.2718, simple_loss=0.3489, pruned_loss=0.09733, over 16678.00 frames. ], tot_loss[loss=0.2794, simple_loss=0.3474, pruned_loss=0.1057, over 3031186.16 frames. ], batch size: 134, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:42:12,361 INFO [train.py:904] (0/8) Epoch 4, batch 5850, loss[loss=0.2234, simple_loss=0.3038, pruned_loss=0.07151, over 16546.00 frames. ], tot_loss[loss=0.2753, simple_loss=0.3444, pruned_loss=0.1031, over 3033448.93 frames. ], batch size: 68, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:42:20,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4885, 3.9225, 3.2658, 3.7801, 3.3457, 3.4868, 3.6036, 3.7964], device='cuda:0'), covar=tensor([0.1753, 0.1313, 0.2569, 0.0819, 0.1440, 0.2154, 0.1200, 0.1484], device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0422, 0.0374, 0.0280, 0.0275, 0.0280, 0.0340, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:42:47,697 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7465, 4.6208, 4.4733, 3.7919, 4.5114, 1.7219, 4.2673, 4.5127], device='cuda:0'), covar=tensor([0.0046, 0.0043, 0.0064, 0.0294, 0.0048, 0.1527, 0.0070, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0070, 0.0107, 0.0118, 0.0080, 0.0129, 0.0094, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:43:29,071 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 4.315e+02 5.493e+02 6.918e+02 1.909e+03, threshold=1.099e+03, percent-clipped=3.0 2023-04-28 03:43:34,007 INFO [train.py:904] (0/8) Epoch 4, batch 5900, loss[loss=0.2327, simple_loss=0.3192, pruned_loss=0.07314, over 16727.00 frames. ], tot_loss[loss=0.2742, simple_loss=0.3439, pruned_loss=0.1022, over 3047881.78 frames. ], batch size: 83, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:43:40,581 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 03:43:53,596 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4439, 3.3686, 3.3363, 2.8475, 3.3146, 2.0995, 3.1564, 3.0761], device='cuda:0'), covar=tensor([0.0066, 0.0055, 0.0083, 0.0197, 0.0061, 0.1198, 0.0078, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0069, 0.0107, 0.0118, 0.0080, 0.0128, 0.0094, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:44:56,230 INFO [train.py:904] (0/8) Epoch 4, batch 5950, loss[loss=0.3116, simple_loss=0.3652, pruned_loss=0.129, over 11672.00 frames. ], tot_loss[loss=0.2726, simple_loss=0.3441, pruned_loss=0.1005, over 3039302.47 frames. ], batch size: 248, lr: 1.61e-02, grad_scale: 4.0 2023-04-28 03:45:37,410 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36427.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:46:10,129 INFO [optim.py:368] (0/8) 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 03:46:14,587 INFO [train.py:904] (0/8) Epoch 4, batch 6000, loss[loss=0.2554, simple_loss=0.3332, pruned_loss=0.08877, over 16621.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3428, pruned_loss=0.09955, over 3059041.99 frames. ], batch size: 134, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:46:14,588 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 03:46:25,215 INFO [train.py:938] (0/8) Epoch 4, validation: loss=0.1966, simple_loss=0.3069, pruned_loss=0.04309, over 944034.00 frames. 2023-04-28 03:46:25,216 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17747MB 2023-04-28 03:46:49,610 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36467.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:01,321 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36475.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:47:44,060 INFO [train.py:904] (0/8) Epoch 4, batch 6050, loss[loss=0.2432, simple_loss=0.3245, pruned_loss=0.08092, over 17047.00 frames. ], tot_loss[loss=0.2678, simple_loss=0.3401, pruned_loss=0.09779, over 3082319.94 frames. ], batch size: 41, lr: 1.61e-02, grad_scale: 8.0 2023-04-28 03:48:06,426 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:24,472 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-28 03:48:42,934 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0872, 4.1074, 4.1826, 4.1932, 4.2121, 4.6839, 4.3335, 4.0204], device='cuda:0'), covar=tensor([0.1451, 0.1534, 0.1358, 0.1665, 0.2133, 0.0991, 0.1255, 0.2370], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0352, 0.0336, 0.0310, 0.0410, 0.0368, 0.0286, 0.0418], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:48:42,948 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36537.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:48:58,313 INFO [optim.py:368] (0/8) 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 03:49:03,049 INFO [train.py:904] (0/8) Epoch 4, batch 6100, loss[loss=0.2672, simple_loss=0.3377, pruned_loss=0.09838, over 16512.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.34, pruned_loss=0.09719, over 3091833.49 frames. ], batch size: 68, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:50:22,083 INFO [train.py:904] (0/8) Epoch 4, batch 6150, loss[loss=0.2621, simple_loss=0.3348, pruned_loss=0.09472, over 16859.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3386, pruned_loss=0.09691, over 3065557.40 frames. ], batch size: 116, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:50:26,042 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-04-28 03:50:29,661 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36605.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:35,052 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36608.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:50:41,768 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:51:38,880 INFO [optim.py:368] (0/8) 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 03:51:44,332 INFO [train.py:904] (0/8) Epoch 4, batch 6200, loss[loss=0.2995, simple_loss=0.3617, pruned_loss=0.1187, over 16359.00 frames. ], tot_loss[loss=0.2652, simple_loss=0.3372, pruned_loss=0.09662, over 3071410.60 frames. ], batch size: 146, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:52:07,189 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:11,556 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:17,456 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36673.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:52:57,844 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36700.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:52:59,120 INFO [train.py:904] (0/8) Epoch 4, batch 6250, loss[loss=0.2471, simple_loss=0.3263, pruned_loss=0.08395, over 16263.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3359, pruned_loss=0.09507, over 3109974.17 frames. ], batch size: 35, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:53:16,179 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2791, 4.3385, 4.1014, 4.1433, 3.7472, 4.1907, 4.0535, 3.9273], device='cuda:0'), covar=tensor([0.0412, 0.0232, 0.0200, 0.0161, 0.0691, 0.0290, 0.0342, 0.0418], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0156, 0.0195, 0.0161, 0.0223, 0.0185, 0.0144, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:54:09,144 INFO [optim.py:368] (0/8) 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 03:54:14,290 INFO [train.py:904] (0/8) Epoch 4, batch 6300, loss[loss=0.2315, simple_loss=0.3191, pruned_loss=0.07191, over 16850.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3355, pruned_loss=0.09424, over 3109617.42 frames. ], batch size: 102, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:54:30,640 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36761.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:55:32,015 INFO [train.py:904] (0/8) Epoch 4, batch 6350, loss[loss=0.2647, simple_loss=0.3337, pruned_loss=0.09783, over 16403.00 frames. ], tot_loss[loss=0.2646, simple_loss=0.3369, pruned_loss=0.09621, over 3096556.97 frames. ], batch size: 146, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:55:35,098 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3988, 3.3612, 2.6573, 2.1487, 2.4803, 2.1066, 3.3452, 3.6631], device='cuda:0'), covar=tensor([0.2043, 0.0606, 0.1176, 0.1373, 0.1578, 0.1297, 0.0429, 0.0553], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0245, 0.0259, 0.0230, 0.0305, 0.0194, 0.0229, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 03:55:38,814 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36805.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 03:56:15,057 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6565, 2.9336, 2.2606, 4.2878, 3.8317, 3.9633, 1.6955, 2.9165], device='cuda:0'), covar=tensor([0.1521, 0.0584, 0.1315, 0.0120, 0.0241, 0.0336, 0.1395, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0137, 0.0164, 0.0076, 0.0157, 0.0161, 0.0156, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 03:56:23,802 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9559, 3.6239, 2.6345, 5.1111, 4.6802, 4.4507, 2.0351, 3.4177], device='cuda:0'), covar=tensor([0.1370, 0.0468, 0.1204, 0.0098, 0.0164, 0.0294, 0.1221, 0.0643], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0137, 0.0165, 0.0076, 0.0158, 0.0161, 0.0157, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 03:56:27,998 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=36837.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:56:42,969 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 03:56:43,691 INFO [optim.py:368] (0/8) 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 03:56:48,104 INFO [train.py:904] (0/8) Epoch 4, batch 6400, loss[loss=0.2526, simple_loss=0.331, pruned_loss=0.08712, over 16448.00 frames. ], tot_loss[loss=0.266, simple_loss=0.3372, pruned_loss=0.09745, over 3089162.40 frames. ], batch size: 146, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:56:48,674 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7577, 4.5317, 4.5954, 2.2651, 4.8871, 4.9957, 3.4917, 3.7406], device='cuda:0'), covar=tensor([0.0748, 0.0093, 0.0141, 0.1086, 0.0034, 0.0025, 0.0273, 0.0343], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0085, 0.0081, 0.0143, 0.0068, 0.0073, 0.0115, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 03:57:10,843 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36866.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 03:57:39,967 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=36885.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:01,700 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4683, 3.5049, 3.2414, 3.2855, 3.0736, 3.3795, 3.2524, 3.2685], device='cuda:0'), covar=tensor([0.0357, 0.0220, 0.0173, 0.0147, 0.0462, 0.0173, 0.0841, 0.0295], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0158, 0.0196, 0.0161, 0.0223, 0.0187, 0.0146, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:58:04,376 INFO [train.py:904] (0/8) Epoch 4, batch 6450, loss[loss=0.237, simple_loss=0.3123, pruned_loss=0.08082, over 16906.00 frames. ], tot_loss[loss=0.2637, simple_loss=0.3359, pruned_loss=0.09576, over 3093936.81 frames. ], batch size: 116, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:58:20,096 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36911.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:58:40,393 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8221, 4.4836, 4.5105, 1.9876, 4.8151, 4.9022, 3.4199, 3.7370], device='cuda:0'), covar=tensor([0.0844, 0.0096, 0.0140, 0.1315, 0.0040, 0.0036, 0.0293, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0083, 0.0079, 0.0141, 0.0066, 0.0072, 0.0112, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 03:59:18,068 INFO [optim.py:368] (0/8) 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 03:59:23,784 INFO [train.py:904] (0/8) Epoch 4, batch 6500, loss[loss=0.2643, simple_loss=0.3441, pruned_loss=0.09226, over 16671.00 frames. ], tot_loss[loss=0.261, simple_loss=0.3334, pruned_loss=0.09426, over 3099176.09 frames. ], batch size: 89, lr: 1.60e-02, grad_scale: 8.0 2023-04-28 03:59:34,554 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=36958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:38,830 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36961.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:43,469 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:49,608 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=36968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 03:59:54,823 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6112, 4.9187, 4.9761, 5.0240, 4.9533, 5.5033, 5.1228, 4.9128], device='cuda:0'), covar=tensor([0.0764, 0.1246, 0.1234, 0.1365, 0.1992, 0.0729, 0.0886, 0.1755], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0340, 0.0332, 0.0309, 0.0396, 0.0358, 0.0276, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 03:59:56,675 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=36972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:00:44,419 INFO [train.py:904] (0/8) Epoch 4, batch 6550, loss[loss=0.2883, simple_loss=0.3695, pruned_loss=0.1035, over 16625.00 frames. ], tot_loss[loss=0.2632, simple_loss=0.3364, pruned_loss=0.09504, over 3122536.48 frames. ], batch size: 62, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:00:50,037 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:13,054 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:01:41,952 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5152, 4.2969, 4.5253, 3.0459, 4.4611, 1.5382, 4.1985, 4.3218], device='cuda:0'), covar=tensor([0.0143, 0.0107, 0.0087, 0.0584, 0.0077, 0.2192, 0.0115, 0.0206], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0069, 0.0104, 0.0114, 0.0077, 0.0127, 0.0094, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:01:50,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2255, 1.5643, 2.2444, 2.8340, 3.1838, 3.5783, 1.5139, 3.5139], device='cuda:0'), covar=tensor([0.0048, 0.0218, 0.0125, 0.0092, 0.0059, 0.0043, 0.0203, 0.0034], device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0132, 0.0119, 0.0113, 0.0117, 0.0081, 0.0130, 0.0075], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 04:01:56,567 INFO [optim.py:368] (0/8) 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 04:02:01,495 INFO [train.py:904] (0/8) Epoch 4, batch 6600, loss[loss=0.2563, simple_loss=0.3348, pruned_loss=0.08891, over 16758.00 frames. ], tot_loss[loss=0.265, simple_loss=0.3385, pruned_loss=0.09577, over 3112462.84 frames. ], batch size: 83, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:02:09,075 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37056.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 04:02:22,601 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:02:41,151 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9773, 5.2591, 4.9736, 4.9386, 4.6939, 4.4340, 4.7807, 5.3165], device='cuda:0'), covar=tensor([0.0580, 0.0597, 0.0888, 0.0470, 0.0495, 0.0654, 0.0559, 0.0689], device='cuda:0'), in_proj_covar=tensor([0.0322, 0.0427, 0.0373, 0.0283, 0.0275, 0.0281, 0.0349, 0.0303], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:02:50,454 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-28 04:03:19,962 INFO [train.py:904] (0/8) Epoch 4, batch 6650, loss[loss=0.2543, simple_loss=0.3294, pruned_loss=0.08962, over 16725.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.3397, pruned_loss=0.09747, over 3089336.73 frames. ], batch size: 76, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:04:33,027 INFO [optim.py:368] (0/8) 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 04:04:37,787 INFO [train.py:904] (0/8) Epoch 4, batch 6700, loss[loss=0.3406, simple_loss=0.3837, pruned_loss=0.1488, over 11730.00 frames. ], tot_loss[loss=0.2669, simple_loss=0.3383, pruned_loss=0.09773, over 3084765.77 frames. ], batch size: 246, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:04:53,107 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37161.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 04:05:19,315 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9634, 1.6295, 2.2442, 2.7620, 2.7647, 3.2018, 1.7126, 3.3315], device='cuda:0'), covar=tensor([0.0053, 0.0203, 0.0122, 0.0092, 0.0088, 0.0047, 0.0206, 0.0032], device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0130, 0.0119, 0.0111, 0.0118, 0.0082, 0.0130, 0.0075], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 04:05:53,502 INFO [train.py:904] (0/8) Epoch 4, batch 6750, loss[loss=0.2474, simple_loss=0.3171, pruned_loss=0.08881, over 16770.00 frames. ], tot_loss[loss=0.2659, simple_loss=0.3367, pruned_loss=0.09754, over 3094263.14 frames. ], batch size: 57, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:06:45,961 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-28 04:07:06,101 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.556e+02 4.104e+02 5.227e+02 6.545e+02 1.583e+03, threshold=1.045e+03, percent-clipped=5.0 2023-04-28 04:07:10,560 INFO [train.py:904] (0/8) Epoch 4, batch 6800, loss[loss=0.309, simple_loss=0.3604, pruned_loss=0.1288, over 11634.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3373, pruned_loss=0.09756, over 3087666.19 frames. ], batch size: 248, lr: 1.59e-02, grad_scale: 8.0 2023-04-28 04:07:27,145 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37261.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:31,013 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:35,281 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37267.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:36,652 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:36,675 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37268.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:07:58,223 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9041, 2.2351, 2.2168, 3.1292, 2.1140, 2.7885, 2.3138, 1.9926], device='cuda:0'), covar=tensor([0.0451, 0.1235, 0.0677, 0.0285, 0.1925, 0.0552, 0.1205, 0.1787], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0279, 0.0234, 0.0294, 0.0348, 0.0258, 0.0258, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:08:27,085 INFO [train.py:904] (0/8) Epoch 4, batch 6850, loss[loss=0.2617, simple_loss=0.351, pruned_loss=0.08621, over 17014.00 frames. ], tot_loss[loss=0.2676, simple_loss=0.3386, pruned_loss=0.0983, over 3086152.51 frames. ], batch size: 55, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:08:27,882 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 04:08:38,767 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37309.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:43,093 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37312.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:45,129 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6192, 3.9044, 2.9451, 2.3827, 2.8279, 2.2317, 3.9980, 4.0566], device='cuda:0'), covar=tensor([0.2062, 0.0519, 0.1186, 0.1439, 0.1856, 0.1290, 0.0371, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0245, 0.0262, 0.0234, 0.0309, 0.0195, 0.0229, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:08:46,844 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37314.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:08:49,157 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37316.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:08,803 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37329.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:36,165 INFO [optim.py:368] (0/8) 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 04:09:37,138 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 04:09:39,839 INFO [train.py:904] (0/8) Epoch 4, batch 6900, loss[loss=0.2757, simple_loss=0.3417, pruned_loss=0.1048, over 16486.00 frames. ], tot_loss[loss=0.2663, simple_loss=0.3401, pruned_loss=0.09626, over 3120559.36 frames. ], batch size: 75, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:09:48,174 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37356.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:09:52,854 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:09:59,839 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:10:55,361 INFO [train.py:904] (0/8) Epoch 4, batch 6950, loss[loss=0.3266, simple_loss=0.3874, pruned_loss=0.1329, over 15327.00 frames. ], tot_loss[loss=0.271, simple_loss=0.3434, pruned_loss=0.09931, over 3112224.74 frames. ], batch size: 190, lr: 1.59e-02, grad_scale: 4.0 2023-04-28 04:11:00,668 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37404.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:11:32,936 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=37425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:12:10,360 INFO [optim.py:368] (0/8) 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 04:12:11,791 INFO [train.py:904] (0/8) Epoch 4, batch 7000, loss[loss=0.2364, simple_loss=0.3257, pruned_loss=0.07356, over 17023.00 frames. ], tot_loss[loss=0.2709, simple_loss=0.3434, pruned_loss=0.09922, over 3082239.17 frames. ], batch size: 41, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:12:28,135 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37461.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 04:12:39,536 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 04:13:26,173 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-28 04:13:29,632 INFO [train.py:904] (0/8) Epoch 4, batch 7050, loss[loss=0.2396, simple_loss=0.3172, pruned_loss=0.08098, over 16639.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.3433, pruned_loss=0.09805, over 3101054.13 frames. ], batch size: 62, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:13:42,601 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37509.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 04:13:46,869 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8534, 4.1392, 3.8706, 3.8786, 3.5250, 3.6880, 3.8309, 4.0415], device='cuda:0'), covar=tensor([0.0611, 0.0668, 0.0921, 0.0492, 0.0606, 0.1146, 0.0563, 0.0780], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0425, 0.0374, 0.0282, 0.0273, 0.0282, 0.0346, 0.0302], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:14:45,395 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 3.107e+02 4.211e+02 5.076e+02 6.594e+02 1.566e+03, threshold=1.015e+03, percent-clipped=2.0 2023-04-28 04:14:46,717 INFO [train.py:904] (0/8) Epoch 4, batch 7100, loss[loss=0.3052, simple_loss=0.3494, pruned_loss=0.1305, over 11389.00 frames. ], tot_loss[loss=0.2695, simple_loss=0.3423, pruned_loss=0.09834, over 3076008.72 frames. ], batch size: 247, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:15:13,009 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:02,672 INFO [train.py:904] (0/8) Epoch 4, batch 7150, loss[loss=0.2474, simple_loss=0.3264, pruned_loss=0.08422, over 16489.00 frames. ], tot_loss[loss=0.2673, simple_loss=0.3397, pruned_loss=0.0975, over 3082457.94 frames. ], batch size: 68, lr: 1.58e-02, grad_scale: 2.0 2023-04-28 04:16:22,613 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:24,182 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:16:38,277 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37624.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:17,124 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.158e+02 3.757e+02 4.855e+02 6.210e+02 1.451e+03, threshold=9.710e+02, percent-clipped=1.0 2023-04-28 04:17:18,898 INFO [train.py:904] (0/8) Epoch 4, batch 7200, loss[loss=0.2056, simple_loss=0.2862, pruned_loss=0.06245, over 16652.00 frames. ], tot_loss[loss=0.2624, simple_loss=0.3361, pruned_loss=0.09432, over 3098651.94 frames. ], batch size: 57, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:17:33,230 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:17:35,458 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37662.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:18:40,026 INFO [train.py:904] (0/8) Epoch 4, batch 7250, loss[loss=0.2745, simple_loss=0.3399, pruned_loss=0.1045, over 15361.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3342, pruned_loss=0.09361, over 3101612.65 frames. ], batch size: 191, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:18:51,359 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:10,339 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=37720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:19:55,473 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 4, batch 7300, loss[loss=0.2685, simple_loss=0.3428, pruned_loss=0.09707, over 15404.00 frames. ], tot_loss[loss=0.2596, simple_loss=0.3329, pruned_loss=0.09316, over 3082533.60 frames. ], batch size: 190, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:21:14,983 INFO [train.py:904] (0/8) Epoch 4, batch 7350, loss[loss=0.24, simple_loss=0.3215, pruned_loss=0.07922, over 16954.00 frames. ], tot_loss[loss=0.2593, simple_loss=0.3326, pruned_loss=0.09296, over 3064735.93 frames. ], batch size: 109, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:21:15,947 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 04:21:41,017 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9246, 4.1706, 1.7895, 4.7764, 2.7474, 4.6059, 2.1570, 2.8225], device='cuda:0'), covar=tensor([0.0118, 0.0205, 0.1666, 0.0018, 0.0752, 0.0212, 0.1430, 0.0666], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0143, 0.0175, 0.0077, 0.0160, 0.0173, 0.0186, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 04:22:29,893 INFO [optim.py:368] (0/8) 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 04:22:31,820 INFO [train.py:904] (0/8) Epoch 4, batch 7400, loss[loss=0.264, simple_loss=0.3452, pruned_loss=0.0914, over 16868.00 frames. ], tot_loss[loss=0.2608, simple_loss=0.3337, pruned_loss=0.09389, over 3066831.76 frames. ], batch size: 102, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:23:22,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4576, 4.0500, 4.1668, 1.7740, 4.4119, 4.4182, 3.0973, 3.1503], device='cuda:0'), covar=tensor([0.0822, 0.0091, 0.0133, 0.1197, 0.0039, 0.0041, 0.0326, 0.0389], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0085, 0.0084, 0.0143, 0.0070, 0.0074, 0.0114, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 04:23:49,012 INFO [train.py:904] (0/8) Epoch 4, batch 7450, loss[loss=0.2661, simple_loss=0.3409, pruned_loss=0.09569, over 16927.00 frames. ], tot_loss[loss=0.2628, simple_loss=0.3352, pruned_loss=0.09521, over 3082149.34 frames. ], batch size: 109, lr: 1.58e-02, grad_scale: 4.0 2023-04-28 04:23:53,043 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6606, 4.1654, 3.9115, 2.2252, 3.0847, 2.4959, 3.8113, 4.0395], device='cuda:0'), covar=tensor([0.0216, 0.0456, 0.0452, 0.1517, 0.0705, 0.0906, 0.0594, 0.0696], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0122, 0.0155, 0.0141, 0.0135, 0.0127, 0.0142, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 04:24:26,711 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=37924.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:06,405 INFO [optim.py:368] (0/8) 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 04:25:07,724 INFO [train.py:904] (0/8) Epoch 4, batch 7500, loss[loss=0.2673, simple_loss=0.3449, pruned_loss=0.09487, over 16832.00 frames. ], tot_loss[loss=0.263, simple_loss=0.3362, pruned_loss=0.09485, over 3089108.00 frames. ], batch size: 116, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:25:40,564 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=37972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:25:50,262 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=37979.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:26:22,959 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-38000.pt 2023-04-28 04:26:27,346 INFO [train.py:904] (0/8) Epoch 4, batch 7550, loss[loss=0.2297, simple_loss=0.3143, pruned_loss=0.07262, over 16524.00 frames. ], tot_loss[loss=0.2619, simple_loss=0.3348, pruned_loss=0.09453, over 3091698.76 frames. ], batch size: 75, lr: 1.57e-02, grad_scale: 4.0 2023-04-28 04:26:51,533 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8437, 2.1694, 2.2040, 3.1622, 2.0661, 2.7757, 2.3353, 2.0442], device='cuda:0'), covar=tensor([0.0465, 0.1323, 0.0669, 0.0343, 0.1972, 0.0573, 0.1285, 0.1665], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0281, 0.0236, 0.0296, 0.0352, 0.0262, 0.0259, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:26:54,422 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:22,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5856, 4.5462, 4.5157, 3.7694, 4.4129, 1.6650, 4.2948, 4.4734], device='cuda:0'), covar=tensor([0.0068, 0.0044, 0.0066, 0.0273, 0.0052, 0.1585, 0.0070, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0070, 0.0108, 0.0116, 0.0080, 0.0131, 0.0095, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:27:23,996 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:27:38,146 INFO [optim.py:368] (0/8) 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 04:27:40,072 INFO [train.py:904] (0/8) Epoch 4, batch 7600, loss[loss=0.2624, simple_loss=0.3295, pruned_loss=0.09763, over 16201.00 frames. ], tot_loss[loss=0.2626, simple_loss=0.3345, pruned_loss=0.09541, over 3086566.20 frames. ], batch size: 165, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:28:05,544 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:28:55,070 INFO [train.py:904] (0/8) Epoch 4, batch 7650, loss[loss=0.3148, simple_loss=0.3604, pruned_loss=0.1346, over 10826.00 frames. ], tot_loss[loss=0.2658, simple_loss=0.3367, pruned_loss=0.09743, over 3070405.80 frames. ], batch size: 247, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:29:53,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3087, 3.1636, 3.2764, 3.4549, 3.4748, 3.1846, 3.4375, 3.4824], device='cuda:0'), covar=tensor([0.0589, 0.0513, 0.0868, 0.0396, 0.0404, 0.1719, 0.0683, 0.0451], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0403, 0.0523, 0.0419, 0.0312, 0.0304, 0.0339, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:30:08,165 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8407, 2.6631, 2.5212, 1.6448, 2.6541, 2.6705, 2.3688, 2.2241], device='cuda:0'), covar=tensor([0.0831, 0.0135, 0.0183, 0.1038, 0.0108, 0.0119, 0.0325, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0084, 0.0083, 0.0142, 0.0069, 0.0073, 0.0114, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 04:30:08,820 INFO [optim.py:368] (0/8) 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 04:30:09,987 INFO [train.py:904] (0/8) Epoch 4, batch 7700, loss[loss=0.255, simple_loss=0.3267, pruned_loss=0.09162, over 16769.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3372, pruned_loss=0.09817, over 3079505.88 frames. ], batch size: 124, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:31:26,737 INFO [train.py:904] (0/8) Epoch 4, batch 7750, loss[loss=0.2449, simple_loss=0.3201, pruned_loss=0.08489, over 16752.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3376, pruned_loss=0.0979, over 3081450.11 frames. ], batch size: 124, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:32:40,403 INFO [optim.py:368] (0/8) 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 04:32:42,188 INFO [train.py:904] (0/8) Epoch 4, batch 7800, loss[loss=0.3558, simple_loss=0.3874, pruned_loss=0.1621, over 11491.00 frames. ], tot_loss[loss=0.2683, simple_loss=0.3388, pruned_loss=0.09886, over 3075268.70 frames. ], batch size: 247, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:33:07,123 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 04:33:58,983 INFO [train.py:904] (0/8) Epoch 4, batch 7850, loss[loss=0.2563, simple_loss=0.3395, pruned_loss=0.08655, over 16552.00 frames. ], tot_loss[loss=0.2665, simple_loss=0.3387, pruned_loss=0.0972, over 3094240.76 frames. ], batch size: 68, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:34:28,441 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2572, 4.3186, 4.8357, 4.7883, 4.7381, 4.2860, 4.4165, 4.2398], device='cuda:0'), covar=tensor([0.0227, 0.0279, 0.0204, 0.0290, 0.0354, 0.0248, 0.0662, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0208, 0.0213, 0.0214, 0.0256, 0.0222, 0.0325, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 04:34:50,569 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38335.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:12,445 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.779e+02 3.909e+02 4.709e+02 5.998e+02 1.164e+03, threshold=9.418e+02, percent-clipped=1.0 2023-04-28 04:35:14,727 INFO [train.py:904] (0/8) Epoch 4, batch 7900, loss[loss=0.2579, simple_loss=0.3313, pruned_loss=0.09225, over 16413.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3372, pruned_loss=0.09652, over 3079021.97 frames. ], batch size: 68, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:35:16,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2040, 3.8886, 3.9976, 1.6242, 4.2155, 4.1466, 3.0708, 3.0252], device='cuda:0'), covar=tensor([0.0892, 0.0091, 0.0131, 0.1332, 0.0050, 0.0061, 0.0286, 0.0411], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0085, 0.0084, 0.0145, 0.0072, 0.0076, 0.0116, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 04:35:28,522 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:35:30,065 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 04:35:40,894 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1114, 3.0133, 2.7005, 2.0752, 2.5347, 2.1755, 2.7520, 2.9361], device='cuda:0'), covar=tensor([0.0267, 0.0421, 0.0414, 0.1226, 0.0594, 0.0762, 0.0531, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0120, 0.0152, 0.0139, 0.0131, 0.0125, 0.0141, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 04:36:34,681 INFO [train.py:904] (0/8) Epoch 4, batch 7950, loss[loss=0.327, simple_loss=0.3693, pruned_loss=0.1423, over 11751.00 frames. ], tot_loss[loss=0.2662, simple_loss=0.3377, pruned_loss=0.09736, over 3081781.75 frames. ], batch size: 248, lr: 1.57e-02, grad_scale: 8.0 2023-04-28 04:36:35,112 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5268, 1.3702, 1.9040, 2.2740, 2.3421, 2.7108, 1.4891, 2.5854], device='cuda:0'), covar=tensor([0.0053, 0.0206, 0.0133, 0.0116, 0.0090, 0.0060, 0.0187, 0.0045], device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0131, 0.0117, 0.0110, 0.0115, 0.0081, 0.0128, 0.0073], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 04:37:05,114 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:37:49,211 INFO [optim.py:368] (0/8) 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 04:37:50,956 INFO [train.py:904] (0/8) Epoch 4, batch 8000, loss[loss=0.2334, simple_loss=0.3165, pruned_loss=0.07515, over 16529.00 frames. ], tot_loss[loss=0.2667, simple_loss=0.3381, pruned_loss=0.09762, over 3084663.21 frames. ], batch size: 68, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:39:04,867 INFO [train.py:904] (0/8) Epoch 4, batch 8050, loss[loss=0.2888, simple_loss=0.355, pruned_loss=0.1113, over 15358.00 frames. ], tot_loss[loss=0.2638, simple_loss=0.3361, pruned_loss=0.09575, over 3095429.20 frames. ], batch size: 190, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:40:21,963 INFO [optim.py:368] (0/8) 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 04:40:23,276 INFO [train.py:904] (0/8) Epoch 4, batch 8100, loss[loss=0.2585, simple_loss=0.3289, pruned_loss=0.09404, over 16229.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.336, pruned_loss=0.09537, over 3097018.67 frames. ], batch size: 165, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:41:37,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0511, 3.4827, 3.4280, 2.4041, 3.2394, 3.4879, 3.2544, 1.8003], device='cuda:0'), covar=tensor([0.0323, 0.0023, 0.0034, 0.0215, 0.0044, 0.0066, 0.0041, 0.0298], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0050, 0.0058, 0.0112, 0.0057, 0.0067, 0.0061, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 04:41:41,530 INFO [train.py:904] (0/8) Epoch 4, batch 8150, loss[loss=0.2361, simple_loss=0.3062, pruned_loss=0.08299, over 16715.00 frames. ], tot_loss[loss=0.2625, simple_loss=0.3345, pruned_loss=0.09524, over 3077385.64 frames. ], batch size: 57, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:41:48,015 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5631, 4.3324, 4.5111, 4.8237, 4.9117, 4.3822, 4.8889, 4.8853], device='cuda:0'), covar=tensor([0.0847, 0.0730, 0.1220, 0.0458, 0.0445, 0.0521, 0.0494, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0422, 0.0547, 0.0433, 0.0324, 0.0311, 0.0350, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:42:07,192 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0290, 4.1241, 4.2292, 4.2310, 4.2210, 4.7151, 4.3355, 4.0998], device='cuda:0'), covar=tensor([0.1505, 0.1490, 0.1420, 0.1759, 0.2434, 0.1020, 0.1180, 0.2309], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0359, 0.0346, 0.0314, 0.0410, 0.0384, 0.0291, 0.0420], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 04:42:34,028 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=38635.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:42:54,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1042, 3.8148, 3.4952, 2.0688, 2.8169, 2.2889, 3.4267, 3.6203], device='cuda:0'), covar=tensor([0.0239, 0.0447, 0.0410, 0.1352, 0.0660, 0.0881, 0.0586, 0.0622], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0122, 0.0156, 0.0140, 0.0134, 0.0126, 0.0142, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 04:42:57,104 INFO [optim.py:368] (0/8) 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 04:42:59,082 INFO [train.py:904] (0/8) Epoch 4, batch 8200, loss[loss=0.2068, simple_loss=0.2831, pruned_loss=0.06529, over 17105.00 frames. ], tot_loss[loss=0.2611, simple_loss=0.3327, pruned_loss=0.09481, over 3082125.43 frames. ], batch size: 49, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:43:53,410 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=38683.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:44:13,135 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9055, 4.2158, 4.4871, 4.4365, 4.4008, 4.1271, 3.7046, 3.9593], device='cuda:0'), covar=tensor([0.0472, 0.0546, 0.0438, 0.0597, 0.0769, 0.0515, 0.1336, 0.0477], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0207, 0.0214, 0.0216, 0.0259, 0.0224, 0.0326, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 04:44:23,330 INFO [train.py:904] (0/8) Epoch 4, batch 8250, loss[loss=0.218, simple_loss=0.3083, pruned_loss=0.06381, over 16856.00 frames. ], tot_loss[loss=0.2578, simple_loss=0.3314, pruned_loss=0.09215, over 3079414.14 frames. ], batch size: 96, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:44:48,502 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=38716.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:45:43,966 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 4, batch 8300, loss[loss=0.2352, simple_loss=0.3163, pruned_loss=0.07699, over 15312.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3271, pruned_loss=0.08796, over 3077959.13 frames. ], batch size: 190, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:46:08,130 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0792, 3.6425, 3.6595, 2.6937, 3.5970, 3.6446, 3.5729, 2.0889], device='cuda:0'), covar=tensor([0.0297, 0.0016, 0.0024, 0.0166, 0.0021, 0.0039, 0.0028, 0.0262], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0048, 0.0055, 0.0108, 0.0055, 0.0065, 0.0059, 0.0102], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 04:47:07,358 INFO [train.py:904] (0/8) Epoch 4, batch 8350, loss[loss=0.2752, simple_loss=0.3281, pruned_loss=0.1111, over 12065.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3258, pruned_loss=0.08543, over 3073585.29 frames. ], batch size: 247, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:47:24,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1240, 1.2182, 1.7017, 1.9185, 2.0079, 2.0813, 1.4046, 2.1387], device='cuda:0'), covar=tensor([0.0061, 0.0225, 0.0131, 0.0122, 0.0096, 0.0082, 0.0192, 0.0061], device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0131, 0.0117, 0.0112, 0.0115, 0.0081, 0.0129, 0.0074], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 04:47:34,112 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8018, 4.6610, 4.6944, 3.3599, 4.5044, 1.6602, 4.3547, 4.5872], device='cuda:0'), covar=tensor([0.0096, 0.0085, 0.0077, 0.0484, 0.0091, 0.2101, 0.0106, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0080, 0.0069, 0.0105, 0.0112, 0.0079, 0.0130, 0.0094, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:47:44,513 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=38823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:48:29,415 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.235e+02 3.176e+02 3.884e+02 4.976e+02 1.172e+03, threshold=7.768e+02, percent-clipped=5.0 2023-04-28 04:48:30,629 INFO [train.py:904] (0/8) Epoch 4, batch 8400, loss[loss=0.2136, simple_loss=0.3086, pruned_loss=0.05925, over 16733.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3213, pruned_loss=0.08134, over 3090209.90 frames. ], batch size: 124, lr: 1.56e-02, grad_scale: 8.0 2023-04-28 04:48:41,836 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5349, 3.4755, 3.4722, 3.0573, 3.4462, 2.0610, 3.2835, 3.1012], device='cuda:0'), covar=tensor([0.0079, 0.0067, 0.0082, 0.0189, 0.0064, 0.1400, 0.0084, 0.0127], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0069, 0.0105, 0.0111, 0.0079, 0.0131, 0.0095, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:49:26,862 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=38884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:49:42,581 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 04:49:53,149 INFO [train.py:904] (0/8) Epoch 4, batch 8450, loss[loss=0.2069, simple_loss=0.2928, pruned_loss=0.06054, over 16649.00 frames. ], tot_loss[loss=0.2384, simple_loss=0.3188, pruned_loss=0.07899, over 3101405.02 frames. ], batch size: 134, lr: 1.56e-02, grad_scale: 4.0 2023-04-28 04:50:19,931 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-04-28 04:50:34,197 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-28 04:51:13,838 INFO [optim.py:368] (0/8) 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 04:51:13,855 INFO [train.py:904] (0/8) Epoch 4, batch 8500, loss[loss=0.1944, simple_loss=0.2895, pruned_loss=0.04969, over 16575.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3136, pruned_loss=0.07561, over 3081869.97 frames. ], batch size: 68, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:51:57,065 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.16 vs. limit=5.0 2023-04-28 04:52:18,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6613, 1.2161, 1.4287, 1.6074, 1.7824, 1.7840, 1.4664, 1.8302], device='cuda:0'), covar=tensor([0.0100, 0.0155, 0.0110, 0.0117, 0.0097, 0.0077, 0.0162, 0.0047], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0130, 0.0118, 0.0113, 0.0116, 0.0081, 0.0130, 0.0074], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 04:52:34,873 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2404, 4.1587, 4.0363, 3.9645, 3.6666, 4.1232, 3.9532, 3.8196], device='cuda:0'), covar=tensor([0.0393, 0.0321, 0.0193, 0.0154, 0.0679, 0.0282, 0.0367, 0.0373], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0153, 0.0188, 0.0154, 0.0211, 0.0180, 0.0139, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 04:52:39,875 INFO [train.py:904] (0/8) Epoch 4, batch 8550, loss[loss=0.2511, simple_loss=0.3157, pruned_loss=0.09326, over 12017.00 frames. ], tot_loss[loss=0.2299, simple_loss=0.3107, pruned_loss=0.07453, over 3040114.75 frames. ], batch size: 248, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:53:09,990 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:54:04,165 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6325, 3.9729, 3.9228, 1.8351, 4.1613, 4.1906, 3.2549, 3.1167], device='cuda:0'), covar=tensor([0.0629, 0.0099, 0.0166, 0.1207, 0.0043, 0.0037, 0.0212, 0.0344], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0081, 0.0142, 0.0069, 0.0073, 0.0110, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 04:54:21,293 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.792e+02 3.391e+02 3.909e+02 5.367e+02 1.114e+03, threshold=7.819e+02, percent-clipped=4.0 2023-04-28 04:54:21,313 INFO [train.py:904] (0/8) Epoch 4, batch 8600, loss[loss=0.2492, simple_loss=0.3258, pruned_loss=0.08626, over 16839.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3116, pruned_loss=0.07385, over 3045030.78 frames. ], batch size: 124, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:54:50,962 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:55:58,618 INFO [train.py:904] (0/8) Epoch 4, batch 8650, loss[loss=0.2044, simple_loss=0.2986, pruned_loss=0.05507, over 16864.00 frames. ], tot_loss[loss=0.2258, simple_loss=0.3086, pruned_loss=0.07146, over 3032399.01 frames. ], batch size: 116, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:57:44,933 INFO [optim.py:368] (0/8) 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 04:57:44,953 INFO [train.py:904] (0/8) Epoch 4, batch 8700, loss[loss=0.2077, simple_loss=0.2819, pruned_loss=0.06675, over 12236.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3049, pruned_loss=0.06942, over 3030608.09 frames. ], batch size: 248, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 04:57:46,189 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2180, 1.3117, 1.7113, 2.1876, 2.2891, 2.3038, 1.4306, 2.2269], device='cuda:0'), covar=tensor([0.0069, 0.0223, 0.0127, 0.0115, 0.0085, 0.0064, 0.0209, 0.0053], device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0132, 0.0117, 0.0112, 0.0117, 0.0080, 0.0129, 0.0072], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 04:57:52,254 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-28 04:58:36,861 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39179.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:49,737 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39185.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 04:58:55,217 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 04:59:20,258 INFO [train.py:904] (0/8) Epoch 4, batch 8750, loss[loss=0.2417, simple_loss=0.3334, pruned_loss=0.075, over 16786.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3055, pruned_loss=0.06925, over 3038655.92 frames. ], batch size: 124, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:01:04,980 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:01:10,664 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2986, 5.6729, 5.4406, 5.3894, 5.0040, 4.7899, 5.1791, 5.6308], device='cuda:0'), covar=tensor([0.0568, 0.0556, 0.0558, 0.0368, 0.0553, 0.0539, 0.0535, 0.0674], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0409, 0.0353, 0.0267, 0.0267, 0.0277, 0.0333, 0.0297], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:01:14,445 INFO [train.py:904] (0/8) Epoch 4, batch 8800, loss[loss=0.2231, simple_loss=0.3203, pruned_loss=0.06291, over 16875.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3034, pruned_loss=0.06785, over 3053746.37 frames. ], batch size: 96, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:01:15,991 INFO [optim.py:368] (0/8) 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 05:02:09,752 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 05:02:23,551 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39284.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:02:58,300 INFO [train.py:904] (0/8) Epoch 4, batch 8850, loss[loss=0.2343, simple_loss=0.3201, pruned_loss=0.07427, over 15304.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3047, pruned_loss=0.06693, over 3029327.21 frames. ], batch size: 191, lr: 1.55e-02, grad_scale: 4.0 2023-04-28 05:04:32,384 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:04:44,188 INFO [train.py:904] (0/8) Epoch 4, batch 8900, loss[loss=0.2223, simple_loss=0.308, pruned_loss=0.06836, over 16976.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3046, pruned_loss=0.06638, over 3016871.92 frames. ], batch size: 109, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:04:49,526 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.906e+02 3.441e+02 4.038e+02 4.851e+02 9.886e+02, threshold=8.076e+02, percent-clipped=2.0 2023-04-28 05:06:47,810 INFO [train.py:904] (0/8) Epoch 4, batch 8950, loss[loss=0.1966, simple_loss=0.2874, pruned_loss=0.05294, over 16350.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3048, pruned_loss=0.06642, over 3051064.26 frames. ], batch size: 146, lr: 1.55e-02, grad_scale: 2.0 2023-04-28 05:08:35,749 INFO [train.py:904] (0/8) Epoch 4, batch 9000, loss[loss=0.2121, simple_loss=0.2888, pruned_loss=0.06766, over 12128.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.3016, pruned_loss=0.06484, over 3063298.24 frames. ], batch size: 247, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:08:35,750 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 05:08:45,784 INFO [train.py:938] (0/8) Epoch 4, validation: loss=0.1802, simple_loss=0.283, pruned_loss=0.0387, over 944034.00 frames. 2023-04-28 05:08:45,785 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 05:08:49,857 INFO [optim.py:368] (0/8) 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 05:09:11,284 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:09:43,528 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:10:17,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8658, 3.9198, 4.3716, 4.3269, 4.3509, 3.9729, 4.0769, 3.9148], device='cuda:0'), covar=tensor([0.0218, 0.0398, 0.0261, 0.0329, 0.0292, 0.0261, 0.0535, 0.0320], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0194, 0.0202, 0.0200, 0.0240, 0.0214, 0.0296, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-28 05:10:26,778 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9507, 2.1007, 2.2238, 3.2685, 1.9665, 2.7777, 2.2204, 1.8905], device='cuda:0'), covar=tensor([0.0461, 0.1566, 0.0749, 0.0322, 0.2469, 0.0590, 0.1573, 0.2090], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0282, 0.0230, 0.0287, 0.0342, 0.0254, 0.0257, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:10:29,890 INFO [train.py:904] (0/8) Epoch 4, batch 9050, loss[loss=0.209, simple_loss=0.2886, pruned_loss=0.06465, over 16344.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.3017, pruned_loss=0.06535, over 3041140.75 frames. ], batch size: 146, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:11:17,301 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:21,900 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:11:54,721 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39541.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:12:14,632 INFO [train.py:904] (0/8) Epoch 4, batch 9100, loss[loss=0.2049, simple_loss=0.3004, pruned_loss=0.05464, over 16274.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.3005, pruned_loss=0.06512, over 3056167.56 frames. ], batch size: 165, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:12:18,750 INFO [optim.py:368] (0/8) 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 05:13:30,733 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-28 05:14:08,099 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 05:14:15,382 INFO [train.py:904] (0/8) Epoch 4, batch 9150, loss[loss=0.1971, simple_loss=0.2859, pruned_loss=0.05417, over 16323.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.3006, pruned_loss=0.06436, over 3049719.91 frames. ], batch size: 166, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:14:19,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7353, 2.4140, 2.4950, 2.3629, 3.1411, 2.7952, 3.6936, 3.3475], device='cuda:0'), covar=tensor([0.0014, 0.0191, 0.0179, 0.0216, 0.0091, 0.0155, 0.0051, 0.0087], device='cuda:0'), in_proj_covar=tensor([0.0065, 0.0144, 0.0143, 0.0141, 0.0135, 0.0145, 0.0100, 0.0117], device='cuda:0'), out_proj_covar=tensor([8.1073e-05, 1.8456e-04, 1.7930e-04, 1.7709e-04, 1.7367e-04, 1.8709e-04, 1.2193e-04, 1.5032e-04], device='cuda:0') 2023-04-28 05:15:40,830 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39639.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:15:42,721 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39640.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:16:00,931 INFO [train.py:904] (0/8) Epoch 4, batch 9200, loss[loss=0.2044, simple_loss=0.2922, pruned_loss=0.05829, over 16257.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2956, pruned_loss=0.06282, over 3062964.31 frames. ], batch size: 165, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:16:04,315 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.381e+02 3.414e+02 4.371e+02 6.281e+02 1.474e+03, threshold=8.741e+02, percent-clipped=12.0 2023-04-28 05:16:53,689 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7626, 1.4104, 1.9076, 2.4311, 2.4902, 2.6726, 1.4887, 2.5956], device='cuda:0'), covar=tensor([0.0053, 0.0263, 0.0162, 0.0126, 0.0111, 0.0082, 0.0224, 0.0055], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0132, 0.0118, 0.0113, 0.0116, 0.0081, 0.0132, 0.0073], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 05:17:14,252 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8902, 4.1597, 3.9679, 4.0461, 3.7077, 3.7254, 3.8493, 4.0750], device='cuda:0'), covar=tensor([0.0653, 0.0784, 0.0773, 0.0427, 0.0601, 0.1434, 0.0623, 0.0924], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0418, 0.0350, 0.0265, 0.0267, 0.0284, 0.0333, 0.0302], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:17:25,223 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3816, 3.3018, 2.7382, 2.0561, 2.1694, 2.0673, 3.3414, 3.2400], device='cuda:0'), covar=tensor([0.1933, 0.0649, 0.1003, 0.1516, 0.1963, 0.1382, 0.0369, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0234, 0.0249, 0.0226, 0.0244, 0.0189, 0.0222, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:17:34,393 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39700.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:17:35,647 INFO [train.py:904] (0/8) Epoch 4, batch 9250, loss[loss=0.2058, simple_loss=0.2979, pruned_loss=0.05685, over 16428.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2959, pruned_loss=0.06331, over 3058979.30 frames. ], batch size: 146, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:17:59,475 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 05:17:59,556 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 05:19:26,139 INFO [train.py:904] (0/8) Epoch 4, batch 9300, loss[loss=0.2036, simple_loss=0.2928, pruned_loss=0.05721, over 17018.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2943, pruned_loss=0.06246, over 3054066.23 frames. ], batch size: 41, lr: 1.54e-02, grad_scale: 4.0 2023-04-28 05:19:30,023 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.165e+02 3.034e+02 3.554e+02 4.321e+02 7.893e+02, threshold=7.107e+02, percent-clipped=0.0 2023-04-28 05:20:05,544 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39767.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:11,033 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8399, 4.1560, 3.9444, 4.0022, 3.5763, 3.7219, 3.8528, 4.1335], device='cuda:0'), covar=tensor([0.0806, 0.0752, 0.0917, 0.0419, 0.0610, 0.1222, 0.0559, 0.0825], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0415, 0.0347, 0.0263, 0.0262, 0.0278, 0.0329, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:21:11,788 INFO [train.py:904] (0/8) Epoch 4, batch 9350, loss[loss=0.2184, simple_loss=0.2902, pruned_loss=0.07329, over 12329.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.295, pruned_loss=0.06281, over 3050327.67 frames. ], batch size: 250, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:21:41,235 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=39815.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:21:48,580 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:05,612 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39828.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:21,141 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 05:22:30,168 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39841.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:22:48,888 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 05:22:51,024 INFO [train.py:904] (0/8) Epoch 4, batch 9400, loss[loss=0.1867, simple_loss=0.2624, pruned_loss=0.05552, over 12639.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2944, pruned_loss=0.06226, over 3049171.86 frames. ], batch size: 247, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:22:57,593 INFO [optim.py:368] (0/8) 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 05:23:41,259 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=39876.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:09,634 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:24:32,548 INFO [train.py:904] (0/8) Epoch 4, batch 9450, loss[loss=0.2029, simple_loss=0.2849, pruned_loss=0.0604, over 12449.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.297, pruned_loss=0.06224, over 3065758.42 frames. ], batch size: 248, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:25:09,064 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 05:25:27,178 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 05:25:53,611 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=39940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:26:13,860 INFO [train.py:904] (0/8) Epoch 4, batch 9500, loss[loss=0.1936, simple_loss=0.2866, pruned_loss=0.05029, over 16906.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2955, pruned_loss=0.06142, over 3072782.78 frames. ], batch size: 116, lr: 1.54e-02, grad_scale: 2.0 2023-04-28 05:26:21,221 INFO [optim.py:368] (0/8) 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 05:27:07,584 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 05:27:29,670 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 05:27:31,421 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=39988.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:27:48,802 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=39995.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 05:28:00,432 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-40000.pt 2023-04-28 05:28:04,409 INFO [train.py:904] (0/8) Epoch 4, batch 9550, loss[loss=0.2211, simple_loss=0.3001, pruned_loss=0.07103, over 12716.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.295, pruned_loss=0.06188, over 3050430.46 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 2.0 2023-04-28 05:29:36,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9082, 3.7128, 3.8768, 4.1109, 4.1666, 3.7499, 4.1911, 4.1904], device='cuda:0'), covar=tensor([0.0705, 0.0624, 0.1017, 0.0446, 0.0415, 0.0966, 0.0409, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0393, 0.0500, 0.0407, 0.0304, 0.0293, 0.0325, 0.0330], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:29:45,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9946, 2.7888, 2.6299, 1.7526, 2.8865, 2.8946, 2.5314, 2.4164], device='cuda:0'), covar=tensor([0.0715, 0.0140, 0.0185, 0.0980, 0.0091, 0.0101, 0.0361, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0081, 0.0075, 0.0136, 0.0067, 0.0072, 0.0108, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 05:29:46,547 INFO [train.py:904] (0/8) Epoch 4, batch 9600, loss[loss=0.2174, simple_loss=0.292, pruned_loss=0.07136, over 12435.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2969, pruned_loss=0.06332, over 3042521.39 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:29:52,056 INFO [optim.py:368] (0/8) 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 05:31:33,071 INFO [train.py:904] (0/8) Epoch 4, batch 9650, loss[loss=0.2269, simple_loss=0.2998, pruned_loss=0.07702, over 12408.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2996, pruned_loss=0.06422, over 3049922.60 frames. ], batch size: 248, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:31:48,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7924, 3.7330, 4.2375, 4.2230, 4.1909, 3.8589, 3.9434, 3.7719], device='cuda:0'), covar=tensor([0.0233, 0.0373, 0.0278, 0.0338, 0.0335, 0.0285, 0.0585, 0.0363], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0189, 0.0196, 0.0202, 0.0229, 0.0207, 0.0291, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-28 05:32:17,685 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:32:25,201 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:33:21,182 INFO [train.py:904] (0/8) Epoch 4, batch 9700, loss[loss=0.1998, simple_loss=0.2878, pruned_loss=0.05587, over 16934.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2982, pruned_loss=0.0634, over 3053512.38 frames. ], batch size: 109, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:33:26,547 INFO [optim.py:368] (0/8) 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 05:33:45,974 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2722, 4.5997, 4.3669, 4.3946, 4.0738, 4.0542, 4.1355, 4.6296], device='cuda:0'), covar=tensor([0.0650, 0.0720, 0.0821, 0.0437, 0.0593, 0.0913, 0.0678, 0.0664], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0418, 0.0352, 0.0267, 0.0269, 0.0278, 0.0333, 0.0300], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:33:53,100 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:00,988 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:34:12,512 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8716, 4.1536, 3.9312, 4.0167, 3.6797, 3.6759, 3.8172, 4.0978], device='cuda:0'), covar=tensor([0.0616, 0.0682, 0.0784, 0.0412, 0.0520, 0.1285, 0.0600, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0415, 0.0350, 0.0265, 0.0268, 0.0277, 0.0331, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:34:54,804 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:35:03,575 INFO [train.py:904] (0/8) Epoch 4, batch 9750, loss[loss=0.1954, simple_loss=0.2721, pruned_loss=0.05938, over 12560.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2976, pruned_loss=0.064, over 3057918.51 frames. ], batch size: 250, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:35:38,312 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:36:45,117 INFO [train.py:904] (0/8) Epoch 4, batch 9800, loss[loss=0.222, simple_loss=0.3206, pruned_loss=0.06168, over 15412.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2968, pruned_loss=0.06252, over 3054779.47 frames. ], batch size: 191, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:36:51,070 INFO [optim.py:368] (0/8) 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 05:36:56,126 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40256.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:24,833 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:37:37,831 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40279.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:38:17,358 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40295.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 05:38:29,692 INFO [train.py:904] (0/8) Epoch 4, batch 9850, loss[loss=0.2082, simple_loss=0.2971, pruned_loss=0.05967, over 16952.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2977, pruned_loss=0.06205, over 3058238.84 frames. ], batch size: 109, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:39:12,025 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 05:39:30,939 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40330.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:39:39,162 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:40:05,816 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40343.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 05:40:21,700 INFO [train.py:904] (0/8) Epoch 4, batch 9900, loss[loss=0.2097, simple_loss=0.2919, pruned_loss=0.06371, over 12631.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.298, pruned_loss=0.06197, over 3052397.90 frames. ], batch size: 247, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:40:27,911 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 3.210e+02 3.891e+02 5.069e+02 8.589e+02, threshold=7.781e+02, percent-clipped=1.0 2023-04-28 05:41:24,187 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4588, 1.3431, 1.6517, 2.2113, 2.3042, 2.4208, 1.4883, 2.4917], device='cuda:0'), covar=tensor([0.0066, 0.0215, 0.0154, 0.0132, 0.0094, 0.0078, 0.0201, 0.0051], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0131, 0.0119, 0.0113, 0.0112, 0.0080, 0.0131, 0.0071], device='cuda:0'), out_proj_covar=tensor([1.4794e-04, 1.8655e-04, 1.7308e-04, 1.6315e-04, 1.5992e-04, 1.1146e-04, 1.8527e-04, 9.9404e-05], device='cuda:0') 2023-04-28 05:41:41,220 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1160, 4.9958, 4.9065, 4.3833, 4.7566, 1.7751, 4.6219, 4.8511], device='cuda:0'), covar=tensor([0.0039, 0.0035, 0.0052, 0.0187, 0.0050, 0.1562, 0.0057, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0066, 0.0102, 0.0102, 0.0078, 0.0132, 0.0092, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:41:56,849 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:42:18,090 INFO [train.py:904] (0/8) Epoch 4, batch 9950, loss[loss=0.2595, simple_loss=0.3294, pruned_loss=0.09476, over 12272.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2996, pruned_loss=0.06243, over 3046127.21 frames. ], batch size: 246, lr: 1.53e-02, grad_scale: 4.0 2023-04-28 05:43:13,940 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:44:20,292 INFO [train.py:904] (0/8) Epoch 4, batch 10000, loss[loss=0.1671, simple_loss=0.2563, pruned_loss=0.03898, over 17255.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2978, pruned_loss=0.06162, over 3067490.75 frames. ], batch size: 52, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:44:26,644 INFO [optim.py:368] (0/8) 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 05:44:30,516 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 05:45:01,305 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:01,349 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40471.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:45:06,463 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1509, 2.0499, 2.0758, 3.4034, 1.6416, 2.7629, 1.9687, 1.8070], device='cuda:0'), covar=tensor([0.0552, 0.1776, 0.0952, 0.0411, 0.3158, 0.0828, 0.1930, 0.2556], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0282, 0.0230, 0.0287, 0.0342, 0.0256, 0.0258, 0.0327], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:45:46,750 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:46:02,928 INFO [train.py:904] (0/8) Epoch 4, batch 10050, loss[loss=0.2014, simple_loss=0.292, pruned_loss=0.05538, over 16472.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2973, pruned_loss=0.06119, over 3057084.34 frames. ], batch size: 68, lr: 1.53e-02, grad_scale: 8.0 2023-04-28 05:46:38,834 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40519.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:38,807 INFO [train.py:904] (0/8) Epoch 4, batch 10100, loss[loss=0.1908, simple_loss=0.2702, pruned_loss=0.05572, over 12272.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2983, pruned_loss=0.06194, over 3066426.27 frames. ], batch size: 248, lr: 1.52e-02, grad_scale: 8.0 2023-04-28 05:47:39,312 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40553.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:47:42,821 INFO [optim.py:368] (0/8) 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 05:48:27,656 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40574.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:48:59,880 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-4.pt 2023-04-28 05:49:25,178 INFO [train.py:904] (0/8) Epoch 5, batch 0, loss[loss=0.4012, simple_loss=0.4074, pruned_loss=0.1975, over 16901.00 frames. ], tot_loss[loss=0.4012, simple_loss=0.4074, pruned_loss=0.1975, over 16901.00 frames. ], batch size: 109, lr: 1.42e-02, grad_scale: 8.0 2023-04-28 05:49:25,179 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 05:49:32,556 INFO [train.py:938] (0/8) Epoch 5, validation: loss=0.1789, simple_loss=0.2817, pruned_loss=0.03802, over 944034.00 frames. 2023-04-28 05:49:32,557 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 05:49:38,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7593, 4.5872, 4.1098, 2.1217, 3.1956, 2.7659, 3.9461, 4.1595], device='cuda:0'), covar=tensor([0.0190, 0.0408, 0.0379, 0.1380, 0.0623, 0.0806, 0.0521, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0111, 0.0154, 0.0142, 0.0132, 0.0126, 0.0136, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-28 05:49:53,230 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 05:50:11,893 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40628.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:42,814 INFO [train.py:904] (0/8) Epoch 5, batch 50, loss[loss=0.2036, simple_loss=0.2852, pruned_loss=0.06102, over 17187.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3193, pruned_loss=0.09091, over 755706.62 frames. ], batch size: 46, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:50:43,186 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:50:49,839 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.847e+02 4.852e+02 6.011e+02 1.299e+03, threshold=9.705e+02, percent-clipped=3.0 2023-04-28 05:50:51,545 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2243, 2.2724, 2.4250, 4.3523, 1.7787, 3.1535, 2.0959, 2.1261], device='cuda:0'), covar=tensor([0.0459, 0.2020, 0.0974, 0.0354, 0.3574, 0.0922, 0.1997, 0.3022], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0290, 0.0234, 0.0292, 0.0348, 0.0260, 0.0260, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:51:29,212 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=40686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:51:48,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4825, 4.7419, 4.6833, 4.8352, 4.7491, 5.2679, 5.0112, 4.7076], device='cuda:0'), covar=tensor([0.1057, 0.1516, 0.1363, 0.1638, 0.2577, 0.0999, 0.1043, 0.2013], device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0364, 0.0354, 0.0319, 0.0414, 0.0376, 0.0290, 0.0419], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 05:51:50,913 INFO [train.py:904] (0/8) Epoch 5, batch 100, loss[loss=0.2121, simple_loss=0.2986, pruned_loss=0.06286, over 17139.00 frames. ], tot_loss[loss=0.2411, simple_loss=0.3126, pruned_loss=0.08482, over 1333077.77 frames. ], batch size: 48, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:52:06,722 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:52:59,310 INFO [train.py:904] (0/8) Epoch 5, batch 150, loss[loss=0.2283, simple_loss=0.3062, pruned_loss=0.07517, over 16387.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.308, pruned_loss=0.08051, over 1771178.12 frames. ], batch size: 68, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:53:08,169 INFO [optim.py:368] (0/8) 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 05:54:09,234 INFO [train.py:904] (0/8) Epoch 5, batch 200, loss[loss=0.2991, simple_loss=0.335, pruned_loss=0.1316, over 16850.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.3078, pruned_loss=0.08068, over 2125038.67 frames. ], batch size: 96, lr: 1.42e-02, grad_scale: 1.0 2023-04-28 05:54:59,414 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-28 05:55:01,371 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8527, 2.9098, 3.5383, 2.0124, 3.2367, 3.3953, 3.3481, 1.8320], device='cuda:0'), covar=tensor([0.0292, 0.0111, 0.0024, 0.0231, 0.0049, 0.0043, 0.0053, 0.0276], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0058, 0.0060, 0.0114, 0.0058, 0.0065, 0.0062, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 05:55:03,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3452, 5.6754, 5.3921, 5.4715, 4.9693, 4.7960, 5.1383, 5.7882], device='cuda:0'), covar=tensor([0.0609, 0.0656, 0.0813, 0.0428, 0.0643, 0.0571, 0.0626, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0459, 0.0396, 0.0301, 0.0298, 0.0307, 0.0373, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 05:55:13,932 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 5, batch 250, loss[loss=0.2577, simple_loss=0.3387, pruned_loss=0.08832, over 17044.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3073, pruned_loss=0.08145, over 2393999.71 frames. ], batch size: 55, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:55:18,082 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:25,561 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.087e+02 3.707e+02 4.511e+02 5.174e+02 8.449e+02, threshold=9.022e+02, percent-clipped=1.0 2023-04-28 05:55:37,414 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40865.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:55:50,427 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40874.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:24,971 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:26,425 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:56:27,696 INFO [train.py:904] (0/8) Epoch 5, batch 300, loss[loss=0.2379, simple_loss=0.3076, pruned_loss=0.08409, over 16626.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.304, pruned_loss=0.07991, over 2600105.77 frames. ], batch size: 68, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:56:58,127 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40922.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:04,466 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40926.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:07,233 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40928.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:57:39,679 INFO [train.py:904] (0/8) Epoch 5, batch 350, loss[loss=0.2334, simple_loss=0.3142, pruned_loss=0.07633, over 17130.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3008, pruned_loss=0.07773, over 2758687.91 frames. ], batch size: 49, lr: 1.41e-02, grad_scale: 1.0 2023-04-28 05:57:48,105 INFO [optim.py:368] (0/8) 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 05:57:54,163 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=40961.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 05:58:04,054 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 2023-04-28 05:58:14,882 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=40976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:28,098 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=40986.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:44,225 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=40997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:58:48,869 INFO [train.py:904] (0/8) Epoch 5, batch 400, loss[loss=0.1866, simple_loss=0.2704, pruned_loss=0.05136, over 17180.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.2991, pruned_loss=0.07697, over 2876470.63 frames. ], batch size: 46, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 05:58:49,438 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3448, 2.4296, 1.8037, 2.2169, 2.7380, 2.5387, 3.4399, 3.0341], device='cuda:0'), covar=tensor([0.0027, 0.0185, 0.0253, 0.0195, 0.0117, 0.0178, 0.0069, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0148, 0.0149, 0.0145, 0.0141, 0.0148, 0.0112, 0.0126], device='cuda:0'), out_proj_covar=tensor([8.9650e-05, 1.8673e-04, 1.8504e-04, 1.7939e-04, 1.7912e-04, 1.8890e-04, 1.3641e-04, 1.5915e-04], device='cuda:0') 2023-04-28 05:58:58,459 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:20,558 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 05:59:27,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7964, 3.3249, 2.6489, 4.9428, 4.4527, 4.2437, 1.6169, 3.4200], device='cuda:0'), covar=tensor([0.1277, 0.0513, 0.1072, 0.0066, 0.0225, 0.0264, 0.1314, 0.0586], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0138, 0.0163, 0.0078, 0.0156, 0.0163, 0.0154, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 05:59:36,592 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41034.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 05:59:51,058 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41045.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:00:00,537 INFO [train.py:904] (0/8) Epoch 5, batch 450, loss[loss=0.1892, simple_loss=0.27, pruned_loss=0.05414, over 16415.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.2972, pruned_loss=0.07626, over 2976472.55 frames. ], batch size: 75, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:00:09,413 INFO [optim.py:368] (0/8) 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 06:00:11,143 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:10,614 INFO [train.py:904] (0/8) Epoch 5, batch 500, loss[loss=0.2097, simple_loss=0.293, pruned_loss=0.06322, over 16732.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.2951, pruned_loss=0.07491, over 3055359.59 frames. ], batch size: 62, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:01:12,246 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:01:16,732 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41106.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:02:14,440 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:02:19,025 INFO [train.py:904] (0/8) Epoch 5, batch 550, loss[loss=0.199, simple_loss=0.274, pruned_loss=0.06205, over 16995.00 frames. ], tot_loss[loss=0.221, simple_loss=0.2938, pruned_loss=0.07405, over 3113788.30 frames. ], batch size: 41, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:02:27,245 INFO [optim.py:368] (0/8) 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 06:02:35,982 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41163.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:13,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8829, 4.2965, 4.2843, 1.5866, 4.6367, 4.7291, 3.5440, 3.6195], device='cuda:0'), covar=tensor([0.0626, 0.0112, 0.0243, 0.1323, 0.0042, 0.0046, 0.0279, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0087, 0.0083, 0.0145, 0.0073, 0.0078, 0.0115, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 06:03:21,441 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41196.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:03:27,690 INFO [train.py:904] (0/8) Epoch 5, batch 600, loss[loss=0.215, simple_loss=0.2937, pruned_loss=0.06818, over 16641.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.293, pruned_loss=0.07374, over 3163644.79 frames. ], batch size: 62, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:03:54,893 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:04:33,950 INFO [train.py:904] (0/8) Epoch 5, batch 650, loss[loss=0.1892, simple_loss=0.2735, pruned_loss=0.05246, over 16984.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2917, pruned_loss=0.07357, over 3190470.07 frames. ], batch size: 41, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:04:40,907 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41256.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:04:42,452 INFO [optim.py:368] (0/8) 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 06:05:01,389 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4599, 4.4585, 4.5567, 4.5621, 4.4344, 5.0247, 4.7063, 4.3199], device='cuda:0'), covar=tensor([0.1124, 0.1556, 0.1331, 0.1770, 0.2719, 0.1056, 0.1024, 0.2418], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0392, 0.0377, 0.0335, 0.0446, 0.0405, 0.0309, 0.0447], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:05:39,910 INFO [train.py:904] (0/8) Epoch 5, batch 700, loss[loss=0.2187, simple_loss=0.282, pruned_loss=0.07765, over 16889.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2904, pruned_loss=0.07216, over 3226482.63 frames. ], batch size: 116, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:05:49,033 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41307.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:05:52,187 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3966, 4.3448, 4.3281, 4.3818, 4.3044, 4.9003, 4.5765, 4.2594], device='cuda:0'), covar=tensor([0.1186, 0.1553, 0.1459, 0.1864, 0.2719, 0.1022, 0.1137, 0.2513], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0394, 0.0379, 0.0336, 0.0447, 0.0410, 0.0308, 0.0448], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:05:57,718 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3840, 3.2440, 2.6064, 2.1932, 2.3209, 2.0732, 3.1511, 3.2987], device='cuda:0'), covar=tensor([0.2067, 0.0669, 0.1155, 0.1454, 0.2021, 0.1510, 0.0485, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0252, 0.0265, 0.0238, 0.0291, 0.0201, 0.0236, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:06:03,501 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6558, 4.2978, 3.9707, 1.9814, 3.2800, 2.3669, 4.0248, 4.0709], device='cuda:0'), covar=tensor([0.0240, 0.0485, 0.0433, 0.1586, 0.0653, 0.0983, 0.0538, 0.0750], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0130, 0.0156, 0.0143, 0.0135, 0.0127, 0.0141, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 06:06:49,386 INFO [train.py:904] (0/8) Epoch 5, batch 750, loss[loss=0.238, simple_loss=0.2996, pruned_loss=0.08815, over 16865.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2909, pruned_loss=0.07209, over 3234403.66 frames. ], batch size: 96, lr: 1.41e-02, grad_scale: 2.0 2023-04-28 06:06:52,764 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:55,091 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:06:57,968 INFO [optim.py:368] (0/8) 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 06:07:25,442 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4561, 2.7197, 2.5277, 4.8242, 2.1142, 3.6865, 2.6761, 2.8168], device='cuda:0'), covar=tensor([0.0438, 0.1728, 0.0981, 0.0240, 0.2868, 0.0867, 0.1676, 0.2421], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0308, 0.0248, 0.0307, 0.0359, 0.0288, 0.0273, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:07:58,824 INFO [train.py:904] (0/8) Epoch 5, batch 800, loss[loss=0.2159, simple_loss=0.3035, pruned_loss=0.06419, over 17051.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2899, pruned_loss=0.07176, over 3242479.47 frames. ], batch size: 50, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:07:59,168 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:09:06,136 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3755, 4.3823, 4.2523, 4.1865, 3.8708, 4.2722, 4.1615, 4.0179], device='cuda:0'), covar=tensor([0.0485, 0.0269, 0.0245, 0.0170, 0.0800, 0.0307, 0.0410, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0178, 0.0225, 0.0191, 0.0256, 0.0213, 0.0162, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:09:08,623 INFO [train.py:904] (0/8) Epoch 5, batch 850, loss[loss=0.1854, simple_loss=0.2698, pruned_loss=0.0505, over 16852.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2889, pruned_loss=0.07069, over 3264536.75 frames. ], batch size: 42, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:09:16,418 INFO [optim.py:368] (0/8) 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 06:09:17,887 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=41458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:10:16,021 INFO [train.py:904] (0/8) Epoch 5, batch 900, loss[loss=0.213, simple_loss=0.3005, pruned_loss=0.0627, over 17040.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2879, pruned_loss=0.06925, over 3285448.27 frames. ], batch size: 50, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:10:44,991 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:11:27,568 INFO [train.py:904] (0/8) Epoch 5, batch 950, loss[loss=0.2589, simple_loss=0.3117, pruned_loss=0.103, over 16852.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2878, pruned_loss=0.06915, over 3292352.94 frames. ], batch size: 109, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:11:34,555 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41556.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:11:35,290 INFO [optim.py:368] (0/8) 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 06:11:52,921 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:12:24,892 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7613, 2.3377, 2.3468, 4.2975, 1.9272, 3.3202, 2.2370, 2.3974], device='cuda:0'), covar=tensor([0.0490, 0.1689, 0.0956, 0.0251, 0.2660, 0.0704, 0.1852, 0.1992], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0308, 0.0249, 0.0310, 0.0359, 0.0290, 0.0272, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:12:37,224 INFO [train.py:904] (0/8) Epoch 5, batch 1000, loss[loss=0.2262, simple_loss=0.3111, pruned_loss=0.0707, over 17062.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2865, pruned_loss=0.06852, over 3289749.28 frames. ], batch size: 55, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:12:41,515 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41604.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:39,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3238, 2.4712, 1.8485, 2.0747, 2.8564, 2.6660, 3.5045, 3.1856], device='cuda:0'), covar=tensor([0.0033, 0.0175, 0.0214, 0.0220, 0.0116, 0.0157, 0.0081, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0155, 0.0151, 0.0150, 0.0147, 0.0152, 0.0122, 0.0132], device='cuda:0'), out_proj_covar=tensor([9.9290e-05, 1.9613e-04, 1.8618e-04, 1.8375e-04, 1.8747e-04, 1.9293e-04, 1.4986e-04, 1.6711e-04], device='cuda:0') 2023-04-28 06:13:45,697 INFO [train.py:904] (0/8) Epoch 5, batch 1050, loss[loss=0.1922, simple_loss=0.2667, pruned_loss=0.05887, over 16990.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2862, pruned_loss=0.06897, over 3291204.91 frames. ], batch size: 41, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:13:48,303 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41653.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:50,308 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41654.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:13:54,772 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.880e+02 3.783e+02 4.446e+02 1.011e+03, threshold=7.566e+02, percent-clipped=5.0 2023-04-28 06:14:56,097 INFO [train.py:904] (0/8) Epoch 5, batch 1100, loss[loss=0.1872, simple_loss=0.2593, pruned_loss=0.0575, over 16761.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2862, pruned_loss=0.06877, over 3305998.30 frames. ], batch size: 39, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:14:56,431 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:14:56,535 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:15:15,128 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41715.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:15:37,028 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1922, 1.6159, 2.5025, 2.9101, 2.8282, 3.4409, 1.7101, 3.2931], device='cuda:0'), covar=tensor([0.0073, 0.0221, 0.0137, 0.0126, 0.0113, 0.0077, 0.0231, 0.0045], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0140, 0.0128, 0.0126, 0.0125, 0.0090, 0.0139, 0.0077], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 06:16:02,697 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:16:05,309 INFO [train.py:904] (0/8) Epoch 5, batch 1150, loss[loss=0.2145, simple_loss=0.2978, pruned_loss=0.06561, over 17102.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2852, pruned_loss=0.06784, over 3298916.29 frames. ], batch size: 49, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:16:12,836 INFO [optim.py:368] (0/8) 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 06:16:13,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9006, 2.7106, 2.7134, 1.6150, 2.8774, 2.8001, 2.5690, 2.3094], device='cuda:0'), covar=tensor([0.0878, 0.0190, 0.0215, 0.1149, 0.0102, 0.0124, 0.0385, 0.0562], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0087, 0.0085, 0.0143, 0.0071, 0.0080, 0.0115, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 06:16:14,967 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=41758.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:17:14,384 INFO [train.py:904] (0/8) Epoch 5, batch 1200, loss[loss=0.1961, simple_loss=0.2927, pruned_loss=0.04974, over 17093.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2843, pruned_loss=0.06776, over 3301563.94 frames. ], batch size: 48, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:17:21,143 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=41806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:18:10,535 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7506, 3.4060, 2.8716, 5.1093, 4.6701, 4.3383, 1.7775, 3.4131], device='cuda:0'), covar=tensor([0.1261, 0.0506, 0.0954, 0.0070, 0.0268, 0.0280, 0.1208, 0.0599], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0139, 0.0164, 0.0082, 0.0171, 0.0169, 0.0156, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 06:18:23,577 INFO [train.py:904] (0/8) Epoch 5, batch 1250, loss[loss=0.2075, simple_loss=0.2928, pruned_loss=0.0611, over 16698.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2847, pruned_loss=0.06831, over 3307859.11 frames. ], batch size: 57, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:18:31,528 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.380e+02 4.131e+02 4.913e+02 1.055e+03, threshold=8.263e+02, percent-clipped=6.0 2023-04-28 06:19:30,798 INFO [train.py:904] (0/8) Epoch 5, batch 1300, loss[loss=0.2207, simple_loss=0.2813, pruned_loss=0.08009, over 16765.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2842, pruned_loss=0.06825, over 3308375.49 frames. ], batch size: 83, lr: 1.40e-02, grad_scale: 8.0 2023-04-28 06:19:38,920 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:20:34,207 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2254, 4.8478, 4.9968, 5.3923, 5.5245, 4.7450, 5.5028, 5.5032], device='cuda:0'), covar=tensor([0.0728, 0.0811, 0.1395, 0.0429, 0.0351, 0.0488, 0.0318, 0.0299], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0503, 0.0640, 0.0507, 0.0379, 0.0372, 0.0403, 0.0418], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:20:38,054 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=41948.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:20:42,276 INFO [train.py:904] (0/8) Epoch 5, batch 1350, loss[loss=0.2162, simple_loss=0.2834, pruned_loss=0.07444, over 16855.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.284, pruned_loss=0.06731, over 3321904.59 frames. ], batch size: 102, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:20:51,209 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.407e+02 4.000e+02 4.882e+02 1.065e+03, threshold=8.000e+02, percent-clipped=1.0 2023-04-28 06:21:04,953 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=41967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:21:42,984 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 06:21:46,318 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 06:21:51,362 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-42000.pt 2023-04-28 06:21:55,643 INFO [train.py:904] (0/8) Epoch 5, batch 1400, loss[loss=0.2096, simple_loss=0.2865, pruned_loss=0.06636, over 16664.00 frames. ], tot_loss[loss=0.209, simple_loss=0.284, pruned_loss=0.06702, over 3329445.78 frames. ], batch size: 57, lr: 1.40e-02, grad_scale: 4.0 2023-04-28 06:22:08,306 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42009.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 06:22:09,284 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42010.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:22:21,526 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2531, 4.4408, 3.6646, 2.6435, 3.3181, 2.5175, 4.6163, 4.4037], device='cuda:0'), covar=tensor([0.1753, 0.0503, 0.0946, 0.1394, 0.2220, 0.1370, 0.0259, 0.0533], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0255, 0.0271, 0.0243, 0.0300, 0.0203, 0.0240, 0.0261], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:23:05,359 INFO [train.py:904] (0/8) Epoch 5, batch 1450, loss[loss=0.2128, simple_loss=0.2687, pruned_loss=0.07848, over 16250.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.283, pruned_loss=0.06608, over 3323837.62 frames. ], batch size: 165, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:23:15,559 INFO [optim.py:368] (0/8) 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 06:24:05,099 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6053, 4.1428, 4.3069, 1.9931, 4.6325, 4.4700, 3.2634, 3.3343], device='cuda:0'), covar=tensor([0.0724, 0.0116, 0.0248, 0.1136, 0.0049, 0.0075, 0.0314, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0085, 0.0084, 0.0141, 0.0069, 0.0079, 0.0113, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 06:24:14,181 INFO [train.py:904] (0/8) Epoch 5, batch 1500, loss[loss=0.2023, simple_loss=0.2713, pruned_loss=0.06665, over 16761.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2833, pruned_loss=0.06636, over 3323012.84 frames. ], batch size: 134, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:25:21,167 INFO [train.py:904] (0/8) Epoch 5, batch 1550, loss[loss=0.1849, simple_loss=0.2704, pruned_loss=0.04973, over 17180.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2842, pruned_loss=0.06751, over 3318049.42 frames. ], batch size: 44, lr: 1.39e-02, grad_scale: 4.0 2023-04-28 06:25:32,958 INFO [optim.py:368] (0/8) 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 06:25:50,970 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8386, 4.0304, 1.7993, 4.2225, 2.6927, 4.2186, 1.9364, 2.9272], device='cuda:0'), covar=tensor([0.0123, 0.0212, 0.1587, 0.0077, 0.0694, 0.0266, 0.1429, 0.0588], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0154, 0.0174, 0.0087, 0.0158, 0.0184, 0.0185, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 06:26:09,204 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8244, 4.4669, 4.6477, 3.3845, 4.1058, 4.7339, 4.3913, 2.9241], device='cuda:0'), covar=tensor([0.0273, 0.0019, 0.0028, 0.0175, 0.0036, 0.0023, 0.0021, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0060, 0.0060, 0.0113, 0.0060, 0.0066, 0.0062, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:26:15,632 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1851, 3.5610, 3.7206, 1.8423, 3.8780, 3.7913, 3.0324, 2.8426], device='cuda:0'), covar=tensor([0.0797, 0.0112, 0.0110, 0.1058, 0.0052, 0.0074, 0.0306, 0.0423], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0086, 0.0083, 0.0141, 0.0070, 0.0078, 0.0112, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 06:26:32,637 INFO [train.py:904] (0/8) Epoch 5, batch 1600, loss[loss=0.2032, simple_loss=0.2796, pruned_loss=0.06342, over 16890.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2867, pruned_loss=0.06907, over 3318678.50 frames. ], batch size: 96, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:27:07,019 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-04-28 06:27:27,668 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 06:27:39,766 INFO [train.py:904] (0/8) Epoch 5, batch 1650, loss[loss=0.2408, simple_loss=0.3006, pruned_loss=0.0905, over 16379.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2893, pruned_loss=0.07042, over 3313868.35 frames. ], batch size: 146, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:27:41,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4143, 4.2974, 3.6367, 1.8088, 2.7153, 2.3291, 3.6131, 3.8863], device='cuda:0'), covar=tensor([0.0275, 0.0599, 0.0568, 0.1684, 0.0880, 0.1056, 0.0721, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0132, 0.0153, 0.0139, 0.0133, 0.0125, 0.0140, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 06:27:49,824 INFO [optim.py:368] (0/8) 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 06:27:55,781 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42262.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:28:31,528 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0420, 5.4290, 5.4786, 5.4172, 5.3379, 5.9841, 5.5184, 5.2289], device='cuda:0'), covar=tensor([0.0764, 0.1731, 0.1408, 0.1681, 0.2790, 0.0882, 0.1012, 0.2231], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0404, 0.0383, 0.0344, 0.0458, 0.0409, 0.0312, 0.0455], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:28:49,299 INFO [train.py:904] (0/8) Epoch 5, batch 1700, loss[loss=0.2224, simple_loss=0.2968, pruned_loss=0.07396, over 16704.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2909, pruned_loss=0.07139, over 3313889.49 frames. ], batch size: 89, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:28:53,671 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42304.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:28:55,582 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1743, 1.6737, 2.2959, 2.9505, 2.8013, 3.3708, 1.6296, 3.1391], device='cuda:0'), covar=tensor([0.0064, 0.0205, 0.0137, 0.0097, 0.0095, 0.0058, 0.0204, 0.0052], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0138, 0.0126, 0.0126, 0.0123, 0.0089, 0.0138, 0.0078], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 06:29:01,270 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42310.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 06:29:56,502 INFO [train.py:904] (0/8) Epoch 5, batch 1750, loss[loss=0.1885, simple_loss=0.2672, pruned_loss=0.05491, over 16764.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2917, pruned_loss=0.07091, over 3318014.49 frames. ], batch size: 39, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:30:05,726 INFO [optim.py:368] (0/8) 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 06:30:05,951 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:30:35,457 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:31:06,108 INFO [train.py:904] (0/8) Epoch 5, batch 1800, loss[loss=0.2273, simple_loss=0.3165, pruned_loss=0.06904, over 17030.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2928, pruned_loss=0.0704, over 3327071.21 frames. ], batch size: 55, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:31:20,156 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3485, 5.7975, 5.4925, 5.6333, 5.1241, 4.9476, 5.2415, 5.8971], device='cuda:0'), covar=tensor([0.0684, 0.0651, 0.0806, 0.0399, 0.0600, 0.0550, 0.0577, 0.0641], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0502, 0.0414, 0.0316, 0.0313, 0.0318, 0.0393, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:31:24,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7968, 3.9062, 2.8259, 2.4066, 2.8924, 2.2462, 3.9425, 3.9527], device='cuda:0'), covar=tensor([0.1869, 0.0499, 0.1195, 0.1463, 0.2094, 0.1426, 0.0354, 0.0596], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0253, 0.0270, 0.0244, 0.0302, 0.0202, 0.0240, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:31:28,534 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6544, 4.6741, 5.1571, 5.1438, 5.1696, 4.5956, 4.6680, 4.4506], device='cuda:0'), covar=tensor([0.0249, 0.0357, 0.0305, 0.0381, 0.0416, 0.0282, 0.0871, 0.0338], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0237, 0.0243, 0.0242, 0.0292, 0.0258, 0.0369, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 06:32:01,379 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:32:16,843 INFO [train.py:904] (0/8) Epoch 5, batch 1850, loss[loss=0.2121, simple_loss=0.2869, pruned_loss=0.06866, over 16738.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.294, pruned_loss=0.07112, over 3319902.48 frames. ], batch size: 83, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:32:26,220 INFO [optim.py:368] (0/8) 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 06:32:34,633 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 06:33:25,579 INFO [train.py:904] (0/8) Epoch 5, batch 1900, loss[loss=0.256, simple_loss=0.3288, pruned_loss=0.09156, over 11959.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.293, pruned_loss=0.07062, over 3317213.27 frames. ], batch size: 246, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:33:32,605 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42506.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:33:40,226 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8510, 2.6385, 2.6670, 1.8229, 2.5475, 2.6197, 2.6349, 1.8495], device='cuda:0'), covar=tensor([0.0262, 0.0062, 0.0036, 0.0208, 0.0046, 0.0049, 0.0047, 0.0219], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0061, 0.0060, 0.0114, 0.0061, 0.0068, 0.0063, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:34:35,502 INFO [train.py:904] (0/8) Epoch 5, batch 1950, loss[loss=0.1845, simple_loss=0.272, pruned_loss=0.04852, over 17109.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.2938, pruned_loss=0.07026, over 3310144.86 frames. ], batch size: 47, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:34:47,179 INFO [optim.py:368] (0/8) 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 06:34:52,160 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:34:59,413 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:35:48,045 INFO [train.py:904] (0/8) Epoch 5, batch 2000, loss[loss=0.1983, simple_loss=0.2793, pruned_loss=0.05864, over 17253.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2946, pruned_loss=0.07059, over 3308651.92 frames. ], batch size: 45, lr: 1.39e-02, grad_scale: 8.0 2023-04-28 06:35:51,717 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=42604.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:36:00,221 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:55,064 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:36:57,550 INFO [train.py:904] (0/8) Epoch 5, batch 2050, loss[loss=0.2227, simple_loss=0.2864, pruned_loss=0.07944, over 16816.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2942, pruned_loss=0.07113, over 3313812.92 frames. ], batch size: 102, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:36:59,032 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=42652.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 06:37:06,890 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.791e+02 3.076e+02 3.576e+02 4.219e+02 7.856e+02, threshold=7.152e+02, percent-clipped=1.0 2023-04-28 06:37:44,214 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-04-28 06:38:05,327 INFO [train.py:904] (0/8) Epoch 5, batch 2100, loss[loss=0.2501, simple_loss=0.3176, pruned_loss=0.09131, over 16505.00 frames. ], tot_loss[loss=0.22, simple_loss=0.2957, pruned_loss=0.07214, over 3320983.83 frames. ], batch size: 75, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:38:11,157 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2589, 5.6175, 5.3338, 5.4594, 4.9080, 4.6697, 5.1234, 5.7292], device='cuda:0'), covar=tensor([0.0667, 0.0622, 0.0811, 0.0404, 0.0652, 0.0650, 0.0574, 0.0632], device='cuda:0'), in_proj_covar=tensor([0.0371, 0.0502, 0.0418, 0.0322, 0.0319, 0.0320, 0.0399, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:38:18,162 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42710.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:38:33,790 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8222, 3.3460, 2.7688, 4.6532, 4.2318, 4.2145, 1.6377, 3.1254], device='cuda:0'), covar=tensor([0.1244, 0.0399, 0.0947, 0.0057, 0.0241, 0.0263, 0.1251, 0.0596], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0139, 0.0165, 0.0084, 0.0174, 0.0168, 0.0156, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 06:38:52,291 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:38:58,825 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 06:39:15,223 INFO [train.py:904] (0/8) Epoch 5, batch 2150, loss[loss=0.339, simple_loss=0.3814, pruned_loss=0.1483, over 12046.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.2959, pruned_loss=0.07264, over 3323971.50 frames. ], batch size: 246, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:39:24,100 INFO [optim.py:368] (0/8) 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 06:39:36,706 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4954, 3.6376, 2.7466, 2.2475, 2.7054, 2.1163, 3.4968, 3.6691], device='cuda:0'), covar=tensor([0.2010, 0.0516, 0.1080, 0.1507, 0.1919, 0.1393, 0.0428, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0249, 0.0267, 0.0239, 0.0301, 0.0200, 0.0237, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:40:21,238 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3448, 3.6810, 3.8005, 1.8363, 3.9716, 3.9148, 3.2787, 2.9810], device='cuda:0'), covar=tensor([0.0717, 0.0107, 0.0123, 0.1077, 0.0055, 0.0078, 0.0255, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0086, 0.0084, 0.0140, 0.0070, 0.0079, 0.0114, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 06:40:23,582 INFO [train.py:904] (0/8) Epoch 5, batch 2200, loss[loss=0.2339, simple_loss=0.3157, pruned_loss=0.07601, over 17012.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.296, pruned_loss=0.07278, over 3333821.58 frames. ], batch size: 53, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:40:41,415 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3839, 2.1315, 2.2479, 3.8390, 1.8270, 2.9962, 2.0501, 2.2024], device='cuda:0'), covar=tensor([0.0573, 0.1809, 0.0918, 0.0350, 0.2554, 0.0879, 0.1866, 0.1933], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0312, 0.0251, 0.0308, 0.0358, 0.0300, 0.0276, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:41:16,484 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 06:41:21,725 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8641, 4.5231, 4.7741, 5.0600, 5.1799, 4.4876, 5.1561, 5.1085], device='cuda:0'), covar=tensor([0.0804, 0.0786, 0.1380, 0.0482, 0.0445, 0.0644, 0.0468, 0.0385], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0490, 0.0631, 0.0512, 0.0377, 0.0369, 0.0396, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 06:41:34,703 INFO [train.py:904] (0/8) Epoch 5, batch 2250, loss[loss=0.1885, simple_loss=0.2663, pruned_loss=0.05531, over 15857.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.2964, pruned_loss=0.0737, over 3328903.12 frames. ], batch size: 35, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:41:43,707 INFO [optim.py:368] (0/8) 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 06:41:45,364 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:49,467 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=42862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:41:57,046 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5053, 4.3992, 4.3761, 3.8990, 4.3679, 1.7708, 4.1940, 4.2407], device='cuda:0'), covar=tensor([0.0066, 0.0053, 0.0082, 0.0235, 0.0062, 0.1493, 0.0076, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0082, 0.0127, 0.0133, 0.0097, 0.0138, 0.0111, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:42:11,038 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4117, 4.1053, 3.4481, 2.1136, 2.8229, 2.3117, 3.6558, 3.8395], device='cuda:0'), covar=tensor([0.0222, 0.0441, 0.0561, 0.1438, 0.0715, 0.0999, 0.0482, 0.0657], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0133, 0.0152, 0.0139, 0.0133, 0.0126, 0.0139, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 06:42:14,365 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=42879.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:42:44,597 INFO [train.py:904] (0/8) Epoch 5, batch 2300, loss[loss=0.2328, simple_loss=0.2978, pruned_loss=0.08389, over 16844.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.2971, pruned_loss=0.07407, over 3323363.20 frames. ], batch size: 102, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:43:11,645 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:38,275 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=42940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:43:53,250 INFO [train.py:904] (0/8) Epoch 5, batch 2350, loss[loss=0.2347, simple_loss=0.3049, pruned_loss=0.08226, over 16251.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2969, pruned_loss=0.07366, over 3327676.20 frames. ], batch size: 165, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:44:03,300 INFO [optim.py:368] (0/8) 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 06:45:02,237 INFO [train.py:904] (0/8) Epoch 5, batch 2400, loss[loss=0.2466, simple_loss=0.3175, pruned_loss=0.0878, over 16691.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2983, pruned_loss=0.07431, over 3324074.29 frames. ], batch size: 83, lr: 1.38e-02, grad_scale: 8.0 2023-04-28 06:45:08,150 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:45:14,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2437, 3.8591, 3.2607, 5.3356, 4.8308, 4.4253, 2.0414, 3.6356], device='cuda:0'), covar=tensor([0.1057, 0.0390, 0.0770, 0.0066, 0.0222, 0.0328, 0.1104, 0.0525], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0140, 0.0162, 0.0084, 0.0176, 0.0169, 0.0156, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 06:45:52,164 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:46:13,645 INFO [train.py:904] (0/8) Epoch 5, batch 2450, loss[loss=0.2047, simple_loss=0.2962, pruned_loss=0.05658, over 17060.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.2981, pruned_loss=0.07378, over 3319023.95 frames. ], batch size: 55, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:46:26,011 INFO [optim.py:368] (0/8) 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 06:46:28,685 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0974, 5.1684, 4.9662, 4.8015, 4.3348, 5.0044, 5.0476, 4.6810], device='cuda:0'), covar=tensor([0.0505, 0.0217, 0.0210, 0.0193, 0.1082, 0.0278, 0.0199, 0.0487], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0192, 0.0227, 0.0198, 0.0262, 0.0218, 0.0165, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 06:46:57,967 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:47:23,978 INFO [train.py:904] (0/8) Epoch 5, batch 2500, loss[loss=0.2486, simple_loss=0.3192, pruned_loss=0.08904, over 16718.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2975, pruned_loss=0.07331, over 3319325.08 frames. ], batch size: 124, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:33,487 INFO [train.py:904] (0/8) Epoch 5, batch 2550, loss[loss=0.2482, simple_loss=0.3106, pruned_loss=0.09288, over 16851.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.2981, pruned_loss=0.07366, over 3319980.43 frames. ], batch size: 116, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:48:45,562 INFO [optim.py:368] (0/8) 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 06:48:48,810 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:49:43,650 INFO [train.py:904] (0/8) Epoch 5, batch 2600, loss[loss=0.2005, simple_loss=0.2828, pruned_loss=0.05908, over 17211.00 frames. ], tot_loss[loss=0.2215, simple_loss=0.2979, pruned_loss=0.07251, over 3330962.40 frames. ], batch size: 44, lr: 1.38e-02, grad_scale: 4.0 2023-04-28 06:49:55,552 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43210.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:01,756 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:50:31,277 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 5, batch 2650, loss[loss=0.2072, simple_loss=0.2941, pruned_loss=0.06013, over 17141.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.2973, pruned_loss=0.07123, over 3328673.82 frames. ], batch size: 48, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:51:05,582 INFO [optim.py:368] (0/8) 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 06:52:02,509 INFO [train.py:904] (0/8) Epoch 5, batch 2700, loss[loss=0.2313, simple_loss=0.3061, pruned_loss=0.07821, over 16855.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.2965, pruned_loss=0.07035, over 3332270.61 frames. ], batch size: 83, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:52:09,017 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 5, batch 2750, loss[loss=0.2223, simple_loss=0.3134, pruned_loss=0.06556, over 16644.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2961, pruned_loss=0.06966, over 3339319.06 frames. ], batch size: 62, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 06:53:15,876 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43353.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:53:25,430 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.808e+02 2.783e+02 3.325e+02 4.453e+02 8.515e+02, threshold=6.649e+02, percent-clipped=2.0 2023-04-28 06:54:22,971 INFO [train.py:904] (0/8) Epoch 5, batch 2800, loss[loss=0.2188, simple_loss=0.2889, pruned_loss=0.07434, over 16739.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2958, pruned_loss=0.07014, over 3330448.97 frames. ], batch size: 134, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:33,358 INFO [train.py:904] (0/8) Epoch 5, batch 2850, loss[loss=0.2331, simple_loss=0.3173, pruned_loss=0.07449, over 16668.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.296, pruned_loss=0.07052, over 3330931.30 frames. ], batch size: 57, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:55:45,525 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.398e+02 3.090e+02 3.971e+02 4.824e+02 1.597e+03, threshold=7.942e+02, percent-clipped=16.0 2023-04-28 06:56:41,596 INFO [train.py:904] (0/8) Epoch 5, batch 2900, loss[loss=0.2163, simple_loss=0.3042, pruned_loss=0.06418, over 16717.00 frames. ], tot_loss[loss=0.219, simple_loss=0.2959, pruned_loss=0.07102, over 3327875.97 frames. ], batch size: 57, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:57:00,406 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:01,418 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:25,047 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-28 06:57:28,495 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=43535.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:57:49,471 INFO [train.py:904] (0/8) Epoch 5, batch 2950, loss[loss=0.1816, simple_loss=0.2614, pruned_loss=0.05092, over 16874.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.2946, pruned_loss=0.07078, over 3337344.37 frames. ], batch size: 42, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:02,030 INFO [optim.py:368] (0/8) 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 06:58:05,998 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:24,150 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:28,451 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8034, 4.7907, 4.1071, 2.4272, 3.3839, 2.9187, 4.1202, 4.4090], device='cuda:0'), covar=tensor([0.0235, 0.0345, 0.0421, 0.1353, 0.0609, 0.0794, 0.0530, 0.0614], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0133, 0.0155, 0.0141, 0.0132, 0.0126, 0.0140, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 06:58:29,759 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 06:58:34,164 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=43583.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 06:58:59,872 INFO [train.py:904] (0/8) Epoch 5, batch 3000, loss[loss=0.1987, simple_loss=0.2761, pruned_loss=0.06063, over 15936.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2956, pruned_loss=0.0719, over 3329655.01 frames. ], batch size: 35, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 06:58:59,873 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 06:59:05,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9009, 3.8659, 4.2788, 4.2153, 4.2078, 3.9419, 3.9400, 3.8336], device='cuda:0'), covar=tensor([0.0254, 0.0372, 0.0252, 0.0358, 0.0361, 0.0316, 0.0718, 0.0431], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0243, 0.0246, 0.0249, 0.0298, 0.0260, 0.0372, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 06:59:08,835 INFO [train.py:938] (0/8) Epoch 5, validation: loss=0.1564, simple_loss=0.2632, pruned_loss=0.02477, over 944034.00 frames. 2023-04-28 06:59:08,835 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 06:59:25,162 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-28 07:00:18,530 INFO [train.py:904] (0/8) Epoch 5, batch 3050, loss[loss=0.2033, simple_loss=0.2949, pruned_loss=0.0558, over 17138.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.2954, pruned_loss=0.07159, over 3333712.29 frames. ], batch size: 49, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:00:31,426 INFO [optim.py:368] (0/8) 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 07:00:49,581 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3498, 4.1397, 4.3109, 4.5885, 4.6400, 4.2169, 4.4081, 4.6250], device='cuda:0'), covar=tensor([0.0713, 0.0648, 0.1047, 0.0428, 0.0449, 0.0748, 0.1194, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0506, 0.0656, 0.0526, 0.0392, 0.0388, 0.0406, 0.0430], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:01:02,939 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7689, 2.7501, 2.4187, 4.2026, 3.8094, 3.9096, 1.4671, 2.9296], device='cuda:0'), covar=tensor([0.1168, 0.0492, 0.0983, 0.0071, 0.0211, 0.0288, 0.1200, 0.0607], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0140, 0.0163, 0.0085, 0.0181, 0.0170, 0.0155, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 07:01:24,016 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4095, 4.3370, 4.3111, 3.8415, 4.3733, 1.8149, 4.1685, 4.2370], device='cuda:0'), covar=tensor([0.0086, 0.0068, 0.0087, 0.0271, 0.0061, 0.1530, 0.0091, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0087, 0.0130, 0.0137, 0.0100, 0.0141, 0.0115, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:01:25,955 INFO [train.py:904] (0/8) Epoch 5, batch 3100, loss[loss=0.1871, simple_loss=0.2627, pruned_loss=0.05574, over 16803.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.295, pruned_loss=0.07166, over 3323037.89 frames. ], batch size: 39, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:02:33,551 INFO [train.py:904] (0/8) Epoch 5, batch 3150, loss[loss=0.2138, simple_loss=0.2966, pruned_loss=0.06549, over 16732.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2933, pruned_loss=0.07117, over 3327862.02 frames. ], batch size: 57, lr: 1.37e-02, grad_scale: 4.0 2023-04-28 07:02:46,879 INFO [optim.py:368] (0/8) 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 07:03:22,284 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:03:42,789 INFO [train.py:904] (0/8) Epoch 5, batch 3200, loss[loss=0.2265, simple_loss=0.3043, pruned_loss=0.07433, over 17045.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2929, pruned_loss=0.07134, over 3327813.83 frames. ], batch size: 50, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:04:25,331 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9403, 4.8648, 4.6763, 4.1667, 4.7247, 2.0557, 4.5055, 4.7813], device='cuda:0'), covar=tensor([0.0065, 0.0050, 0.0086, 0.0285, 0.0057, 0.1374, 0.0082, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0086, 0.0130, 0.0136, 0.0100, 0.0139, 0.0114, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:04:39,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9355, 3.7610, 3.9608, 4.1469, 4.1853, 3.7518, 3.9875, 4.2216], device='cuda:0'), covar=tensor([0.0785, 0.0735, 0.0985, 0.0444, 0.0443, 0.1132, 0.1126, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0499, 0.0646, 0.0519, 0.0382, 0.0384, 0.0396, 0.0427], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:04:46,921 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4043, 4.4148, 4.4176, 4.3842, 4.2978, 4.9180, 4.5829, 4.2540], device='cuda:0'), covar=tensor([0.1348, 0.1516, 0.1584, 0.1623, 0.2762, 0.1042, 0.1218, 0.2461], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0409, 0.0393, 0.0345, 0.0461, 0.0414, 0.0330, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:04:48,189 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:04:52,120 INFO [train.py:904] (0/8) Epoch 5, batch 3250, loss[loss=0.1702, simple_loss=0.2586, pruned_loss=0.04092, over 17226.00 frames. ], tot_loss[loss=0.218, simple_loss=0.293, pruned_loss=0.07156, over 3329558.57 frames. ], batch size: 45, lr: 1.37e-02, grad_scale: 8.0 2023-04-28 07:05:06,670 INFO [optim.py:368] (0/8) 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 07:05:07,110 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1824, 3.9635, 4.1567, 4.3895, 4.4865, 4.0151, 4.2801, 4.4815], device='cuda:0'), covar=tensor([0.0825, 0.0792, 0.1285, 0.0532, 0.0473, 0.0972, 0.1087, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0498, 0.0646, 0.0520, 0.0382, 0.0384, 0.0396, 0.0427], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:05:19,449 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=43870.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:06:03,326 INFO [train.py:904] (0/8) Epoch 5, batch 3300, loss[loss=0.2446, simple_loss=0.3146, pruned_loss=0.08735, over 16304.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2933, pruned_loss=0.07165, over 3318043.82 frames. ], batch size: 165, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:06:14,797 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:06,490 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 07:07:12,756 INFO [train.py:904] (0/8) Epoch 5, batch 3350, loss[loss=0.1771, simple_loss=0.2625, pruned_loss=0.04589, over 17203.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.2934, pruned_loss=0.07096, over 3313556.32 frames. ], batch size: 44, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:07:24,263 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=43959.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:07:26,942 INFO [optim.py:368] (0/8) 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 07:07:39,251 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=43970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:08:17,912 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-44000.pt 2023-04-28 07:08:22,250 INFO [train.py:904] (0/8) Epoch 5, batch 3400, loss[loss=0.2034, simple_loss=0.2825, pruned_loss=0.06216, over 17214.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2921, pruned_loss=0.07025, over 3316190.98 frames. ], batch size: 45, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:08:47,628 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:09:01,966 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6553, 3.8331, 2.0503, 3.8712, 2.6097, 3.9262, 2.0248, 2.8594], device='cuda:0'), covar=tensor([0.0122, 0.0251, 0.1306, 0.0118, 0.0677, 0.0355, 0.1175, 0.0524], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0158, 0.0175, 0.0090, 0.0158, 0.0190, 0.0183, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 07:09:31,576 INFO [train.py:904] (0/8) Epoch 5, batch 3450, loss[loss=0.2024, simple_loss=0.2945, pruned_loss=0.0551, over 17104.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2918, pruned_loss=0.07019, over 3311567.46 frames. ], batch size: 49, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:09:44,814 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.795e+02 3.034e+02 3.683e+02 4.468e+02 1.074e+03, threshold=7.367e+02, percent-clipped=2.0 2023-04-28 07:09:54,253 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 07:09:56,192 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0182, 5.0017, 5.5260, 5.5182, 5.5443, 5.1309, 5.0861, 4.8790], device='cuda:0'), covar=tensor([0.0278, 0.0298, 0.0377, 0.0465, 0.0361, 0.0253, 0.0769, 0.0305], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0244, 0.0250, 0.0252, 0.0296, 0.0261, 0.0374, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 07:10:39,137 INFO [train.py:904] (0/8) Epoch 5, batch 3500, loss[loss=0.1784, simple_loss=0.2597, pruned_loss=0.0485, over 17199.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2901, pruned_loss=0.06942, over 3314178.13 frames. ], batch size: 44, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:10:58,554 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3021, 5.2076, 5.0956, 4.8888, 4.6564, 5.1322, 5.0634, 4.7384], device='cuda:0'), covar=tensor([0.0327, 0.0218, 0.0162, 0.0161, 0.0779, 0.0219, 0.0216, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0200, 0.0231, 0.0205, 0.0271, 0.0229, 0.0170, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:11:08,843 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1936, 1.3644, 1.8906, 2.1379, 2.3003, 2.2579, 1.4012, 2.2271], device='cuda:0'), covar=tensor([0.0070, 0.0194, 0.0111, 0.0115, 0.0076, 0.0073, 0.0191, 0.0044], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0145, 0.0129, 0.0133, 0.0129, 0.0095, 0.0140, 0.0083], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 07:11:34,388 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2928, 4.1000, 4.3069, 4.5591, 4.6423, 4.1675, 4.3940, 4.5947], device='cuda:0'), covar=tensor([0.0847, 0.0743, 0.1177, 0.0499, 0.0421, 0.0828, 0.1020, 0.0449], device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0505, 0.0649, 0.0521, 0.0385, 0.0391, 0.0394, 0.0430], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:11:37,798 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44143.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:11:49,616 INFO [train.py:904] (0/8) Epoch 5, batch 3550, loss[loss=0.1991, simple_loss=0.2733, pruned_loss=0.06244, over 16815.00 frames. ], tot_loss[loss=0.215, simple_loss=0.2899, pruned_loss=0.07008, over 3298542.88 frames. ], batch size: 83, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:12:03,024 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.484e+02 3.064e+02 3.621e+02 4.319e+02 6.916e+02, threshold=7.242e+02, percent-clipped=0.0 2023-04-28 07:12:17,180 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44170.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:12:42,680 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 07:12:59,218 INFO [train.py:904] (0/8) Epoch 5, batch 3600, loss[loss=0.2327, simple_loss=0.2999, pruned_loss=0.08276, over 16254.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2887, pruned_loss=0.06908, over 3305212.05 frames. ], batch size: 165, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:13:01,496 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 07:13:08,472 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 07:13:23,276 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44218.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:13:31,761 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1167, 1.5242, 2.3591, 2.8338, 2.8510, 3.0118, 1.6532, 3.0407], device='cuda:0'), covar=tensor([0.0062, 0.0259, 0.0148, 0.0106, 0.0092, 0.0083, 0.0231, 0.0056], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0145, 0.0129, 0.0133, 0.0129, 0.0095, 0.0141, 0.0082], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 07:13:53,778 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4274, 2.1054, 2.2892, 3.8707, 1.9349, 2.8987, 2.1693, 2.1814], device='cuda:0'), covar=tensor([0.0531, 0.1788, 0.0915, 0.0312, 0.2474, 0.0905, 0.1807, 0.1977], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0323, 0.0258, 0.0316, 0.0365, 0.0314, 0.0289, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:14:09,679 INFO [train.py:904] (0/8) Epoch 5, batch 3650, loss[loss=0.2246, simple_loss=0.3023, pruned_loss=0.07345, over 16631.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2873, pruned_loss=0.06976, over 3307948.32 frames. ], batch size: 57, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:14:16,308 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 07:14:25,437 INFO [optim.py:368] (0/8) 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 07:14:30,513 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:15:22,481 INFO [train.py:904] (0/8) Epoch 5, batch 3700, loss[loss=0.2145, simple_loss=0.2889, pruned_loss=0.07008, over 15486.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2866, pruned_loss=0.07097, over 3289928.04 frames. ], batch size: 190, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:15:44,549 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:15:45,002 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-28 07:16:36,543 INFO [train.py:904] (0/8) Epoch 5, batch 3750, loss[loss=0.2711, simple_loss=0.3302, pruned_loss=0.106, over 11860.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2872, pruned_loss=0.07245, over 3281636.15 frames. ], batch size: 248, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:16:52,716 INFO [optim.py:368] (0/8) 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 07:17:01,621 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6619, 4.4305, 3.7988, 1.9495, 2.9772, 2.5181, 3.8586, 4.1727], device='cuda:0'), covar=tensor([0.0169, 0.0301, 0.0502, 0.1697, 0.0731, 0.0858, 0.0522, 0.0539], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0133, 0.0151, 0.0139, 0.0131, 0.0124, 0.0139, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 07:17:45,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7478, 3.9706, 1.9228, 4.0713, 2.7319, 4.0331, 2.0159, 2.8792], device='cuda:0'), covar=tensor([0.0111, 0.0179, 0.1452, 0.0058, 0.0682, 0.0340, 0.1277, 0.0559], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0157, 0.0175, 0.0088, 0.0160, 0.0190, 0.0182, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 07:17:49,991 INFO [train.py:904] (0/8) Epoch 5, batch 3800, loss[loss=0.2607, simple_loss=0.313, pruned_loss=0.1041, over 16860.00 frames. ], tot_loss[loss=0.2176, simple_loss=0.288, pruned_loss=0.07357, over 3280853.69 frames. ], batch size: 109, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:18:24,441 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4067, 2.1504, 2.1876, 3.9734, 1.9384, 2.9232, 2.1090, 2.1992], device='cuda:0'), covar=tensor([0.0617, 0.2034, 0.1076, 0.0302, 0.2579, 0.0937, 0.2230, 0.2074], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0324, 0.0258, 0.0313, 0.0365, 0.0312, 0.0291, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:18:31,486 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5941, 2.2988, 2.3237, 4.2611, 1.9702, 3.0582, 2.1794, 2.3499], device='cuda:0'), covar=tensor([0.0616, 0.1923, 0.1031, 0.0277, 0.2738, 0.0972, 0.2147, 0.2231], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0324, 0.0258, 0.0313, 0.0365, 0.0312, 0.0291, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:18:49,806 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:19:01,028 INFO [train.py:904] (0/8) Epoch 5, batch 3850, loss[loss=0.2141, simple_loss=0.2825, pruned_loss=0.07279, over 16882.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.2877, pruned_loss=0.07392, over 3287666.70 frames. ], batch size: 116, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:19:16,583 INFO [optim.py:368] (0/8) 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 07:19:17,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0417, 4.4080, 1.9205, 4.6390, 2.8710, 4.6534, 2.3041, 2.8973], device='cuda:0'), covar=tensor([0.0116, 0.0141, 0.1687, 0.0021, 0.0731, 0.0143, 0.1275, 0.0619], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0159, 0.0175, 0.0088, 0.0162, 0.0191, 0.0185, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 07:19:25,625 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4961, 4.7880, 4.5235, 4.5793, 3.7108, 4.6998, 4.6948, 4.2953], device='cuda:0'), covar=tensor([0.0734, 0.0351, 0.0301, 0.0237, 0.1533, 0.0328, 0.0346, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0195, 0.0224, 0.0199, 0.0261, 0.0223, 0.0166, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:19:35,247 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0232, 4.7512, 5.0386, 5.2489, 5.3545, 4.7999, 5.3195, 5.3393], device='cuda:0'), covar=tensor([0.0802, 0.0675, 0.1104, 0.0440, 0.0332, 0.0456, 0.0357, 0.0392], device='cuda:0'), in_proj_covar=tensor([0.0389, 0.0478, 0.0606, 0.0494, 0.0367, 0.0364, 0.0381, 0.0409], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:19:57,672 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44491.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:20:11,854 INFO [train.py:904] (0/8) Epoch 5, batch 3900, loss[loss=0.2486, simple_loss=0.3062, pruned_loss=0.09549, over 16872.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.2868, pruned_loss=0.07388, over 3291768.75 frames. ], batch size: 116, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:20:49,854 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44527.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:21:09,442 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 07:21:22,446 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=44550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:21:23,250 INFO [train.py:904] (0/8) Epoch 5, batch 3950, loss[loss=0.2796, simple_loss=0.3378, pruned_loss=0.1107, over 12556.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2869, pruned_loss=0.07459, over 3275212.93 frames. ], batch size: 246, lr: 1.36e-02, grad_scale: 8.0 2023-04-28 07:21:33,171 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 07:21:37,719 INFO [optim.py:368] (0/8) 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 07:21:44,201 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:15,254 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=44588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:34,883 INFO [train.py:904] (0/8) Epoch 5, batch 4000, loss[loss=0.2217, simple_loss=0.2861, pruned_loss=0.07862, over 16775.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.2873, pruned_loss=0.07485, over 3272989.43 frames. ], batch size: 124, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:22:45,233 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2284, 3.9865, 3.9906, 1.5569, 4.1680, 4.3210, 3.2144, 2.8612], device='cuda:0'), covar=tensor([0.1037, 0.0112, 0.0175, 0.1402, 0.0056, 0.0041, 0.0336, 0.0549], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0088, 0.0084, 0.0142, 0.0073, 0.0078, 0.0117, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 07:22:48,823 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:22:54,963 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=44615.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:23:37,834 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5307, 4.1066, 4.1981, 2.7036, 3.7007, 4.0546, 4.0719, 2.6285], device='cuda:0'), covar=tensor([0.0288, 0.0020, 0.0017, 0.0227, 0.0033, 0.0032, 0.0018, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0059, 0.0060, 0.0114, 0.0061, 0.0069, 0.0064, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:23:45,848 INFO [train.py:904] (0/8) Epoch 5, batch 4050, loss[loss=0.2054, simple_loss=0.2834, pruned_loss=0.06366, over 16823.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2862, pruned_loss=0.07262, over 3281838.14 frames. ], batch size: 116, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:24:02,948 INFO [optim.py:368] (0/8) 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 07:24:04,436 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=44663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:24:05,845 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5078, 2.1360, 2.0529, 4.0911, 1.7868, 2.8491, 2.1373, 2.1166], device='cuda:0'), covar=tensor([0.0585, 0.2115, 0.1154, 0.0271, 0.3041, 0.1045, 0.2109, 0.2528], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0325, 0.0260, 0.0310, 0.0365, 0.0314, 0.0291, 0.0395], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:24:11,701 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6929, 2.6961, 2.3738, 4.1957, 3.6224, 3.9618, 1.4607, 2.8561], device='cuda:0'), covar=tensor([0.1226, 0.0592, 0.1045, 0.0071, 0.0221, 0.0249, 0.1340, 0.0734], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0141, 0.0165, 0.0086, 0.0179, 0.0171, 0.0158, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 07:25:01,119 INFO [train.py:904] (0/8) Epoch 5, batch 4100, loss[loss=0.2514, simple_loss=0.3137, pruned_loss=0.09454, over 12132.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2871, pruned_loss=0.07154, over 3272807.12 frames. ], batch size: 248, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:25:01,955 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.25 vs. limit=5.0 2023-04-28 07:26:16,735 INFO [train.py:904] (0/8) Epoch 5, batch 4150, loss[loss=0.2415, simple_loss=0.3212, pruned_loss=0.08095, over 16414.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.2958, pruned_loss=0.07559, over 3235532.05 frames. ], batch size: 146, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:26:20,967 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0061, 3.2115, 1.7091, 3.2777, 2.2170, 3.2839, 1.8138, 2.5154], device='cuda:0'), covar=tensor([0.0173, 0.0269, 0.1489, 0.0065, 0.0725, 0.0415, 0.1349, 0.0630], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0157, 0.0178, 0.0084, 0.0161, 0.0187, 0.0186, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 07:26:34,908 INFO [optim.py:368] (0/8) 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 07:27:10,005 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 07:27:11,704 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4560, 1.9170, 1.5085, 1.7279, 2.2427, 2.0114, 2.5173, 2.5327], device='cuda:0'), covar=tensor([0.0044, 0.0185, 0.0241, 0.0213, 0.0110, 0.0181, 0.0059, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0150, 0.0152, 0.0148, 0.0147, 0.0154, 0.0124, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:27:18,444 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-28 07:27:35,382 INFO [train.py:904] (0/8) Epoch 5, batch 4200, loss[loss=0.2708, simple_loss=0.3508, pruned_loss=0.09539, over 15365.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3035, pruned_loss=0.07759, over 3205587.98 frames. ], batch size: 191, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:28:51,243 INFO [train.py:904] (0/8) Epoch 5, batch 4250, loss[loss=0.2472, simple_loss=0.3132, pruned_loss=0.09056, over 12392.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.307, pruned_loss=0.07799, over 3186633.77 frames. ], batch size: 246, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:28:51,629 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6701, 3.6451, 3.9955, 4.0011, 4.0140, 3.6537, 3.7001, 3.6027], device='cuda:0'), covar=tensor([0.0265, 0.0451, 0.0345, 0.0390, 0.0380, 0.0316, 0.0890, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0225, 0.0228, 0.0233, 0.0278, 0.0244, 0.0349, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 07:29:07,192 INFO [optim.py:368] (0/8) 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 07:29:38,867 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44883.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:30:06,006 INFO [train.py:904] (0/8) Epoch 5, batch 4300, loss[loss=0.2425, simple_loss=0.3233, pruned_loss=0.08083, over 17029.00 frames. ], tot_loss[loss=0.231, simple_loss=0.3081, pruned_loss=0.07689, over 3185252.99 frames. ], batch size: 50, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:30:13,075 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=44906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:30:49,996 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9060, 4.1768, 3.8936, 3.9641, 3.5910, 3.6333, 3.8305, 4.1066], device='cuda:0'), covar=tensor([0.0623, 0.0634, 0.0883, 0.0438, 0.0598, 0.1281, 0.0613, 0.0775], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0468, 0.0409, 0.0310, 0.0301, 0.0311, 0.0382, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:31:14,574 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7875, 4.8049, 4.6245, 4.4705, 4.1875, 4.6811, 4.5635, 4.3090], device='cuda:0'), covar=tensor([0.0335, 0.0152, 0.0160, 0.0138, 0.0715, 0.0174, 0.0248, 0.0373], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0175, 0.0206, 0.0182, 0.0240, 0.0202, 0.0151, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:31:19,639 INFO [train.py:904] (0/8) Epoch 5, batch 4350, loss[loss=0.241, simple_loss=0.3226, pruned_loss=0.07974, over 16805.00 frames. ], tot_loss[loss=0.234, simple_loss=0.3114, pruned_loss=0.07829, over 3181413.35 frames. ], batch size: 83, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:31:36,574 INFO [optim.py:368] (0/8) 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 07:32:35,465 INFO [train.py:904] (0/8) Epoch 5, batch 4400, loss[loss=0.217, simple_loss=0.2984, pruned_loss=0.06775, over 17261.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3132, pruned_loss=0.07878, over 3188812.05 frames. ], batch size: 52, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:49,616 INFO [train.py:904] (0/8) Epoch 5, batch 4450, loss[loss=0.2242, simple_loss=0.309, pruned_loss=0.06969, over 16413.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3161, pruned_loss=0.07923, over 3188220.96 frames. ], batch size: 68, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:33:56,487 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:34:06,120 INFO [optim.py:368] (0/8) 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 07:34:12,439 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5648, 3.2515, 2.8432, 1.6403, 2.6056, 1.9784, 3.0173, 3.1882], device='cuda:0'), covar=tensor([0.0275, 0.0473, 0.0554, 0.1738, 0.0744, 0.1038, 0.0691, 0.0597], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0131, 0.0153, 0.0141, 0.0134, 0.0126, 0.0141, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 07:34:33,672 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45081.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:34:34,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8871, 2.6516, 1.9965, 2.5077, 3.1730, 2.7762, 3.9351, 3.5400], device='cuda:0'), covar=tensor([0.0014, 0.0166, 0.0255, 0.0181, 0.0079, 0.0161, 0.0044, 0.0076], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0150, 0.0153, 0.0149, 0.0147, 0.0155, 0.0123, 0.0136], device='cuda:0'), out_proj_covar=tensor([9.9622e-05, 1.8468e-04, 1.8532e-04, 1.8018e-04, 1.8441e-04, 1.9328e-04, 1.4942e-04, 1.6951e-04], device='cuda:0') 2023-04-28 07:35:01,854 INFO [train.py:904] (0/8) Epoch 5, batch 4500, loss[loss=0.2275, simple_loss=0.3023, pruned_loss=0.07631, over 16099.00 frames. ], tot_loss[loss=0.2362, simple_loss=0.3153, pruned_loss=0.0786, over 3205346.84 frames. ], batch size: 165, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:35:26,565 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45117.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 07:35:33,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 07:35:48,839 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 07:36:02,245 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45142.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:36:13,671 INFO [train.py:904] (0/8) Epoch 5, batch 4550, loss[loss=0.2555, simple_loss=0.3319, pruned_loss=0.08958, over 16732.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3158, pruned_loss=0.079, over 3222134.88 frames. ], batch size: 83, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:36:25,612 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 07:36:30,365 INFO [optim.py:368] (0/8) 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 07:37:00,746 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:37:27,630 INFO [train.py:904] (0/8) Epoch 5, batch 4600, loss[loss=0.227, simple_loss=0.3127, pruned_loss=0.07065, over 17022.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3163, pruned_loss=0.07875, over 3216909.21 frames. ], batch size: 50, lr: 1.35e-02, grad_scale: 8.0 2023-04-28 07:37:35,023 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:12,265 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45231.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:40,738 INFO [train.py:904] (0/8) Epoch 5, batch 4650, loss[loss=0.2213, simple_loss=0.3046, pruned_loss=0.06899, over 16528.00 frames. ], tot_loss[loss=0.2353, simple_loss=0.3147, pruned_loss=0.07799, over 3229591.33 frames. ], batch size: 75, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:38:45,220 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:38:48,026 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.49 vs. limit=5.0 2023-04-28 07:38:57,178 INFO [optim.py:368] (0/8) 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 07:39:55,300 INFO [train.py:904] (0/8) Epoch 5, batch 4700, loss[loss=0.2253, simple_loss=0.3086, pruned_loss=0.07097, over 16822.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3121, pruned_loss=0.07708, over 3215977.27 frames. ], batch size: 116, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:40:48,634 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4031, 3.2942, 3.3373, 2.6511, 3.3461, 1.9755, 3.1339, 2.9055], device='cuda:0'), covar=tensor([0.0127, 0.0108, 0.0108, 0.0371, 0.0082, 0.1603, 0.0117, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0080, 0.0118, 0.0129, 0.0090, 0.0137, 0.0106, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:40:49,822 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0106, 5.4588, 5.6114, 5.3974, 5.3420, 5.9883, 5.4951, 5.2874], device='cuda:0'), covar=tensor([0.0621, 0.1238, 0.0884, 0.1448, 0.1975, 0.0816, 0.0814, 0.1761], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0373, 0.0363, 0.0324, 0.0432, 0.0388, 0.0301, 0.0441], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:41:03,151 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45348.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:41:06,815 INFO [train.py:904] (0/8) Epoch 5, batch 4750, loss[loss=0.222, simple_loss=0.2989, pruned_loss=0.07255, over 17031.00 frames. ], tot_loss[loss=0.2291, simple_loss=0.3079, pruned_loss=0.0751, over 3222245.80 frames. ], batch size: 53, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:41:22,808 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.710e+02 3.330e+02 4.145e+02 7.902e+02, threshold=6.661e+02, percent-clipped=5.0 2023-04-28 07:41:40,222 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0643, 3.1714, 3.1374, 1.6171, 3.3783, 3.3473, 2.6904, 2.5720], device='cuda:0'), covar=tensor([0.0843, 0.0129, 0.0173, 0.1144, 0.0067, 0.0067, 0.0347, 0.0468], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0089, 0.0083, 0.0142, 0.0072, 0.0076, 0.0117, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 07:41:49,747 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2301, 5.5189, 5.1892, 5.2725, 4.9463, 4.7021, 4.9849, 5.5392], device='cuda:0'), covar=tensor([0.0609, 0.0522, 0.0823, 0.0388, 0.0519, 0.0565, 0.0502, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0453, 0.0392, 0.0298, 0.0290, 0.0301, 0.0370, 0.0327], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:42:07,626 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-28 07:42:12,524 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9841, 2.6904, 2.6378, 1.7089, 2.8619, 2.8166, 2.4218, 2.3534], device='cuda:0'), covar=tensor([0.0762, 0.0148, 0.0171, 0.1014, 0.0082, 0.0087, 0.0385, 0.0474], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0089, 0.0083, 0.0143, 0.0071, 0.0076, 0.0118, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 07:42:20,553 INFO [train.py:904] (0/8) Epoch 5, batch 4800, loss[loss=0.2466, simple_loss=0.3233, pruned_loss=0.08492, over 16892.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.304, pruned_loss=0.07291, over 3220829.91 frames. ], batch size: 109, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:42:33,336 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45409.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 07:42:38,070 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45412.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:42:55,779 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4897, 4.1964, 4.5011, 4.7049, 4.8451, 4.3215, 4.8232, 4.7952], device='cuda:0'), covar=tensor([0.0621, 0.0703, 0.0821, 0.0340, 0.0257, 0.0551, 0.0304, 0.0328], device='cuda:0'), in_proj_covar=tensor([0.0362, 0.0443, 0.0560, 0.0451, 0.0341, 0.0341, 0.0356, 0.0372], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:43:14,568 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:43:34,382 INFO [train.py:904] (0/8) Epoch 5, batch 4850, loss[loss=0.2483, simple_loss=0.3193, pruned_loss=0.08865, over 12186.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3057, pruned_loss=0.07307, over 3188104.85 frames. ], batch size: 246, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:43:50,677 INFO [optim.py:368] (0/8) 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 07:44:05,271 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7892, 1.2921, 1.4370, 1.7045, 1.8444, 1.9502, 1.3711, 1.7249], device='cuda:0'), covar=tensor([0.0097, 0.0191, 0.0104, 0.0129, 0.0089, 0.0059, 0.0169, 0.0040], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0142, 0.0127, 0.0126, 0.0126, 0.0089, 0.0139, 0.0082], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 07:44:47,407 INFO [train.py:904] (0/8) Epoch 5, batch 4900, loss[loss=0.2061, simple_loss=0.2885, pruned_loss=0.06181, over 17229.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3052, pruned_loss=0.07193, over 3178468.31 frames. ], batch size: 45, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:01,014 INFO [train.py:904] (0/8) Epoch 5, batch 4950, loss[loss=0.2517, simple_loss=0.3235, pruned_loss=0.08992, over 12036.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.305, pruned_loss=0.07158, over 3182286.43 frames. ], batch size: 246, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:46:15,486 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.470e+02 3.018e+02 3.573e+02 4.428e+02 9.346e+02, threshold=7.147e+02, percent-clipped=9.0 2023-04-28 07:46:34,295 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0779, 4.1378, 4.5018, 4.4470, 4.4426, 4.0014, 4.0742, 3.8798], device='cuda:0'), covar=tensor([0.0205, 0.0302, 0.0264, 0.0330, 0.0407, 0.0283, 0.0701, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0218, 0.0221, 0.0225, 0.0270, 0.0237, 0.0335, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 07:46:43,898 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2272, 4.0396, 4.2525, 4.4739, 4.6593, 4.1790, 4.6285, 4.6204], device='cuda:0'), covar=tensor([0.0978, 0.0777, 0.1198, 0.0474, 0.0321, 0.0746, 0.0362, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0447, 0.0568, 0.0455, 0.0340, 0.0344, 0.0354, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:46:56,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4490, 3.3692, 3.3632, 2.7873, 3.3820, 1.9039, 3.1996, 3.0446], device='cuda:0'), covar=tensor([0.0088, 0.0092, 0.0088, 0.0280, 0.0067, 0.1510, 0.0097, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0087, 0.0078, 0.0115, 0.0124, 0.0088, 0.0133, 0.0102, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:47:13,812 INFO [train.py:904] (0/8) Epoch 5, batch 5000, loss[loss=0.2289, simple_loss=0.3139, pruned_loss=0.07188, over 16235.00 frames. ], tot_loss[loss=0.226, simple_loss=0.3074, pruned_loss=0.07227, over 3179244.66 frames. ], batch size: 165, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:47:15,516 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5508, 2.4110, 2.1082, 4.0967, 3.3143, 3.7096, 1.4821, 2.4250], device='cuda:0'), covar=tensor([0.1436, 0.0766, 0.1374, 0.0081, 0.0259, 0.0352, 0.1481, 0.1063], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0139, 0.0164, 0.0083, 0.0168, 0.0169, 0.0159, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 07:47:54,766 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3220, 5.6456, 5.2543, 5.4400, 4.9859, 4.6586, 5.1135, 5.7045], device='cuda:0'), covar=tensor([0.0669, 0.0594, 0.0963, 0.0444, 0.0589, 0.0567, 0.0624, 0.0685], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0459, 0.0400, 0.0304, 0.0295, 0.0303, 0.0375, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:48:12,837 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6044, 3.6327, 2.8694, 2.2261, 2.6108, 2.1210, 3.7383, 3.7265], device='cuda:0'), covar=tensor([0.2056, 0.0632, 0.1277, 0.1568, 0.1730, 0.1391, 0.0423, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0253, 0.0268, 0.0244, 0.0296, 0.0203, 0.0242, 0.0256], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:48:22,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8361, 3.2515, 2.6688, 4.9698, 4.2899, 4.1644, 1.7093, 3.1887], device='cuda:0'), covar=tensor([0.1266, 0.0540, 0.1024, 0.0050, 0.0207, 0.0264, 0.1330, 0.0681], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0141, 0.0166, 0.0084, 0.0171, 0.0170, 0.0161, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 07:48:27,075 INFO [train.py:904] (0/8) Epoch 5, batch 5050, loss[loss=0.2093, simple_loss=0.2869, pruned_loss=0.06586, over 17302.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3073, pruned_loss=0.07193, over 3195859.93 frames. ], batch size: 52, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:48:43,128 INFO [optim.py:368] (0/8) 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 07:48:53,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2326, 4.1359, 4.1034, 3.0779, 4.0785, 1.3751, 3.8666, 3.8919], device='cuda:0'), covar=tensor([0.0088, 0.0088, 0.0095, 0.0528, 0.0080, 0.2067, 0.0119, 0.0208], device='cuda:0'), in_proj_covar=tensor([0.0088, 0.0078, 0.0116, 0.0126, 0.0088, 0.0135, 0.0103, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:49:25,479 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7134, 5.1834, 5.2682, 5.2722, 5.1759, 5.7529, 5.3084, 5.1170], device='cuda:0'), covar=tensor([0.0823, 0.1408, 0.0833, 0.1377, 0.1999, 0.0745, 0.0967, 0.1871], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0373, 0.0365, 0.0321, 0.0427, 0.0389, 0.0307, 0.0447], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 07:49:39,982 INFO [train.py:904] (0/8) Epoch 5, batch 5100, loss[loss=0.2629, simple_loss=0.3295, pruned_loss=0.0981, over 12287.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3045, pruned_loss=0.07043, over 3209652.04 frames. ], batch size: 248, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:49:42,360 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45702.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:49:44,454 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=45704.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:49:56,039 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45712.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:50:33,139 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=45737.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:50:33,311 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1204, 1.7662, 1.9914, 3.5407, 1.5681, 2.6708, 1.9064, 1.7853], device='cuda:0'), covar=tensor([0.0736, 0.2421, 0.1287, 0.0426, 0.3726, 0.1154, 0.2620, 0.2905], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0314, 0.0256, 0.0304, 0.0362, 0.0300, 0.0281, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:50:53,052 INFO [train.py:904] (0/8) Epoch 5, batch 5150, loss[loss=0.2129, simple_loss=0.2969, pruned_loss=0.06446, over 16460.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3037, pruned_loss=0.06917, over 3209732.30 frames. ], batch size: 68, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:51:06,960 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=45760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:08,849 INFO [optim.py:368] (0/8) 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 07:51:10,533 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45763.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:51:26,011 INFO [zipformer.py:625] (0/8) 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] (0/8) 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] (0/8) Epoch 5, batch 5200, loss[loss=0.2182, simple_loss=0.2992, pruned_loss=0.06855, over 16387.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3029, pruned_loss=0.06892, over 3215582.43 frames. ], batch size: 165, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:52:54,245 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45835.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:16,338 INFO [train.py:904] (0/8) Epoch 5, batch 5250, loss[loss=0.205, simple_loss=0.2736, pruned_loss=0.06821, over 16598.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.2997, pruned_loss=0.06844, over 3223070.06 frames. ], batch size: 57, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:53:26,677 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=45859.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:53:31,056 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.833e+02 2.734e+02 3.218e+02 3.993e+02 9.159e+02, threshold=6.436e+02, percent-clipped=4.0 2023-04-28 07:53:45,232 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0149, 3.2808, 3.3914, 1.5590, 3.5579, 3.6322, 2.7981, 2.5353], device='cuda:0'), covar=tensor([0.0842, 0.0146, 0.0134, 0.1262, 0.0051, 0.0048, 0.0296, 0.0468], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0088, 0.0079, 0.0142, 0.0070, 0.0073, 0.0115, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 07:54:26,351 INFO [train.py:904] (0/8) Epoch 5, batch 5300, loss[loss=0.1782, simple_loss=0.2674, pruned_loss=0.04443, over 16824.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.296, pruned_loss=0.06665, over 3232648.38 frames. ], batch size: 102, lr: 1.34e-02, grad_scale: 8.0 2023-04-28 07:54:53,141 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=45920.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:55:26,606 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8457, 4.2726, 1.8257, 4.6963, 2.7313, 4.4523, 2.1914, 2.9649], device='cuda:0'), covar=tensor([0.0135, 0.0170, 0.1872, 0.0020, 0.0775, 0.0229, 0.1442, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0148, 0.0177, 0.0080, 0.0160, 0.0181, 0.0186, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 07:55:36,445 INFO [train.py:904] (0/8) Epoch 5, batch 5350, loss[loss=0.2112, simple_loss=0.3027, pruned_loss=0.05988, over 16880.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2943, pruned_loss=0.06576, over 3228649.86 frames. ], batch size: 116, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:55:53,056 INFO [optim.py:368] (0/8) 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 07:56:30,130 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 07:56:47,994 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-46000.pt 2023-04-28 07:56:52,502 INFO [train.py:904] (0/8) Epoch 5, batch 5400, loss[loss=0.2259, simple_loss=0.3164, pruned_loss=0.06771, over 16802.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2976, pruned_loss=0.06689, over 3223293.90 frames. ], batch size: 102, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:56:57,453 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46004.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 07:57:36,535 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3425, 2.0845, 1.5690, 1.7989, 2.4180, 2.1778, 2.5997, 2.5845], device='cuda:0'), covar=tensor([0.0045, 0.0179, 0.0240, 0.0231, 0.0109, 0.0183, 0.0070, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0154, 0.0157, 0.0153, 0.0150, 0.0156, 0.0124, 0.0137], device='cuda:0'), out_proj_covar=tensor([9.9425e-05, 1.8906e-04, 1.8965e-04, 1.8483e-04, 1.8672e-04, 1.9321e-04, 1.4827e-04, 1.7026e-04], device='cuda:0') 2023-04-28 07:58:03,608 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4313, 3.3282, 2.6952, 2.1978, 2.3572, 2.1125, 3.3255, 3.5357], device='cuda:0'), covar=tensor([0.1853, 0.0604, 0.1119, 0.1387, 0.1564, 0.1240, 0.0377, 0.0500], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0250, 0.0266, 0.0244, 0.0291, 0.0201, 0.0243, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 07:58:09,293 INFO [train.py:904] (0/8) Epoch 5, batch 5450, loss[loss=0.2549, simple_loss=0.3246, pruned_loss=0.09266, over 16290.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3023, pruned_loss=0.06994, over 3213675.62 frames. ], batch size: 35, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 07:58:11,177 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46052.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 07:58:19,260 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46058.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 07:58:24,997 INFO [optim.py:368] (0/8) 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 07:59:16,104 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5793, 2.1881, 1.5679, 1.8104, 2.5555, 2.1924, 2.7904, 2.7552], device='cuda:0'), covar=tensor([0.0049, 0.0161, 0.0248, 0.0232, 0.0113, 0.0191, 0.0075, 0.0115], device='cuda:0'), in_proj_covar=tensor([0.0079, 0.0154, 0.0157, 0.0154, 0.0150, 0.0157, 0.0125, 0.0137], device='cuda:0'), out_proj_covar=tensor([9.9772e-05, 1.8964e-04, 1.8904e-04, 1.8524e-04, 1.8651e-04, 1.9366e-04, 1.4962e-04, 1.6986e-04], device='cuda:0') 2023-04-28 07:59:22,166 INFO [train.py:904] (0/8) Epoch 5, batch 5500, loss[loss=0.3365, simple_loss=0.3852, pruned_loss=0.1439, over 11830.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3121, pruned_loss=0.07697, over 3179088.84 frames. ], batch size: 248, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:07,590 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46130.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:00:39,137 INFO [train.py:904] (0/8) Epoch 5, batch 5550, loss[loss=0.4297, simple_loss=0.4326, pruned_loss=0.2134, over 11338.00 frames. ], tot_loss[loss=0.247, simple_loss=0.3222, pruned_loss=0.08584, over 3127353.50 frames. ], batch size: 247, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:00:56,875 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.500e+02 4.429e+02 5.470e+02 6.755e+02 1.488e+03, threshold=1.094e+03, percent-clipped=17.0 2023-04-28 08:01:58,733 INFO [train.py:904] (0/8) Epoch 5, batch 5600, loss[loss=0.2937, simple_loss=0.3597, pruned_loss=0.1138, over 16776.00 frames. ], tot_loss[loss=0.2559, simple_loss=0.3286, pruned_loss=0.09158, over 3101338.65 frames. ], batch size: 124, lr: 1.33e-02, grad_scale: 16.0 2023-04-28 08:02:23,968 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46215.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:03:09,593 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8065, 1.3839, 1.6123, 1.7697, 1.8160, 1.8747, 1.4992, 1.7836], device='cuda:0'), covar=tensor([0.0096, 0.0160, 0.0089, 0.0110, 0.0090, 0.0056, 0.0145, 0.0045], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0142, 0.0127, 0.0124, 0.0126, 0.0089, 0.0140, 0.0080], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 08:03:23,171 INFO [train.py:904] (0/8) Epoch 5, batch 5650, loss[loss=0.2674, simple_loss=0.3361, pruned_loss=0.09937, over 16367.00 frames. ], tot_loss[loss=0.2633, simple_loss=0.3342, pruned_loss=0.09623, over 3086019.10 frames. ], batch size: 146, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:03:23,657 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4696, 2.0337, 1.5314, 1.7833, 2.3424, 2.1191, 2.4096, 2.4775], device='cuda:0'), covar=tensor([0.0034, 0.0147, 0.0218, 0.0190, 0.0091, 0.0157, 0.0068, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0075, 0.0150, 0.0152, 0.0148, 0.0145, 0.0152, 0.0120, 0.0132], device='cuda:0'), out_proj_covar=tensor([9.4537e-05, 1.8346e-04, 1.8268e-04, 1.7862e-04, 1.8061e-04, 1.8697e-04, 1.4394e-04, 1.6287e-04], device='cuda:0') 2023-04-28 08:03:25,379 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5940, 4.8827, 4.9215, 4.9148, 4.8721, 5.3983, 5.0197, 4.8654], device='cuda:0'), covar=tensor([0.0922, 0.1462, 0.1411, 0.1498, 0.2024, 0.0857, 0.1204, 0.2064], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0373, 0.0370, 0.0325, 0.0435, 0.0392, 0.0308, 0.0448], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:03:42,415 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.653e+02 4.676e+02 6.240e+02 7.908e+02 1.367e+03, threshold=1.248e+03, percent-clipped=1.0 2023-04-28 08:03:49,267 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7895, 3.7761, 4.3058, 4.2533, 4.2558, 3.8238, 3.9333, 3.8551], device='cuda:0'), covar=tensor([0.0302, 0.0432, 0.0367, 0.0473, 0.0500, 0.0355, 0.0842, 0.0427], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0220, 0.0226, 0.0228, 0.0275, 0.0243, 0.0341, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 08:04:04,193 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46276.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:04:29,160 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-28 08:04:43,826 INFO [train.py:904] (0/8) Epoch 5, batch 5700, loss[loss=0.2664, simple_loss=0.3513, pruned_loss=0.09075, over 16950.00 frames. ], tot_loss[loss=0.2668, simple_loss=0.3363, pruned_loss=0.09859, over 3071145.94 frames. ], batch size: 116, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:05:41,858 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:04,714 INFO [train.py:904] (0/8) Epoch 5, batch 5750, loss[loss=0.2991, simple_loss=0.3458, pruned_loss=0.1262, over 10940.00 frames. ], tot_loss[loss=0.2702, simple_loss=0.3395, pruned_loss=0.1005, over 3039162.01 frames. ], batch size: 247, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:06:16,662 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46358.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:06:23,498 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.622e+02 4.192e+02 4.948e+02 6.377e+02 1.174e+03, threshold=9.897e+02, percent-clipped=0.0 2023-04-28 08:07:26,290 INFO [train.py:904] (0/8) Epoch 5, batch 5800, loss[loss=0.2596, simple_loss=0.3402, pruned_loss=0.08947, over 16729.00 frames. ], tot_loss[loss=0.2672, simple_loss=0.3383, pruned_loss=0.09808, over 3048647.05 frames. ], batch size: 89, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:07:35,756 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:13,814 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46430.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:08:46,737 INFO [train.py:904] (0/8) Epoch 5, batch 5850, loss[loss=0.2383, simple_loss=0.3243, pruned_loss=0.07616, over 16715.00 frames. ], tot_loss[loss=0.2623, simple_loss=0.3349, pruned_loss=0.09483, over 3059040.24 frames. ], batch size: 89, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:09:06,686 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.458e+02 3.799e+02 4.852e+02 6.197e+02 1.255e+03, threshold=9.704e+02, percent-clipped=4.0 2023-04-28 08:09:28,722 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:10:09,074 INFO [train.py:904] (0/8) Epoch 5, batch 5900, loss[loss=0.2443, simple_loss=0.3202, pruned_loss=0.0842, over 16259.00 frames. ], tot_loss[loss=0.262, simple_loss=0.3345, pruned_loss=0.09471, over 3051668.68 frames. ], batch size: 165, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:10:16,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6649, 3.8802, 1.8651, 4.1461, 2.5408, 4.0184, 1.9498, 2.8693], device='cuda:0'), covar=tensor([0.0114, 0.0188, 0.1626, 0.0030, 0.0810, 0.0305, 0.1596, 0.0652], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0146, 0.0177, 0.0078, 0.0159, 0.0179, 0.0186, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 08:10:29,176 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8566, 4.1189, 3.8730, 3.8867, 3.6052, 3.7321, 3.8031, 4.0382], device='cuda:0'), covar=tensor([0.0683, 0.0690, 0.0939, 0.0492, 0.0592, 0.1288, 0.0666, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0470, 0.0407, 0.0308, 0.0295, 0.0313, 0.0384, 0.0337], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:10:36,181 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46515.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:10:54,025 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1628, 1.9420, 2.0889, 3.6294, 1.7924, 2.7586, 2.1104, 1.9746], device='cuda:0'), covar=tensor([0.0603, 0.2058, 0.1125, 0.0328, 0.2999, 0.1007, 0.1863, 0.2507], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0309, 0.0252, 0.0298, 0.0364, 0.0296, 0.0276, 0.0369], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:11:31,970 INFO [train.py:904] (0/8) Epoch 5, batch 5950, loss[loss=0.2586, simple_loss=0.3348, pruned_loss=0.09121, over 15275.00 frames. ], tot_loss[loss=0.2595, simple_loss=0.3342, pruned_loss=0.09241, over 3066109.68 frames. ], batch size: 190, lr: 1.33e-02, grad_scale: 4.0 2023-04-28 08:11:52,726 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:11:53,496 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.225e+02 3.478e+02 4.456e+02 5.303e+02 9.511e+02, threshold=8.911e+02, percent-clipped=0.0 2023-04-28 08:12:43,223 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:12:51,982 INFO [train.py:904] (0/8) Epoch 5, batch 6000, loss[loss=0.2397, simple_loss=0.3125, pruned_loss=0.08341, over 17069.00 frames. ], tot_loss[loss=0.2594, simple_loss=0.3337, pruned_loss=0.09252, over 3055883.25 frames. ], batch size: 55, lr: 1.33e-02, grad_scale: 8.0 2023-04-28 08:12:51,983 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 08:13:04,050 INFO [train.py:938] (0/8) Epoch 5, validation: loss=0.1879, simple_loss=0.2992, pruned_loss=0.03826, over 944034.00 frames. 2023-04-28 08:13:04,051 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 08:13:05,717 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:13:51,305 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:13:52,993 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6466, 4.8740, 4.9282, 4.9330, 4.9210, 5.4657, 4.9843, 4.7565], device='cuda:0'), covar=tensor([0.0804, 0.1379, 0.1380, 0.1417, 0.2016, 0.0780, 0.1211, 0.2125], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0366, 0.0367, 0.0320, 0.0425, 0.0385, 0.0301, 0.0439], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:14:27,410 INFO [train.py:904] (0/8) Epoch 5, batch 6050, loss[loss=0.2466, simple_loss=0.3232, pruned_loss=0.08498, over 16282.00 frames. ], tot_loss[loss=0.2571, simple_loss=0.3315, pruned_loss=0.09132, over 3084733.12 frames. ], batch size: 35, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:14:36,907 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46656.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:14:48,249 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46663.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:14:49,000 INFO [optim.py:368] (0/8) 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 08:15:35,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9698, 2.3397, 1.7432, 2.0070, 2.8841, 2.5408, 3.2031, 3.1373], device='cuda:0'), covar=tensor([0.0036, 0.0170, 0.0254, 0.0209, 0.0087, 0.0163, 0.0066, 0.0085], device='cuda:0'), in_proj_covar=tensor([0.0076, 0.0152, 0.0153, 0.0151, 0.0146, 0.0154, 0.0125, 0.0134], device='cuda:0'), out_proj_covar=tensor([9.5301e-05, 1.8602e-04, 1.8332e-04, 1.8132e-04, 1.8058e-04, 1.8932e-04, 1.5015e-04, 1.6527e-04], device='cuda:0') 2023-04-28 08:15:45,451 INFO [train.py:904] (0/8) Epoch 5, batch 6100, loss[loss=0.2173, simple_loss=0.2984, pruned_loss=0.06811, over 17099.00 frames. ], tot_loss[loss=0.2545, simple_loss=0.3303, pruned_loss=0.08938, over 3111297.35 frames. ], batch size: 49, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:16:34,275 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9725, 1.5510, 2.2805, 2.9037, 2.7315, 3.2408, 1.6599, 2.9708], device='cuda:0'), covar=tensor([0.0060, 0.0233, 0.0137, 0.0103, 0.0091, 0.0049, 0.0233, 0.0046], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0138, 0.0123, 0.0119, 0.0122, 0.0086, 0.0135, 0.0076], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 08:16:39,189 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5540, 4.5785, 4.3688, 3.5233, 4.4015, 1.4602, 4.2091, 4.3152], device='cuda:0'), covar=tensor([0.0066, 0.0043, 0.0085, 0.0330, 0.0057, 0.1912, 0.0080, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0078, 0.0119, 0.0126, 0.0091, 0.0142, 0.0105, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:17:01,187 INFO [train.py:904] (0/8) Epoch 5, batch 6150, loss[loss=0.2709, simple_loss=0.3285, pruned_loss=0.1067, over 11450.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3286, pruned_loss=0.08927, over 3104583.03 frames. ], batch size: 246, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:17:07,064 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7131, 3.8370, 3.0581, 2.4520, 2.9091, 2.2420, 4.1299, 4.0002], device='cuda:0'), covar=tensor([0.2271, 0.0704, 0.1352, 0.1492, 0.1981, 0.1534, 0.0368, 0.0600], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0250, 0.0270, 0.0244, 0.0294, 0.0203, 0.0241, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:17:22,722 INFO [optim.py:368] (0/8) 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 08:17:24,384 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5213, 4.7898, 4.4693, 4.5070, 4.2292, 4.1767, 4.3311, 4.7799], device='cuda:0'), covar=tensor([0.0641, 0.0617, 0.0916, 0.0482, 0.0566, 0.0917, 0.0596, 0.0675], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0478, 0.0416, 0.0314, 0.0300, 0.0318, 0.0391, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:18:21,100 INFO [train.py:904] (0/8) Epoch 5, batch 6200, loss[loss=0.259, simple_loss=0.3275, pruned_loss=0.09528, over 16663.00 frames. ], tot_loss[loss=0.2515, simple_loss=0.3263, pruned_loss=0.08833, over 3106575.92 frames. ], batch size: 76, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:18:23,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7771, 3.3990, 3.4004, 2.2381, 3.1593, 3.3428, 3.2788, 1.8833], device='cuda:0'), covar=tensor([0.0347, 0.0024, 0.0027, 0.0229, 0.0037, 0.0051, 0.0033, 0.0288], device='cuda:0'), in_proj_covar=tensor([0.0112, 0.0051, 0.0056, 0.0111, 0.0057, 0.0066, 0.0061, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:18:35,820 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=46810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:19:37,513 INFO [train.py:904] (0/8) Epoch 5, batch 6250, loss[loss=0.2439, simple_loss=0.332, pruned_loss=0.07794, over 16932.00 frames. ], tot_loss[loss=0.25, simple_loss=0.3254, pruned_loss=0.08734, over 3126498.63 frames. ], batch size: 116, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:19:57,348 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.405e+02 3.628e+02 4.524e+02 5.799e+02 1.377e+03, threshold=9.049e+02, percent-clipped=5.0 2023-04-28 08:20:09,453 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=46871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:20:55,077 INFO [train.py:904] (0/8) Epoch 5, batch 6300, loss[loss=0.2375, simple_loss=0.3198, pruned_loss=0.07756, over 17061.00 frames. ], tot_loss[loss=0.2507, simple_loss=0.3263, pruned_loss=0.08755, over 3129978.00 frames. ], batch size: 53, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:21:22,365 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.50 vs. limit=5.0 2023-04-28 08:21:43,920 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=46932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:12,105 INFO [train.py:904] (0/8) Epoch 5, batch 6350, loss[loss=0.2252, simple_loss=0.3114, pruned_loss=0.06948, over 17231.00 frames. ], tot_loss[loss=0.2535, simple_loss=0.3278, pruned_loss=0.08961, over 3114657.57 frames. ], batch size: 52, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:22:13,460 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6634, 5.0346, 5.1930, 5.1166, 5.1732, 5.6677, 5.3025, 5.0543], device='cuda:0'), covar=tensor([0.0856, 0.1440, 0.1188, 0.1562, 0.1913, 0.0777, 0.0987, 0.1976], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0377, 0.0380, 0.0331, 0.0438, 0.0400, 0.0309, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:22:13,471 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46951.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:22:14,853 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5989, 2.6550, 1.7224, 2.7471, 2.1705, 2.7267, 1.8925, 2.4022], device='cuda:0'), covar=tensor([0.0172, 0.0306, 0.1265, 0.0098, 0.0619, 0.0645, 0.1144, 0.0578], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0149, 0.0181, 0.0081, 0.0164, 0.0187, 0.0189, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 08:22:22,796 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=46958.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:22:31,677 INFO [optim.py:368] (0/8) 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 08:22:35,465 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6973, 3.6251, 3.7675, 3.6773, 3.7225, 4.1295, 3.9256, 3.6914], device='cuda:0'), covar=tensor([0.1738, 0.1965, 0.1787, 0.2223, 0.2686, 0.1638, 0.1441, 0.2589], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0376, 0.0380, 0.0331, 0.0436, 0.0399, 0.0308, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:22:56,445 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=46980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:23:28,022 INFO [train.py:904] (0/8) Epoch 5, batch 6400, loss[loss=0.2361, simple_loss=0.3111, pruned_loss=0.08058, over 16868.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.3295, pruned_loss=0.09184, over 3088494.67 frames. ], batch size: 42, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:24:42,548 INFO [train.py:904] (0/8) Epoch 5, batch 6450, loss[loss=0.2294, simple_loss=0.3065, pruned_loss=0.07617, over 16386.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3275, pruned_loss=0.08956, over 3101120.58 frames. ], batch size: 146, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:25:02,353 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.185e+02 3.689e+02 4.668e+02 5.826e+02 1.210e+03, threshold=9.337e+02, percent-clipped=4.0 2023-04-28 08:25:59,703 INFO [train.py:904] (0/8) Epoch 5, batch 6500, loss[loss=0.2205, simple_loss=0.299, pruned_loss=0.07099, over 17218.00 frames. ], tot_loss[loss=0.2503, simple_loss=0.3241, pruned_loss=0.0883, over 3084638.38 frames. ], batch size: 45, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:27:10,919 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 08:27:18,021 INFO [train.py:904] (0/8) Epoch 5, batch 6550, loss[loss=0.3053, simple_loss=0.3551, pruned_loss=0.1277, over 11713.00 frames. ], tot_loss[loss=0.2536, simple_loss=0.3279, pruned_loss=0.08968, over 3085218.71 frames. ], batch size: 246, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:27:37,108 INFO [optim.py:368] (0/8) 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 08:27:41,363 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:28:33,665 INFO [train.py:904] (0/8) Epoch 5, batch 6600, loss[loss=0.261, simple_loss=0.3435, pruned_loss=0.08927, over 16523.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3304, pruned_loss=0.09033, over 3079391.64 frames. ], batch size: 68, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:35,157 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:29:51,339 INFO [train.py:904] (0/8) Epoch 5, batch 6650, loss[loss=0.3263, simple_loss=0.3643, pruned_loss=0.1442, over 11315.00 frames. ], tot_loss[loss=0.2565, simple_loss=0.3305, pruned_loss=0.09123, over 3081668.10 frames. ], batch size: 247, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:29:51,895 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:30:02,728 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 08:31:04,891 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:06,914 INFO [train.py:904] (0/8) Epoch 5, batch 6700, loss[loss=0.2566, simple_loss=0.3264, pruned_loss=0.0934, over 16943.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3296, pruned_loss=0.09134, over 3080875.23 frames. ], batch size: 116, lr: 1.32e-02, grad_scale: 8.0 2023-04-28 08:31:09,314 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:31:16,567 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47306.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:32:26,003 INFO [train.py:904] (0/8) Epoch 5, batch 6750, loss[loss=0.2351, simple_loss=0.3102, pruned_loss=0.08004, over 16848.00 frames. ], tot_loss[loss=0.255, simple_loss=0.3282, pruned_loss=0.09095, over 3086029.71 frames. ], batch size: 116, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:32:45,845 INFO [optim.py:368] (0/8) 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 08:33:08,546 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:33:40,663 INFO [train.py:904] (0/8) Epoch 5, batch 6800, loss[loss=0.2652, simple_loss=0.3361, pruned_loss=0.09711, over 15356.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.328, pruned_loss=0.0903, over 3100195.21 frames. ], batch size: 190, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:34:40,976 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:34:57,650 INFO [train.py:904] (0/8) Epoch 5, batch 6850, loss[loss=0.2456, simple_loss=0.3418, pruned_loss=0.07471, over 16741.00 frames. ], tot_loss[loss=0.2544, simple_loss=0.3285, pruned_loss=0.0901, over 3111534.86 frames. ], batch size: 83, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:35:09,572 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7372, 2.2183, 1.6915, 1.8837, 2.7498, 2.4801, 2.8795, 2.8582], device='cuda:0'), covar=tensor([0.0038, 0.0199, 0.0259, 0.0226, 0.0095, 0.0161, 0.0094, 0.0109], device='cuda:0'), in_proj_covar=tensor([0.0074, 0.0151, 0.0153, 0.0149, 0.0143, 0.0154, 0.0125, 0.0133], device='cuda:0'), out_proj_covar=tensor([9.1979e-05, 1.8409e-04, 1.8228e-04, 1.7907e-04, 1.7613e-04, 1.8879e-04, 1.4890e-04, 1.6322e-04], device='cuda:0') 2023-04-28 08:35:17,578 INFO [optim.py:368] (0/8) 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 08:35:20,992 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:36:12,149 INFO [train.py:904] (0/8) Epoch 5, batch 6900, loss[loss=0.291, simple_loss=0.3549, pruned_loss=0.1135, over 15367.00 frames. ], tot_loss[loss=0.2553, simple_loss=0.3312, pruned_loss=0.08975, over 3117485.01 frames. ], batch size: 191, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:36:32,405 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 08:36:33,233 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:37:15,385 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8960, 5.6675, 5.9024, 5.6558, 5.6389, 6.1807, 5.7080, 5.5110], device='cuda:0'), covar=tensor([0.0717, 0.1478, 0.1256, 0.1575, 0.2154, 0.0815, 0.1221, 0.2239], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0375, 0.0388, 0.0334, 0.0443, 0.0403, 0.0314, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:37:30,515 INFO [train.py:904] (0/8) Epoch 5, batch 6950, loss[loss=0.2365, simple_loss=0.3215, pruned_loss=0.07574, over 16526.00 frames. ], tot_loss[loss=0.2593, simple_loss=0.3339, pruned_loss=0.09235, over 3111710.95 frames. ], batch size: 75, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:37:39,139 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7365, 4.7390, 4.5191, 3.8476, 4.6049, 1.5248, 4.3239, 4.4820], device='cuda:0'), covar=tensor([0.0049, 0.0040, 0.0078, 0.0290, 0.0043, 0.1792, 0.0070, 0.0109], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0077, 0.0118, 0.0123, 0.0088, 0.0140, 0.0104, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:37:50,871 INFO [optim.py:368] (0/8) 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 08:38:15,572 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-28 08:38:28,867 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0641, 4.7588, 5.0336, 5.3049, 5.4361, 4.6387, 5.3754, 5.3383], device='cuda:0'), covar=tensor([0.0792, 0.0642, 0.1007, 0.0371, 0.0313, 0.0512, 0.0385, 0.0393], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0448, 0.0568, 0.0460, 0.0345, 0.0336, 0.0369, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:38:42,825 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:38:48,259 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3843, 5.7395, 5.3839, 5.4470, 4.9968, 4.9200, 5.1563, 5.8035], device='cuda:0'), covar=tensor([0.0769, 0.0677, 0.1124, 0.0528, 0.0738, 0.0582, 0.0700, 0.0717], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0473, 0.0416, 0.0307, 0.0296, 0.0321, 0.0389, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:38:49,099 INFO [train.py:904] (0/8) Epoch 5, batch 7000, loss[loss=0.2438, simple_loss=0.3367, pruned_loss=0.07548, over 16687.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3334, pruned_loss=0.09145, over 3113527.85 frames. ], batch size: 57, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:39:53,593 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5079, 3.3412, 2.9235, 1.7357, 2.5852, 2.0915, 3.0081, 3.1248], device='cuda:0'), covar=tensor([0.0280, 0.0433, 0.0527, 0.1578, 0.0747, 0.0874, 0.0623, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0123, 0.0152, 0.0139, 0.0131, 0.0122, 0.0138, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 08:40:06,970 INFO [train.py:904] (0/8) Epoch 5, batch 7050, loss[loss=0.2456, simple_loss=0.3278, pruned_loss=0.08166, over 16711.00 frames. ], tot_loss[loss=0.26, simple_loss=0.3351, pruned_loss=0.09246, over 3094687.09 frames. ], batch size: 124, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:40:26,884 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.583e+02 3.724e+02 4.498e+02 5.703e+02 1.501e+03, threshold=8.996e+02, percent-clipped=4.0 2023-04-28 08:41:26,196 INFO [train.py:904] (0/8) Epoch 5, batch 7100, loss[loss=0.2927, simple_loss=0.3427, pruned_loss=0.1213, over 11740.00 frames. ], tot_loss[loss=0.2597, simple_loss=0.334, pruned_loss=0.0927, over 3079970.31 frames. ], batch size: 248, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:42:14,283 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8203, 4.1057, 3.8181, 3.9090, 3.5391, 3.7307, 3.7582, 4.0317], device='cuda:0'), covar=tensor([0.0719, 0.0740, 0.1004, 0.0551, 0.0690, 0.1308, 0.0673, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0480, 0.0419, 0.0311, 0.0299, 0.0326, 0.0392, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:42:19,700 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=47735.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:24,182 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:42:44,678 INFO [train.py:904] (0/8) Epoch 5, batch 7150, loss[loss=0.2652, simple_loss=0.3407, pruned_loss=0.0949, over 16265.00 frames. ], tot_loss[loss=0.2568, simple_loss=0.3305, pruned_loss=0.09151, over 3077386.91 frames. ], batch size: 165, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:43:03,970 INFO [optim.py:368] (0/8) 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 08:43:16,852 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 08:43:39,959 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47788.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:43:57,130 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:43:57,741 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 08:44:00,022 INFO [train.py:904] (0/8) Epoch 5, batch 7200, loss[loss=0.2459, simple_loss=0.3253, pruned_loss=0.08328, over 16393.00 frames. ], tot_loss[loss=0.2543, simple_loss=0.3286, pruned_loss=0.08994, over 3070299.71 frames. ], batch size: 146, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:44:13,327 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=47809.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 08:45:21,342 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7472, 3.6354, 3.8022, 3.6664, 3.7252, 4.1017, 3.8431, 3.5626], device='cuda:0'), covar=tensor([0.1717, 0.1743, 0.1350, 0.2095, 0.2474, 0.1367, 0.1308, 0.2718], device='cuda:0'), in_proj_covar=tensor([0.0276, 0.0374, 0.0382, 0.0330, 0.0436, 0.0404, 0.0310, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:45:21,479 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:45:24,184 INFO [train.py:904] (0/8) Epoch 5, batch 7250, loss[loss=0.227, simple_loss=0.3056, pruned_loss=0.07416, over 16895.00 frames. ], tot_loss[loss=0.2523, simple_loss=0.3268, pruned_loss=0.08892, over 3058369.65 frames. ], batch size: 96, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:45:43,493 INFO [optim.py:368] (0/8) 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 08:45:53,807 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=47870.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 08:45:55,190 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 08:46:12,574 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2469, 3.1806, 3.1168, 3.3918, 3.3937, 3.1607, 3.3373, 3.4175], device='cuda:0'), covar=tensor([0.0817, 0.0669, 0.1352, 0.0640, 0.0700, 0.1996, 0.0850, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0461, 0.0587, 0.0472, 0.0355, 0.0348, 0.0372, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:46:34,259 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=47897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:46:40,689 INFO [train.py:904] (0/8) Epoch 5, batch 7300, loss[loss=0.285, simple_loss=0.3357, pruned_loss=0.1172, over 11672.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3261, pruned_loss=0.089, over 3045867.42 frames. ], batch size: 248, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:47:49,461 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=47945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:47:58,206 INFO [train.py:904] (0/8) Epoch 5, batch 7350, loss[loss=0.2334, simple_loss=0.3159, pruned_loss=0.07541, over 16751.00 frames. ], tot_loss[loss=0.2516, simple_loss=0.3259, pruned_loss=0.08864, over 3054132.49 frames. ], batch size: 83, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:48:17,667 INFO [optim.py:368] (0/8) 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 08:49:13,824 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-48000.pt 2023-04-28 08:49:18,364 INFO [train.py:904] (0/8) Epoch 5, batch 7400, loss[loss=0.2557, simple_loss=0.3328, pruned_loss=0.0893, over 16629.00 frames. ], tot_loss[loss=0.2529, simple_loss=0.3272, pruned_loss=0.08936, over 3064746.74 frames. ], batch size: 62, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:50:12,582 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48035.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:50:38,176 INFO [train.py:904] (0/8) Epoch 5, batch 7450, loss[loss=0.2357, simple_loss=0.3143, pruned_loss=0.07856, over 17045.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3274, pruned_loss=0.08957, over 3075235.24 frames. ], batch size: 55, lr: 1.31e-02, grad_scale: 8.0 2023-04-28 08:51:00,816 INFO [optim.py:368] (0/8) 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 08:51:14,598 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:27,346 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6338, 2.5396, 2.3309, 3.6907, 2.7341, 3.6056, 1.5305, 2.8198], device='cuda:0'), covar=tensor([0.1569, 0.0631, 0.1180, 0.0109, 0.0277, 0.0382, 0.1638, 0.0767], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0144, 0.0168, 0.0085, 0.0180, 0.0178, 0.0160, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 08:51:31,408 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:51:49,940 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48094.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:52:00,149 INFO [train.py:904] (0/8) Epoch 5, batch 7500, loss[loss=0.2562, simple_loss=0.3312, pruned_loss=0.09061, over 15299.00 frames. ], tot_loss[loss=0.2534, simple_loss=0.3285, pruned_loss=0.08917, over 3083389.13 frames. ], batch size: 191, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:52:25,789 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 08:52:47,782 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5251, 4.7719, 4.8529, 4.8332, 4.8493, 5.4046, 4.9157, 4.7350], device='cuda:0'), covar=tensor([0.0893, 0.1588, 0.1427, 0.1505, 0.2206, 0.0872, 0.1182, 0.2042], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0376, 0.0383, 0.0333, 0.0436, 0.0400, 0.0312, 0.0452], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:52:50,918 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48133.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:08,129 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:53:18,039 INFO [train.py:904] (0/8) Epoch 5, batch 7550, loss[loss=0.2404, simple_loss=0.3221, pruned_loss=0.07931, over 15374.00 frames. ], tot_loss[loss=0.2526, simple_loss=0.3273, pruned_loss=0.08898, over 3083747.87 frames. ], batch size: 190, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:53:38,523 INFO [optim.py:368] (0/8) 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 08:53:40,656 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48165.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 08:54:34,596 INFO [train.py:904] (0/8) Epoch 5, batch 7600, loss[loss=0.2495, simple_loss=0.323, pruned_loss=0.08803, over 16739.00 frames. ], tot_loss[loss=0.2528, simple_loss=0.3269, pruned_loss=0.08936, over 3075652.09 frames. ], batch size: 134, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:54:51,927 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48212.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:55:19,681 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1178, 3.9149, 3.5010, 1.9301, 2.8301, 2.3216, 3.5250, 3.5898], device='cuda:0'), covar=tensor([0.0279, 0.0444, 0.0525, 0.1593, 0.0742, 0.0897, 0.0606, 0.0740], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0128, 0.0155, 0.0141, 0.0134, 0.0126, 0.0142, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 08:55:51,253 INFO [train.py:904] (0/8) Epoch 5, batch 7650, loss[loss=0.2819, simple_loss=0.3485, pruned_loss=0.1077, over 16165.00 frames. ], tot_loss[loss=0.2551, simple_loss=0.3286, pruned_loss=0.09078, over 3071871.89 frames. ], batch size: 165, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 08:55:59,537 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6044, 3.4477, 3.6614, 3.5602, 3.6217, 4.0432, 3.7926, 3.4756], device='cuda:0'), covar=tensor([0.1946, 0.2158, 0.1722, 0.2183, 0.2808, 0.1487, 0.1336, 0.2792], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0379, 0.0386, 0.0334, 0.0440, 0.0401, 0.0311, 0.0452], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 08:56:12,528 INFO [optim.py:368] (0/8) 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 08:56:26,017 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48273.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 08:56:59,243 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4772, 2.0870, 2.2090, 4.2176, 1.9614, 2.8085, 2.2613, 2.1495], device='cuda:0'), covar=tensor([0.0715, 0.2312, 0.1210, 0.0338, 0.3260, 0.1389, 0.2063, 0.2755], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0320, 0.0262, 0.0309, 0.0375, 0.0310, 0.0287, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:57:08,665 INFO [train.py:904] (0/8) Epoch 5, batch 7700, loss[loss=0.2447, simple_loss=0.3254, pruned_loss=0.08199, over 16200.00 frames. ], tot_loss[loss=0.2561, simple_loss=0.3292, pruned_loss=0.09152, over 3058277.90 frames. ], batch size: 165, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:57:44,132 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48324.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:57:52,456 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8527, 4.8081, 4.5949, 3.9222, 4.6425, 1.7753, 4.3999, 4.5038], device='cuda:0'), covar=tensor([0.0058, 0.0044, 0.0077, 0.0323, 0.0053, 0.1754, 0.0075, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0077, 0.0119, 0.0124, 0.0090, 0.0141, 0.0104, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:57:56,452 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 08:58:26,518 INFO [train.py:904] (0/8) Epoch 5, batch 7750, loss[loss=0.2896, simple_loss=0.3411, pruned_loss=0.119, over 11634.00 frames. ], tot_loss[loss=0.2563, simple_loss=0.3294, pruned_loss=0.09156, over 3058417.03 frames. ], batch size: 246, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 08:58:47,830 INFO [optim.py:368] (0/8) 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 08:59:20,048 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:28,869 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1343, 3.9201, 4.1293, 4.4105, 4.4789, 4.0353, 4.4691, 4.4214], device='cuda:0'), covar=tensor([0.1007, 0.0783, 0.1349, 0.0505, 0.0482, 0.0897, 0.0458, 0.0488], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0462, 0.0585, 0.0479, 0.0354, 0.0349, 0.0374, 0.0396], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 08:59:34,677 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:39,603 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48397.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 08:59:45,054 INFO [train.py:904] (0/8) Epoch 5, batch 7800, loss[loss=0.3302, simple_loss=0.3691, pruned_loss=0.1456, over 11345.00 frames. ], tot_loss[loss=0.2572, simple_loss=0.33, pruned_loss=0.09215, over 3057843.30 frames. ], batch size: 248, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:00:20,468 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48423.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:28,143 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48428.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:49,117 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48442.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:00:52,198 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48444.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:01:02,298 INFO [train.py:904] (0/8) Epoch 5, batch 7850, loss[loss=0.3043, simple_loss=0.3554, pruned_loss=0.1265, over 11816.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3312, pruned_loss=0.09296, over 3025443.24 frames. ], batch size: 246, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:01:13,133 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48458.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:01:23,605 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.309e+02 3.753e+02 4.508e+02 5.540e+02 1.129e+03, threshold=9.017e+02, percent-clipped=6.0 2023-04-28 09:01:24,005 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48465.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:01:51,259 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1910, 1.3384, 1.7583, 2.0166, 2.1501, 2.2864, 1.4822, 2.0618], device='cuda:0'), covar=tensor([0.0075, 0.0220, 0.0125, 0.0124, 0.0100, 0.0076, 0.0201, 0.0043], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0143, 0.0125, 0.0122, 0.0129, 0.0092, 0.0140, 0.0079], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 09:01:52,515 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48484.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:04,084 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48492.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:02:16,629 INFO [train.py:904] (0/8) Epoch 5, batch 7900, loss[loss=0.2327, simple_loss=0.3111, pruned_loss=0.07716, over 16555.00 frames. ], tot_loss[loss=0.2573, simple_loss=0.3301, pruned_loss=0.09221, over 3026864.60 frames. ], batch size: 68, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:02:18,977 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8317, 2.5955, 2.6213, 1.8876, 2.5145, 2.6137, 2.5774, 1.6453], device='cuda:0'), covar=tensor([0.0261, 0.0041, 0.0047, 0.0189, 0.0064, 0.0082, 0.0048, 0.0329], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0053, 0.0058, 0.0114, 0.0060, 0.0069, 0.0062, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 09:02:35,255 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48513.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:03:17,070 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7755, 1.2011, 1.5073, 1.6786, 1.8408, 1.8818, 1.4583, 1.6922], device='cuda:0'), covar=tensor([0.0079, 0.0167, 0.0092, 0.0118, 0.0087, 0.0063, 0.0147, 0.0040], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0144, 0.0125, 0.0122, 0.0129, 0.0093, 0.0141, 0.0079], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 09:03:36,606 INFO [train.py:904] (0/8) Epoch 5, batch 7950, loss[loss=0.2316, simple_loss=0.3082, pruned_loss=0.07749, over 16657.00 frames. ], tot_loss[loss=0.2579, simple_loss=0.3303, pruned_loss=0.09279, over 3015856.46 frames. ], batch size: 57, lr: 1.30e-02, grad_scale: 4.0 2023-04-28 09:03:56,965 INFO [optim.py:368] (0/8) 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 09:03:59,580 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3140, 4.1262, 3.6351, 2.0332, 3.1872, 2.5433, 3.6950, 3.8800], device='cuda:0'), covar=tensor([0.0274, 0.0408, 0.0508, 0.1682, 0.0727, 0.0952, 0.0625, 0.0765], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0127, 0.0155, 0.0140, 0.0134, 0.0126, 0.0141, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 09:04:02,297 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48568.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:04:50,729 INFO [train.py:904] (0/8) Epoch 5, batch 8000, loss[loss=0.2195, simple_loss=0.3053, pruned_loss=0.06684, over 16907.00 frames. ], tot_loss[loss=0.2585, simple_loss=0.3307, pruned_loss=0.09319, over 3022561.82 frames. ], batch size: 96, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:05:25,764 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-28 09:06:06,087 INFO [train.py:904] (0/8) Epoch 5, batch 8050, loss[loss=0.2854, simple_loss=0.3445, pruned_loss=0.1132, over 11805.00 frames. ], tot_loss[loss=0.2591, simple_loss=0.331, pruned_loss=0.09362, over 3006660.85 frames. ], batch size: 248, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:06:26,964 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.628e+02 3.970e+02 4.648e+02 5.523e+02 1.617e+03, threshold=9.296e+02, percent-clipped=6.0 2023-04-28 09:06:49,660 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:07:21,599 INFO [train.py:904] (0/8) Epoch 5, batch 8100, loss[loss=0.2264, simple_loss=0.3001, pruned_loss=0.07633, over 16864.00 frames. ], tot_loss[loss=0.2574, simple_loss=0.3297, pruned_loss=0.09255, over 3001078.22 frames. ], batch size: 116, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:07:23,838 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2420, 3.1941, 3.2283, 3.4330, 3.4262, 3.1775, 3.3884, 3.4458], device='cuda:0'), covar=tensor([0.0824, 0.0695, 0.1130, 0.0553, 0.0636, 0.1588, 0.0797, 0.0596], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0473, 0.0592, 0.0490, 0.0367, 0.0352, 0.0382, 0.0411], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:08:05,465 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48728.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:08:40,602 INFO [train.py:904] (0/8) Epoch 5, batch 8150, loss[loss=0.2318, simple_loss=0.306, pruned_loss=0.07885, over 16542.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3262, pruned_loss=0.09022, over 3028498.28 frames. ], batch size: 68, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:08:44,096 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48753.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:01,951 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.147e+02 3.868e+02 4.692e+02 5.916e+02 1.218e+03, threshold=9.385e+02, percent-clipped=3.0 2023-04-28 09:09:20,847 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:09:24,941 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=48779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:10:00,541 INFO [train.py:904] (0/8) Epoch 5, batch 8200, loss[loss=0.2507, simple_loss=0.3196, pruned_loss=0.09094, over 16585.00 frames. ], tot_loss[loss=0.2496, simple_loss=0.3227, pruned_loss=0.0883, over 3060699.03 frames. ], batch size: 62, lr: 1.30e-02, grad_scale: 8.0 2023-04-28 09:10:32,482 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:10:50,523 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3154, 3.5568, 3.6361, 1.6437, 3.8458, 3.9380, 3.0731, 3.0320], device='cuda:0'), covar=tensor([0.0671, 0.0126, 0.0208, 0.1115, 0.0048, 0.0049, 0.0263, 0.0300], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0089, 0.0083, 0.0143, 0.0070, 0.0077, 0.0116, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 09:11:00,867 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=48838.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:11:22,119 INFO [train.py:904] (0/8) Epoch 5, batch 8250, loss[loss=0.2357, simple_loss=0.3232, pruned_loss=0.07407, over 16441.00 frames. ], tot_loss[loss=0.247, simple_loss=0.322, pruned_loss=0.08598, over 3051571.57 frames. ], batch size: 68, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:11:44,536 INFO [optim.py:368] (0/8) 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 09:11:50,753 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48868.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:12:13,776 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48882.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:40,596 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=48899.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:12:43,302 INFO [train.py:904] (0/8) Epoch 5, batch 8300, loss[loss=0.2011, simple_loss=0.2889, pruned_loss=0.05666, over 16302.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3181, pruned_loss=0.08186, over 3061804.23 frames. ], batch size: 35, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:13:08,660 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=48916.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:14:06,484 INFO [train.py:904] (0/8) Epoch 5, batch 8350, loss[loss=0.2426, simple_loss=0.3267, pruned_loss=0.07924, over 15188.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3158, pruned_loss=0.07873, over 3051154.75 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:14:30,523 INFO [optim.py:368] (0/8) 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 09:14:54,818 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=48980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:14:57,252 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 09:15:30,227 INFO [train.py:904] (0/8) Epoch 5, batch 8400, loss[loss=0.2009, simple_loss=0.2878, pruned_loss=0.05696, over 15450.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3123, pruned_loss=0.07591, over 3039412.48 frames. ], batch size: 190, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:14,936 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49028.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:16:51,008 INFO [train.py:904] (0/8) Epoch 5, batch 8450, loss[loss=0.2196, simple_loss=0.3052, pruned_loss=0.06696, over 16327.00 frames. ], tot_loss[loss=0.2287, simple_loss=0.3102, pruned_loss=0.07361, over 3036282.90 frames. ], batch size: 146, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:16:54,817 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:17:13,778 INFO [optim.py:368] (0/8) 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 09:17:19,693 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.23 vs. limit=5.0 2023-04-28 09:17:38,472 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49079.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:17:39,689 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4947, 4.6747, 4.6928, 4.6553, 4.6668, 5.1801, 4.8029, 4.4789], device='cuda:0'), covar=tensor([0.0993, 0.1315, 0.1464, 0.1534, 0.2301, 0.0876, 0.1094, 0.2209], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0362, 0.0365, 0.0315, 0.0419, 0.0383, 0.0298, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:18:13,194 INFO [train.py:904] (0/8) Epoch 5, batch 8500, loss[loss=0.2059, simple_loss=0.2903, pruned_loss=0.06078, over 16175.00 frames. ], tot_loss[loss=0.2232, simple_loss=0.3056, pruned_loss=0.07035, over 3032388.36 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:18:13,683 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:18:52,114 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4384, 3.3940, 2.8196, 2.0777, 2.2230, 2.1180, 3.4888, 3.4322], device='cuda:0'), covar=tensor([0.2143, 0.0597, 0.1124, 0.1678, 0.1930, 0.1473, 0.0374, 0.0622], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0235, 0.0257, 0.0236, 0.0262, 0.0196, 0.0231, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:18:57,913 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49127.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:19:38,548 INFO [train.py:904] (0/8) Epoch 5, batch 8550, loss[loss=0.207, simple_loss=0.2937, pruned_loss=0.06019, over 16553.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3027, pruned_loss=0.06885, over 3022705.73 frames. ], batch size: 62, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:20:04,111 INFO [optim.py:368] (0/8) 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 09:20:05,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 09:20:29,000 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:04,493 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49194.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:21:17,323 INFO [train.py:904] (0/8) Epoch 5, batch 8600, loss[loss=0.2103, simple_loss=0.306, pruned_loss=0.05731, over 16331.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3025, pruned_loss=0.06753, over 3029507.26 frames. ], batch size: 146, lr: 1.29e-02, grad_scale: 8.0 2023-04-28 09:22:28,823 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49236.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:22:58,658 INFO [train.py:904] (0/8) Epoch 5, batch 8650, loss[loss=0.2088, simple_loss=0.289, pruned_loss=0.0643, over 12589.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2998, pruned_loss=0.06501, over 3040858.73 frames. ], batch size: 247, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:23:33,946 INFO [optim.py:368] (0/8) 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 09:23:35,332 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9320, 4.8913, 4.6825, 4.2083, 4.6812, 1.7049, 4.4670, 4.6083], device='cuda:0'), covar=tensor([0.0046, 0.0038, 0.0085, 0.0229, 0.0056, 0.1816, 0.0085, 0.0120], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0076, 0.0121, 0.0118, 0.0089, 0.0143, 0.0104, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:24:38,802 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49297.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:24:41,152 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7019, 3.5581, 3.8211, 3.9345, 3.9871, 3.5537, 3.9997, 4.0184], device='cuda:0'), covar=tensor([0.0914, 0.0755, 0.0953, 0.0507, 0.0476, 0.1332, 0.0495, 0.0409], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0446, 0.0550, 0.0465, 0.0348, 0.0342, 0.0363, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:24:47,028 INFO [train.py:904] (0/8) Epoch 5, batch 8700, loss[loss=0.2322, simple_loss=0.3079, pruned_loss=0.07829, over 12229.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2965, pruned_loss=0.06338, over 3046335.02 frames. ], batch size: 248, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:25:00,392 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9820, 3.9206, 3.8458, 3.1287, 3.8420, 1.6613, 3.6272, 3.6140], device='cuda:0'), covar=tensor([0.0090, 0.0069, 0.0108, 0.0338, 0.0082, 0.2089, 0.0119, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0075, 0.0119, 0.0115, 0.0088, 0.0141, 0.0103, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:25:33,364 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3518, 3.6039, 3.7686, 1.7414, 3.8714, 3.9124, 3.1543, 2.8627], device='cuda:0'), covar=tensor([0.0699, 0.0125, 0.0103, 0.1085, 0.0042, 0.0047, 0.0254, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0085, 0.0077, 0.0137, 0.0066, 0.0073, 0.0112, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 09:26:24,159 INFO [train.py:904] (0/8) Epoch 5, batch 8750, loss[loss=0.2208, simple_loss=0.3114, pruned_loss=0.06513, over 16514.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2963, pruned_loss=0.06304, over 3034867.86 frames. ], batch size: 62, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:27:05,492 INFO [optim.py:368] (0/8) 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 09:27:22,702 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8268, 3.5480, 3.2775, 1.9393, 2.8650, 2.2257, 3.2083, 3.2885], device='cuda:0'), covar=tensor([0.0247, 0.0438, 0.0458, 0.1471, 0.0627, 0.0935, 0.0682, 0.0626], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0117, 0.0147, 0.0136, 0.0128, 0.0124, 0.0133, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 09:27:47,434 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1340, 3.0465, 3.1434, 1.6766, 3.2514, 3.3179, 2.8710, 2.5369], device='cuda:0'), covar=tensor([0.0708, 0.0172, 0.0134, 0.1126, 0.0068, 0.0071, 0.0321, 0.0422], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0086, 0.0077, 0.0139, 0.0067, 0.0074, 0.0113, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 09:28:17,603 INFO [train.py:904] (0/8) Epoch 5, batch 8800, loss[loss=0.2504, simple_loss=0.3302, pruned_loss=0.08528, over 15409.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2949, pruned_loss=0.0617, over 3061670.44 frames. ], batch size: 191, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:29:01,955 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:29:24,461 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49432.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:30:04,518 INFO [train.py:904] (0/8) Epoch 5, batch 8850, loss[loss=0.2098, simple_loss=0.3052, pruned_loss=0.0572, over 16813.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2962, pruned_loss=0.06092, over 3035408.98 frames. ], batch size: 124, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:30:22,449 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 09:30:38,672 INFO [optim.py:368] (0/8) 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 09:31:02,531 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:10,515 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5461, 4.5267, 4.9638, 4.9929, 4.9519, 4.5973, 4.6680, 4.3181], device='cuda:0'), covar=tensor([0.0206, 0.0280, 0.0329, 0.0303, 0.0305, 0.0214, 0.0570, 0.0289], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0208, 0.0213, 0.0212, 0.0259, 0.0227, 0.0314, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-04-28 09:31:16,457 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49483.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:36,720 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49493.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:31:39,720 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:31:54,246 INFO [train.py:904] (0/8) Epoch 5, batch 8900, loss[loss=0.2012, simple_loss=0.2889, pruned_loss=0.05676, over 16609.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2962, pruned_loss=0.05972, over 3031295.05 frames. ], batch size: 62, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:32:48,132 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49525.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:39,424 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:33:52,458 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-04-28 09:33:59,797 INFO [train.py:904] (0/8) Epoch 5, batch 8950, loss[loss=0.1891, simple_loss=0.2789, pruned_loss=0.04969, over 16201.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2962, pruned_loss=0.06006, over 3062855.49 frames. ], batch size: 165, lr: 1.29e-02, grad_scale: 4.0 2023-04-28 09:34:35,635 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.980e+02 2.796e+02 3.504e+02 4.655e+02 7.441e+02, threshold=7.007e+02, percent-clipped=0.0 2023-04-28 09:35:32,783 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49592.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:35:51,815 INFO [train.py:904] (0/8) Epoch 5, batch 9000, loss[loss=0.2, simple_loss=0.2833, pruned_loss=0.05836, over 11842.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2932, pruned_loss=0.05863, over 3060770.58 frames. ], batch size: 246, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:35:51,817 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 09:36:02,101 INFO [train.py:938] (0/8) Epoch 5, validation: loss=0.1735, simple_loss=0.2766, pruned_loss=0.0352, over 944034.00 frames. 2023-04-28 09:36:02,102 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 09:36:46,142 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 09:37:44,750 INFO [train.py:904] (0/8) Epoch 5, batch 9050, loss[loss=0.2425, simple_loss=0.3111, pruned_loss=0.08696, over 12759.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2945, pruned_loss=0.05953, over 3069544.05 frames. ], batch size: 246, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:38:18,686 INFO [optim.py:368] (0/8) 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 09:38:24,132 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.13 vs. limit=2.0 2023-04-28 09:38:30,389 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 09:39:29,283 INFO [train.py:904] (0/8) Epoch 5, batch 9100, loss[loss=0.2188, simple_loss=0.2914, pruned_loss=0.07306, over 12490.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2942, pruned_loss=0.06039, over 3075060.34 frames. ], batch size: 246, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:39:30,731 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 09:40:25,915 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.65 vs. limit=2.0 2023-04-28 09:41:26,346 INFO [train.py:904] (0/8) Epoch 5, batch 9150, loss[loss=0.1702, simple_loss=0.2658, pruned_loss=0.03728, over 16806.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2944, pruned_loss=0.05982, over 3058298.23 frames. ], batch size: 76, lr: 1.28e-02, grad_scale: 4.0 2023-04-28 09:42:00,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=49766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:01,428 INFO [optim.py:368] (0/8) 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 09:42:25,172 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:42:47,422 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=49788.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 09:43:02,285 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 09:43:09,766 INFO [train.py:904] (0/8) Epoch 5, batch 9200, loss[loss=0.1758, simple_loss=0.2566, pruned_loss=0.04752, over 12091.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2897, pruned_loss=0.0588, over 3042263.85 frames. ], batch size: 248, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:43:58,006 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=49827.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:44:46,688 INFO [train.py:904] (0/8) Epoch 5, batch 9250, loss[loss=0.2001, simple_loss=0.2889, pruned_loss=0.0556, over 15433.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2897, pruned_loss=0.05923, over 3032183.76 frames. ], batch size: 192, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:45:00,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5507, 3.4546, 3.3926, 2.9550, 3.4309, 1.9732, 3.2391, 2.9355], device='cuda:0'), covar=tensor([0.0085, 0.0066, 0.0112, 0.0192, 0.0072, 0.1568, 0.0102, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0089, 0.0076, 0.0119, 0.0112, 0.0088, 0.0140, 0.0101, 0.0111], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:45:18,929 INFO [optim.py:368] (0/8) 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 09:45:25,808 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6795, 3.6591, 3.8050, 3.9584, 4.0330, 3.5289, 3.9934, 4.0407], device='cuda:0'), covar=tensor([0.0939, 0.0680, 0.0988, 0.0443, 0.0409, 0.1393, 0.0499, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0433, 0.0546, 0.0448, 0.0341, 0.0336, 0.0358, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:45:49,747 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-28 09:46:21,042 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=49892.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:46:40,059 INFO [train.py:904] (0/8) Epoch 5, batch 9300, loss[loss=0.2041, simple_loss=0.2759, pruned_loss=0.06616, over 12321.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2877, pruned_loss=0.05846, over 3016225.83 frames. ], batch size: 247, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:48:05,857 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=49940.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:48:26,420 INFO [train.py:904] (0/8) Epoch 5, batch 9350, loss[loss=0.1964, simple_loss=0.2871, pruned_loss=0.05281, over 16841.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2881, pruned_loss=0.05843, over 3030497.95 frames. ], batch size: 90, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:49:00,530 INFO [optim.py:368] (0/8) 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 09:50:06,886 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-50000.pt 2023-04-28 09:50:11,394 INFO [train.py:904] (0/8) Epoch 5, batch 9400, loss[loss=0.2013, simple_loss=0.3037, pruned_loss=0.04948, over 16670.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2875, pruned_loss=0.05785, over 3029415.77 frames. ], batch size: 83, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:51:01,535 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4525, 1.3855, 1.8452, 2.2631, 2.2541, 2.5662, 1.4824, 2.2527], device='cuda:0'), covar=tensor([0.0080, 0.0247, 0.0156, 0.0137, 0.0120, 0.0077, 0.0260, 0.0053], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0141, 0.0127, 0.0122, 0.0127, 0.0088, 0.0139, 0.0075], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 09:51:51,665 INFO [train.py:904] (0/8) Epoch 5, batch 9450, loss[loss=0.2218, simple_loss=0.3053, pruned_loss=0.06912, over 15412.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2898, pruned_loss=0.05858, over 3034993.83 frames. ], batch size: 191, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:51:52,422 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9463, 2.0126, 2.3190, 3.2724, 1.9761, 2.6211, 2.2980, 2.0250], device='cuda:0'), covar=tensor([0.0554, 0.1957, 0.0932, 0.0343, 0.2785, 0.0982, 0.1780, 0.2287], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0304, 0.0258, 0.0302, 0.0363, 0.0305, 0.0285, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:52:18,604 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3829, 1.4422, 1.8045, 2.2268, 2.2308, 2.3626, 1.4180, 2.1564], device='cuda:0'), covar=tensor([0.0080, 0.0252, 0.0171, 0.0146, 0.0124, 0.0090, 0.0257, 0.0055], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0141, 0.0128, 0.0122, 0.0127, 0.0088, 0.0139, 0.0075], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 09:52:21,842 INFO [optim.py:368] (0/8) 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 09:52:47,395 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:53:09,031 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50088.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:53:33,293 INFO [train.py:904] (0/8) Epoch 5, batch 9500, loss[loss=0.2061, simple_loss=0.29, pruned_loss=0.06112, over 16759.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2893, pruned_loss=0.05825, over 3032543.39 frames. ], batch size: 39, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:53:58,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9046, 4.1877, 3.9598, 3.9914, 3.6330, 3.7556, 3.8712, 4.1439], device='cuda:0'), covar=tensor([0.0764, 0.0765, 0.0826, 0.0510, 0.0638, 0.1298, 0.0578, 0.0798], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0467, 0.0382, 0.0302, 0.0292, 0.0308, 0.0377, 0.0336], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:54:17,577 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:25,577 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:29,289 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50128.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:54:44,420 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50136.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:55:18,514 INFO [train.py:904] (0/8) Epoch 5, batch 9550, loss[loss=0.2311, simple_loss=0.3163, pruned_loss=0.073, over 15324.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2894, pruned_loss=0.05838, over 3051560.68 frames. ], batch size: 190, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:55:52,509 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0903, 4.2296, 4.2416, 4.1823, 4.2500, 4.7086, 4.3481, 4.1181], device='cuda:0'), covar=tensor([0.1360, 0.1584, 0.1210, 0.1755, 0.2392, 0.0982, 0.1046, 0.2067], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0356, 0.0353, 0.0309, 0.0408, 0.0384, 0.0294, 0.0414], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 09:55:53,362 INFO [optim.py:368] (0/8) 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 09:56:38,118 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50189.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 09:56:52,833 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50197.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 09:56:59,053 INFO [train.py:904] (0/8) Epoch 5, batch 9600, loss[loss=0.2135, simple_loss=0.2866, pruned_loss=0.0702, over 12647.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2911, pruned_loss=0.05947, over 3041182.64 frames. ], batch size: 250, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:57:25,463 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 09:57:25,598 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 09:58:37,877 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7009, 2.7754, 1.6758, 2.7721, 2.0804, 2.8144, 1.8744, 2.3903], device='cuda:0'), covar=tensor([0.0156, 0.0339, 0.1306, 0.0097, 0.0776, 0.0439, 0.1290, 0.0640], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0144, 0.0174, 0.0074, 0.0156, 0.0168, 0.0181, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 09:58:46,680 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 09:58:47,432 INFO [train.py:904] (0/8) Epoch 5, batch 9650, loss[loss=0.1919, simple_loss=0.2785, pruned_loss=0.05264, over 12260.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2935, pruned_loss=0.06, over 3033794.08 frames. ], batch size: 250, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 09:59:05,816 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50258.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 09:59:27,462 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.258e+02 2.991e+02 3.657e+02 4.626e+02 9.582e+02, threshold=7.315e+02, percent-clipped=7.0 2023-04-28 10:00:35,793 INFO [train.py:904] (0/8) Epoch 5, batch 9700, loss[loss=0.1989, simple_loss=0.2884, pruned_loss=0.0547, over 16919.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2921, pruned_loss=0.05952, over 3025074.84 frames. ], batch size: 109, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:01:51,309 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 10:02:18,556 INFO [train.py:904] (0/8) Epoch 5, batch 9750, loss[loss=0.209, simple_loss=0.2992, pruned_loss=0.05937, over 16208.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2915, pruned_loss=0.0595, over 3040860.13 frames. ], batch size: 165, lr: 1.28e-02, grad_scale: 8.0 2023-04-28 10:02:50,596 INFO [optim.py:368] (0/8) 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 10:03:56,539 INFO [train.py:904] (0/8) Epoch 5, batch 9800, loss[loss=0.1997, simple_loss=0.2962, pruned_loss=0.05155, over 15263.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2915, pruned_loss=0.05803, over 3061099.97 frames. ], batch size: 190, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:04:36,903 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:05:00,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1974, 3.3353, 3.3354, 2.5030, 3.1630, 3.2783, 3.2434, 1.9715], device='cuda:0'), covar=tensor([0.0279, 0.0021, 0.0032, 0.0182, 0.0049, 0.0045, 0.0035, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0054, 0.0058, 0.0114, 0.0059, 0.0064, 0.0062, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:05:07,408 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5666, 2.4768, 1.9426, 2.2079, 2.9693, 2.8110, 3.4959, 3.1912], device='cuda:0'), covar=tensor([0.0022, 0.0220, 0.0272, 0.0249, 0.0128, 0.0182, 0.0065, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0072, 0.0156, 0.0156, 0.0152, 0.0148, 0.0154, 0.0121, 0.0135], device='cuda:0'), out_proj_covar=tensor([8.6206e-05, 1.8749e-04, 1.8429e-04, 1.8113e-04, 1.8035e-04, 1.8528e-04, 1.3845e-04, 1.6277e-04], device='cuda:0') 2023-04-28 10:05:41,985 INFO [train.py:904] (0/8) Epoch 5, batch 9850, loss[loss=0.1832, simple_loss=0.268, pruned_loss=0.04918, over 12439.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2924, pruned_loss=0.05744, over 3067917.64 frames. ], batch size: 247, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:06:14,681 INFO [optim.py:368] (0/8) 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 10:06:21,305 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50470.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:06:52,456 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50484.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:07:20,528 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7378, 3.6014, 3.8139, 3.9541, 4.0343, 3.5822, 4.0000, 4.0253], device='cuda:0'), covar=tensor([0.0839, 0.0695, 0.1067, 0.0508, 0.0398, 0.1248, 0.0475, 0.0385], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0441, 0.0552, 0.0458, 0.0339, 0.0339, 0.0360, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:07:31,484 INFO [train.py:904] (0/8) Epoch 5, batch 9900, loss[loss=0.2197, simple_loss=0.3095, pruned_loss=0.06501, over 15401.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2928, pruned_loss=0.0578, over 3041573.06 frames. ], batch size: 192, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:09:17,009 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50546.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:09:27,685 INFO [train.py:904] (0/8) Epoch 5, batch 9950, loss[loss=0.22, simple_loss=0.3088, pruned_loss=0.06557, over 16649.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2941, pruned_loss=0.05794, over 3037089.70 frames. ], batch size: 134, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:09:33,679 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50553.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:10:04,529 INFO [optim.py:368] (0/8) 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 10:10:42,397 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 10:11:22,255 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2671, 5.0561, 5.1155, 4.8554, 4.5251, 5.1410, 5.1308, 4.7715], device='cuda:0'), covar=tensor([0.0413, 0.0379, 0.0180, 0.0144, 0.0987, 0.0296, 0.0134, 0.0436], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0176, 0.0198, 0.0166, 0.0220, 0.0198, 0.0137, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:11:29,341 INFO [train.py:904] (0/8) Epoch 5, batch 10000, loss[loss=0.2061, simple_loss=0.2866, pruned_loss=0.06278, over 12888.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2924, pruned_loss=0.05741, over 3042665.49 frames. ], batch size: 250, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:11:43,847 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50607.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:13:08,524 INFO [train.py:904] (0/8) Epoch 5, batch 10050, loss[loss=0.2166, simple_loss=0.3081, pruned_loss=0.06256, over 12158.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2918, pruned_loss=0.05675, over 3033179.10 frames. ], batch size: 248, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:13:38,886 INFO [optim.py:368] (0/8) 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 10:13:39,751 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3675, 3.2996, 3.3698, 3.5177, 3.5472, 3.2674, 3.5654, 3.5765], device='cuda:0'), covar=tensor([0.0664, 0.0573, 0.0849, 0.0457, 0.0462, 0.1473, 0.0557, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0441, 0.0553, 0.0456, 0.0340, 0.0331, 0.0356, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:14:01,388 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 10:14:06,886 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9002, 3.0010, 2.9976, 2.0780, 2.9284, 2.9728, 2.9126, 1.6124], device='cuda:0'), covar=tensor([0.0325, 0.0034, 0.0040, 0.0255, 0.0068, 0.0097, 0.0093, 0.0396], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0054, 0.0058, 0.0114, 0.0060, 0.0065, 0.0063, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:14:22,795 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4410, 1.9878, 2.1774, 3.9288, 1.8381, 2.7300, 2.2668, 2.1100], device='cuda:0'), covar=tensor([0.0585, 0.2318, 0.1239, 0.0304, 0.3050, 0.1170, 0.1956, 0.2578], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0304, 0.0259, 0.0297, 0.0356, 0.0306, 0.0284, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:14:39,006 INFO [train.py:904] (0/8) Epoch 5, batch 10100, loss[loss=0.1921, simple_loss=0.2761, pruned_loss=0.05409, over 16648.00 frames. ], tot_loss[loss=0.203, simple_loss=0.292, pruned_loss=0.05699, over 3044703.20 frames. ], batch size: 62, lr: 1.27e-02, grad_scale: 8.0 2023-04-28 10:15:56,040 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-5.pt 2023-04-28 10:16:20,275 INFO [train.py:904] (0/8) Epoch 6, batch 0, loss[loss=0.2639, simple_loss=0.3349, pruned_loss=0.09643, over 17031.00 frames. ], tot_loss[loss=0.2639, simple_loss=0.3349, pruned_loss=0.09643, over 17031.00 frames. ], batch size: 50, lr: 1.19e-02, grad_scale: 8.0 2023-04-28 10:16:20,275 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 10:16:27,645 INFO [train.py:938] (0/8) Epoch 6, validation: loss=0.1727, simple_loss=0.2755, pruned_loss=0.03501, over 944034.00 frames. 2023-04-28 10:16:27,646 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 10:16:50,531 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2429, 4.1208, 4.1388, 4.0419, 3.7624, 4.2196, 3.9367, 3.8877], device='cuda:0'), covar=tensor([0.0537, 0.0404, 0.0212, 0.0171, 0.0761, 0.0303, 0.0422, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0175, 0.0196, 0.0167, 0.0221, 0.0200, 0.0138, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:16:52,398 INFO [optim.py:368] (0/8) 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 10:17:10,810 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:13,010 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50784.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:17:19,728 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6876, 3.4172, 3.2045, 5.0127, 4.6850, 4.5766, 1.9809, 3.3252], device='cuda:0'), covar=tensor([0.1390, 0.0480, 0.0885, 0.0111, 0.0212, 0.0363, 0.1176, 0.0692], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0140, 0.0166, 0.0086, 0.0151, 0.0174, 0.0161, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 10:17:35,105 INFO [train.py:904] (0/8) Epoch 6, batch 50, loss[loss=0.2444, simple_loss=0.306, pruned_loss=0.09139, over 16757.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3203, pruned_loss=0.09028, over 743547.06 frames. ], batch size: 124, lr: 1.19e-02, grad_scale: 2.0 2023-04-28 10:17:35,813 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 10:17:46,454 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50808.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:18:20,209 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=50832.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:23,676 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1286, 5.0598, 4.8536, 4.3032, 4.7634, 1.9817, 4.5929, 4.8433], device='cuda:0'), covar=tensor([0.0053, 0.0043, 0.0080, 0.0223, 0.0067, 0.1641, 0.0090, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0078, 0.0122, 0.0113, 0.0090, 0.0145, 0.0104, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:18:31,170 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9700, 3.8943, 3.8108, 3.6599, 3.5058, 3.8881, 3.6494, 3.6229], device='cuda:0'), covar=tensor([0.0458, 0.0331, 0.0180, 0.0174, 0.0704, 0.0300, 0.0616, 0.0441], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0183, 0.0206, 0.0175, 0.0233, 0.0209, 0.0144, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:18:34,260 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50843.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:18:47,060 INFO [train.py:904] (0/8) Epoch 6, batch 100, loss[loss=0.1977, simple_loss=0.2808, pruned_loss=0.05727, over 17189.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.3123, pruned_loss=0.08337, over 1306135.73 frames. ], batch size: 46, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:18:49,590 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=50853.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:19:11,789 INFO [optim.py:368] (0/8) 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 10:19:12,283 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=50869.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:19:54,851 INFO [train.py:904] (0/8) Epoch 6, batch 150, loss[loss=0.2263, simple_loss=0.311, pruned_loss=0.07085, over 17261.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.3078, pruned_loss=0.0793, over 1753462.98 frames. ], batch size: 52, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:19:55,127 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=50902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:20:56,996 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=50946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:21:05,275 INFO [train.py:904] (0/8) Epoch 6, batch 200, loss[loss=0.2196, simple_loss=0.2883, pruned_loss=0.07547, over 16886.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3056, pruned_loss=0.07813, over 2108741.41 frames. ], batch size: 96, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:21:28,628 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.434e+02 3.943e+02 4.796e+02 1.132e+03, threshold=7.886e+02, percent-clipped=3.0 2023-04-28 10:22:12,673 INFO [train.py:904] (0/8) Epoch 6, batch 250, loss[loss=0.2474, simple_loss=0.3019, pruned_loss=0.0964, over 16707.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3025, pruned_loss=0.07736, over 2380908.83 frames. ], batch size: 124, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:22:20,736 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51007.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:23:20,598 INFO [train.py:904] (0/8) Epoch 6, batch 300, loss[loss=0.2576, simple_loss=0.3056, pruned_loss=0.1048, over 16863.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.2984, pruned_loss=0.07529, over 2589024.34 frames. ], batch size: 96, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:23:37,430 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51062.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:23:45,705 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.797e+02 3.691e+02 4.478e+02 8.058e+02, threshold=7.381e+02, percent-clipped=1.0 2023-04-28 10:24:15,158 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 10:24:30,409 INFO [train.py:904] (0/8) Epoch 6, batch 350, loss[loss=0.2106, simple_loss=0.2811, pruned_loss=0.07003, over 16425.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.2954, pruned_loss=0.07294, over 2760449.99 frames. ], batch size: 146, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:25:01,844 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51123.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:19,474 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51138.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:25:37,107 INFO [train.py:904] (0/8) Epoch 6, batch 400, loss[loss=0.171, simple_loss=0.2611, pruned_loss=0.04039, over 17213.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.2925, pruned_loss=0.07139, over 2887311.77 frames. ], batch size: 46, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:25:55,424 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51164.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:26:01,710 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.109e+02 2.940e+02 3.610e+02 4.239e+02 7.005e+02, threshold=7.220e+02, percent-clipped=1.0 2023-04-28 10:26:02,467 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.08 vs. limit=2.0 2023-04-28 10:26:05,633 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51172.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:26:20,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9735, 4.3864, 3.2504, 2.4412, 3.0600, 2.4251, 4.5627, 4.3183], device='cuda:0'), covar=tensor([0.2144, 0.0524, 0.1246, 0.1747, 0.2283, 0.1542, 0.0326, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0250, 0.0271, 0.0246, 0.0271, 0.0203, 0.0244, 0.0256], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:26:45,080 INFO [train.py:904] (0/8) Epoch 6, batch 450, loss[loss=0.2108, simple_loss=0.2777, pruned_loss=0.07189, over 16769.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2901, pruned_loss=0.07023, over 2987989.56 frames. ], batch size: 124, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:26:46,484 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:28,566 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-04-28 10:27:29,515 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51233.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:27:52,965 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:27:53,837 INFO [train.py:904] (0/8) Epoch 6, batch 500, loss[loss=0.2166, simple_loss=0.2811, pruned_loss=0.07603, over 12351.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2888, pruned_loss=0.06966, over 3055126.28 frames. ], batch size: 248, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:28:16,024 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-28 10:28:17,360 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 3.090e+02 3.738e+02 4.449e+02 8.454e+02, threshold=7.475e+02, percent-clipped=2.0 2023-04-28 10:29:01,777 INFO [train.py:904] (0/8) Epoch 6, batch 550, loss[loss=0.2349, simple_loss=0.2907, pruned_loss=0.08952, over 16358.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2882, pruned_loss=0.06966, over 3111901.21 frames. ], batch size: 146, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:29:03,268 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:30:13,391 INFO [train.py:904] (0/8) Epoch 6, batch 600, loss[loss=0.1991, simple_loss=0.2702, pruned_loss=0.06397, over 16840.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.287, pruned_loss=0.06935, over 3144272.12 frames. ], batch size: 102, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:30:38,610 INFO [optim.py:368] (0/8) 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 10:30:55,691 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51382.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:10,959 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:31:21,407 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9047, 4.2454, 4.4232, 4.5252, 4.5167, 4.1323, 3.6644, 4.0934], device='cuda:0'), covar=tensor([0.0577, 0.0630, 0.0612, 0.0594, 0.0605, 0.0498, 0.1650, 0.0592], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0255, 0.0254, 0.0249, 0.0302, 0.0268, 0.0377, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 10:31:23,787 INFO [train.py:904] (0/8) Epoch 6, batch 650, loss[loss=0.2085, simple_loss=0.2774, pruned_loss=0.06977, over 16615.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2856, pruned_loss=0.0691, over 3173840.01 frames. ], batch size: 89, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:31:47,267 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:03,271 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7187, 4.4597, 4.7178, 4.9365, 5.0675, 4.5655, 4.9506, 4.9969], device='cuda:0'), covar=tensor([0.0980, 0.0876, 0.1251, 0.0542, 0.0452, 0.0713, 0.0703, 0.0603], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0526, 0.0671, 0.0533, 0.0397, 0.0391, 0.0421, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:32:14,563 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51438.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:21,295 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51443.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:30,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8147, 1.3160, 1.6454, 1.7374, 1.8953, 1.9730, 1.3328, 1.8293], device='cuda:0'), covar=tensor([0.0100, 0.0173, 0.0093, 0.0125, 0.0084, 0.0073, 0.0179, 0.0046], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0145, 0.0131, 0.0128, 0.0131, 0.0095, 0.0144, 0.0080], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 10:32:31,579 INFO [train.py:904] (0/8) Epoch 6, batch 700, loss[loss=0.3194, simple_loss=0.3587, pruned_loss=0.14, over 12348.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2849, pruned_loss=0.06809, over 3206765.59 frames. ], batch size: 247, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:32:35,179 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51453.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:32:49,264 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51464.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:32:57,149 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.941e+02 3.067e+02 3.781e+02 4.816e+02 1.338e+03, threshold=7.562e+02, percent-clipped=6.0 2023-04-28 10:33:19,515 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51486.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:32,349 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:33:39,732 INFO [train.py:904] (0/8) Epoch 6, batch 750, loss[loss=0.1808, simple_loss=0.2668, pruned_loss=0.04737, over 17109.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2857, pruned_loss=0.06826, over 3228880.18 frames. ], batch size: 48, lr: 1.18e-02, grad_scale: 2.0 2023-04-28 10:33:56,283 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51512.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:34:19,299 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51528.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:34:44,169 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7074, 4.0932, 4.3806, 1.9242, 4.5865, 4.5461, 3.3117, 3.4022], device='cuda:0'), covar=tensor([0.0701, 0.0117, 0.0135, 0.1054, 0.0061, 0.0070, 0.0299, 0.0350], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0091, 0.0081, 0.0139, 0.0071, 0.0080, 0.0115, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 10:34:53,044 INFO [train.py:904] (0/8) Epoch 6, batch 800, loss[loss=0.2075, simple_loss=0.269, pruned_loss=0.07296, over 16841.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2847, pruned_loss=0.06771, over 3251974.12 frames. ], batch size: 102, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:34:55,731 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5147, 3.9367, 4.0775, 1.8060, 4.0795, 4.1655, 3.1902, 3.1742], device='cuda:0'), covar=tensor([0.0746, 0.0103, 0.0102, 0.1081, 0.0090, 0.0087, 0.0310, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0092, 0.0081, 0.0139, 0.0071, 0.0080, 0.0115, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 10:34:59,955 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:35:19,941 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.998e+02 2.921e+02 3.386e+02 4.262e+02 8.289e+02, threshold=6.772e+02, percent-clipped=2.0 2023-04-28 10:35:26,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4507, 4.1726, 4.0130, 1.9053, 3.0163, 2.1449, 3.6867, 3.8507], device='cuda:0'), covar=tensor([0.0261, 0.0541, 0.0433, 0.1640, 0.0690, 0.0986, 0.0670, 0.0910], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0129, 0.0153, 0.0141, 0.0130, 0.0125, 0.0137, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 10:36:01,912 INFO [train.py:904] (0/8) Epoch 6, batch 850, loss[loss=0.2016, simple_loss=0.2784, pruned_loss=0.06242, over 16804.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2841, pruned_loss=0.0673, over 3267125.19 frames. ], batch size: 102, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:36:04,231 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51602.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:36:54,511 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 10:37:10,634 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:37:11,535 INFO [train.py:904] (0/8) Epoch 6, batch 900, loss[loss=0.1741, simple_loss=0.2677, pruned_loss=0.04023, over 17145.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2828, pruned_loss=0.06576, over 3274017.55 frames. ], batch size: 48, lr: 1.18e-02, grad_scale: 4.0 2023-04-28 10:37:39,520 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 2.905e+02 3.526e+02 4.406e+02 8.198e+02, threshold=7.052e+02, percent-clipped=7.0 2023-04-28 10:37:55,518 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1155, 4.7980, 5.1132, 5.3192, 5.5427, 4.7677, 5.4691, 5.4423], device='cuda:0'), covar=tensor([0.0931, 0.0816, 0.1265, 0.0538, 0.0385, 0.0605, 0.0385, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0533, 0.0685, 0.0545, 0.0407, 0.0399, 0.0431, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:38:22,604 INFO [train.py:904] (0/8) Epoch 6, batch 950, loss[loss=0.2014, simple_loss=0.2729, pruned_loss=0.06497, over 15522.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2829, pruned_loss=0.06539, over 3285354.18 frames. ], batch size: 190, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:38:45,634 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:10,639 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 10:39:12,974 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:25,718 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=51747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:26,805 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:30,787 INFO [train.py:904] (0/8) Epoch 6, batch 1000, loss[loss=0.211, simple_loss=0.2684, pruned_loss=0.07682, over 16812.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2813, pruned_loss=0.06479, over 3293272.68 frames. ], batch size: 124, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:39:51,038 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:39:51,457 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-28 10:39:57,332 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.827e+02 3.297e+02 4.150e+02 8.828e+02, threshold=6.593e+02, percent-clipped=4.0 2023-04-28 10:40:12,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1594, 3.0044, 3.4482, 2.2455, 3.1686, 3.4043, 3.3497, 1.9698], device='cuda:0'), covar=tensor([0.0343, 0.0112, 0.0030, 0.0244, 0.0054, 0.0050, 0.0041, 0.0285], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0064, 0.0061, 0.0117, 0.0064, 0.0071, 0.0064, 0.0110], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:40:27,221 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9846, 2.6353, 2.6902, 1.9606, 2.5643, 2.5769, 2.6552, 1.8061], device='cuda:0'), covar=tensor([0.0288, 0.0081, 0.0042, 0.0217, 0.0059, 0.0068, 0.0050, 0.0283], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0064, 0.0061, 0.0118, 0.0064, 0.0072, 0.0065, 0.0110], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:40:39,946 INFO [train.py:904] (0/8) Epoch 6, batch 1050, loss[loss=0.2165, simple_loss=0.2774, pruned_loss=0.07786, over 16850.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2807, pruned_loss=0.06439, over 3297003.79 frames. ], batch size: 102, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:40:49,369 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=51808.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:41:17,990 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=51828.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:41:27,757 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6690, 3.4322, 3.0015, 5.0851, 4.5034, 4.6561, 1.7742, 3.6185], device='cuda:0'), covar=tensor([0.1432, 0.0535, 0.0978, 0.0092, 0.0269, 0.0328, 0.1270, 0.0570], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0143, 0.0168, 0.0093, 0.0182, 0.0185, 0.0161, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 10:41:32,947 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-28 10:41:49,466 INFO [train.py:904] (0/8) Epoch 6, batch 1100, loss[loss=0.1796, simple_loss=0.2603, pruned_loss=0.04944, over 16804.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2803, pruned_loss=0.06363, over 3300422.03 frames. ], batch size: 39, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:41:49,785 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=51851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:42:16,640 INFO [optim.py:368] (0/8) 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 10:42:24,558 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=51876.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 10:42:59,427 INFO [train.py:904] (0/8) Epoch 6, batch 1150, loss[loss=0.1923, simple_loss=0.2847, pruned_loss=0.0499, over 17288.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.28, pruned_loss=0.06321, over 3303487.01 frames. ], batch size: 52, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:43:40,647 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7519, 4.2643, 2.1726, 4.5858, 2.8658, 4.5412, 2.0701, 3.0109], device='cuda:0'), covar=tensor([0.0176, 0.0248, 0.1420, 0.0044, 0.0751, 0.0277, 0.1538, 0.0662], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0162, 0.0180, 0.0086, 0.0162, 0.0192, 0.0187, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 10:43:42,469 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0907, 3.0398, 3.2157, 2.1378, 2.9642, 3.1285, 3.0983, 2.0332], device='cuda:0'), covar=tensor([0.0292, 0.0071, 0.0028, 0.0228, 0.0048, 0.0047, 0.0048, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0062, 0.0060, 0.0114, 0.0062, 0.0069, 0.0064, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:44:08,005 INFO [train.py:904] (0/8) Epoch 6, batch 1200, loss[loss=0.1987, simple_loss=0.2932, pruned_loss=0.05217, over 17114.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2796, pruned_loss=0.06282, over 3298773.60 frames. ], batch size: 49, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:44:11,319 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4562, 4.3919, 4.3221, 4.1772, 4.0215, 4.3486, 4.1696, 4.0746], device='cuda:0'), covar=tensor([0.0416, 0.0341, 0.0201, 0.0205, 0.0658, 0.0349, 0.0471, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0218, 0.0237, 0.0207, 0.0271, 0.0244, 0.0170, 0.0273], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:44:33,631 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.991e+02 2.935e+02 3.395e+02 4.046e+02 1.078e+03, threshold=6.791e+02, percent-clipped=4.0 2023-04-28 10:44:48,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2023-04-28 10:44:54,772 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 10:45:16,460 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-52000.pt 2023-04-28 10:45:20,384 INFO [train.py:904] (0/8) Epoch 6, batch 1250, loss[loss=0.2359, simple_loss=0.3058, pruned_loss=0.08302, over 11752.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2796, pruned_loss=0.06347, over 3305861.38 frames. ], batch size: 246, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:45:28,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0376, 4.0762, 4.3948, 3.2660, 3.9306, 4.3060, 4.0503, 2.7651], device='cuda:0'), covar=tensor([0.0264, 0.0033, 0.0018, 0.0187, 0.0041, 0.0039, 0.0034, 0.0240], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0062, 0.0060, 0.0115, 0.0062, 0.0070, 0.0064, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:46:13,374 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52038.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:15,805 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:20,953 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 10:46:26,803 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52048.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:46:30,469 INFO [train.py:904] (0/8) Epoch 6, batch 1300, loss[loss=0.2352, simple_loss=0.3019, pruned_loss=0.08427, over 16899.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2789, pruned_loss=0.06274, over 3323070.06 frames. ], batch size: 116, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:46:57,056 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7003, 1.5126, 2.0703, 2.5914, 2.6286, 2.3943, 1.6426, 2.5912], device='cuda:0'), covar=tensor([0.0064, 0.0243, 0.0148, 0.0106, 0.0085, 0.0143, 0.0213, 0.0051], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0145, 0.0130, 0.0129, 0.0133, 0.0096, 0.0143, 0.0081], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 10:46:58,327 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.918e+02 3.490e+02 4.202e+02 7.834e+02, threshold=6.979e+02, percent-clipped=4.0 2023-04-28 10:47:19,564 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52086.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:33,415 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52096.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:41,506 INFO [train.py:904] (0/8) Epoch 6, batch 1350, loss[loss=0.218, simple_loss=0.2876, pruned_loss=0.0742, over 16899.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2789, pruned_loss=0.06238, over 3317612.68 frames. ], batch size: 116, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:47:42,020 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:47:44,075 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:48:51,212 INFO [train.py:904] (0/8) Epoch 6, batch 1400, loss[loss=0.2036, simple_loss=0.2979, pruned_loss=0.05466, over 17111.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2785, pruned_loss=0.06256, over 3322878.87 frames. ], batch size: 48, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:48:51,471 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:48:57,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0027, 4.7475, 4.9917, 5.2835, 5.4693, 4.6500, 5.4362, 5.3802], device='cuda:0'), covar=tensor([0.1118, 0.0941, 0.1550, 0.0561, 0.0378, 0.0617, 0.0335, 0.0469], device='cuda:0'), in_proj_covar=tensor([0.0442, 0.0544, 0.0696, 0.0557, 0.0417, 0.0410, 0.0434, 0.0469], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:49:19,189 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.784e+02 2.727e+02 3.349e+02 4.009e+02 8.330e+02, threshold=6.698e+02, percent-clipped=2.0 2023-04-28 10:49:49,823 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52193.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:49:58,143 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:50:00,575 INFO [train.py:904] (0/8) Epoch 6, batch 1450, loss[loss=0.1625, simple_loss=0.2345, pruned_loss=0.04519, over 16998.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2773, pruned_loss=0.06216, over 3331300.44 frames. ], batch size: 41, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:50:33,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0409, 4.4047, 2.4765, 4.8047, 2.9583, 4.6878, 2.3949, 3.4038], device='cuda:0'), covar=tensor([0.0153, 0.0218, 0.1306, 0.0058, 0.0718, 0.0332, 0.1306, 0.0511], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0161, 0.0180, 0.0086, 0.0160, 0.0193, 0.0186, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 10:51:10,837 INFO [train.py:904] (0/8) Epoch 6, batch 1500, loss[loss=0.2077, simple_loss=0.2717, pruned_loss=0.0718, over 16771.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2777, pruned_loss=0.06317, over 3326342.22 frames. ], batch size: 124, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:51:15,469 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52254.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:51:30,603 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 10:51:38,537 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.881e+02 3.021e+02 3.666e+02 4.541e+02 9.509e+02, threshold=7.332e+02, percent-clipped=3.0 2023-04-28 10:51:52,579 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-28 10:52:18,648 INFO [train.py:904] (0/8) Epoch 6, batch 1550, loss[loss=0.2314, simple_loss=0.3067, pruned_loss=0.07809, over 16860.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.281, pruned_loss=0.06589, over 3329870.99 frames. ], batch size: 116, lr: 1.17e-02, grad_scale: 4.0 2023-04-28 10:52:30,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7202, 4.8166, 4.9539, 4.7803, 4.7195, 5.4821, 5.0273, 4.7211], device='cuda:0'), covar=tensor([0.1212, 0.1679, 0.1729, 0.1838, 0.2917, 0.0990, 0.1344, 0.2401], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0423, 0.0420, 0.0364, 0.0489, 0.0453, 0.0343, 0.0486], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:52:33,891 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 6, batch 1600, loss[loss=0.1939, simple_loss=0.2846, pruned_loss=0.0516, over 17246.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2822, pruned_loss=0.06595, over 3328931.91 frames. ], batch size: 52, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:53:55,826 INFO [optim.py:368] (0/8) 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 10:53:59,257 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52373.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:30,447 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:54:37,125 INFO [train.py:904] (0/8) Epoch 6, batch 1650, loss[loss=0.2038, simple_loss=0.2743, pruned_loss=0.06662, over 16944.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2833, pruned_loss=0.06594, over 3320827.06 frames. ], batch size: 96, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:54:40,247 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52403.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:55:25,771 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6325, 3.2510, 3.9073, 2.7654, 3.4447, 3.7316, 3.5878, 2.1926], device='cuda:0'), covar=tensor([0.0246, 0.0080, 0.0021, 0.0196, 0.0048, 0.0043, 0.0038, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0061, 0.0059, 0.0112, 0.0061, 0.0070, 0.0064, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:55:45,572 INFO [train.py:904] (0/8) Epoch 6, batch 1700, loss[loss=0.2132, simple_loss=0.3019, pruned_loss=0.06227, over 17036.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2857, pruned_loss=0.06683, over 3319597.83 frames. ], batch size: 53, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:55:45,914 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52451.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:56:14,196 INFO [optim.py:368] (0/8) 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 10:56:14,909 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 10:56:40,383 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2752, 2.1812, 1.7015, 1.9821, 2.5968, 2.4339, 2.6403, 2.7211], device='cuda:0'), covar=tensor([0.0082, 0.0174, 0.0226, 0.0212, 0.0095, 0.0169, 0.0123, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0095, 0.0167, 0.0164, 0.0162, 0.0162, 0.0168, 0.0153, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:56:57,693 INFO [train.py:904] (0/8) Epoch 6, batch 1750, loss[loss=0.2289, simple_loss=0.2988, pruned_loss=0.07945, over 16701.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2879, pruned_loss=0.06774, over 3306043.75 frames. ], batch size: 134, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:05,486 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52549.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 10:58:07,623 INFO [train.py:904] (0/8) Epoch 6, batch 1800, loss[loss=0.2036, simple_loss=0.2956, pruned_loss=0.05583, over 17034.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2884, pruned_loss=0.06733, over 3316890.95 frames. ], batch size: 50, lr: 1.17e-02, grad_scale: 8.0 2023-04-28 10:58:14,083 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52555.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 10:58:27,258 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8606, 4.6019, 4.8308, 5.0963, 5.2617, 4.6067, 5.2121, 5.1692], device='cuda:0'), covar=tensor([0.1149, 0.0890, 0.1446, 0.0551, 0.0392, 0.0677, 0.0441, 0.0437], device='cuda:0'), in_proj_covar=tensor([0.0443, 0.0548, 0.0695, 0.0560, 0.0416, 0.0406, 0.0435, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:58:36,301 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 3.092e+02 3.773e+02 4.797e+02 9.614e+02, threshold=7.547e+02, percent-clipped=5.0 2023-04-28 10:58:55,247 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8368, 2.1684, 2.3366, 4.5804, 1.9543, 3.2339, 2.3382, 2.4487], device='cuda:0'), covar=tensor([0.0615, 0.2494, 0.1281, 0.0269, 0.3188, 0.1239, 0.2067, 0.2692], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0331, 0.0274, 0.0318, 0.0374, 0.0348, 0.0301, 0.0403], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:58:57,482 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0652, 5.2676, 5.0185, 4.9546, 4.0415, 5.1814, 5.1953, 4.6715], device='cuda:0'), covar=tensor([0.0643, 0.0372, 0.0312, 0.0232, 0.1683, 0.0320, 0.0248, 0.0526], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0221, 0.0243, 0.0215, 0.0277, 0.0248, 0.0173, 0.0276], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 10:59:17,743 INFO [train.py:904] (0/8) Epoch 6, batch 1850, loss[loss=0.2169, simple_loss=0.3006, pruned_loss=0.06663, over 17111.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2893, pruned_loss=0.06754, over 3319850.80 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 10:59:38,390 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52616.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 10:59:39,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8976, 3.6908, 3.7860, 4.0876, 4.1401, 3.7888, 3.9090, 4.1484], device='cuda:0'), covar=tensor([0.1028, 0.0910, 0.1564, 0.0701, 0.0657, 0.1322, 0.1307, 0.0561], device='cuda:0'), in_proj_covar=tensor([0.0441, 0.0545, 0.0692, 0.0559, 0.0415, 0.0408, 0.0434, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 10:59:54,413 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0017, 3.5151, 3.0091, 1.8888, 2.6506, 2.1943, 3.3771, 3.3323], device='cuda:0'), covar=tensor([0.0217, 0.0492, 0.0557, 0.1431, 0.0677, 0.0930, 0.0545, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0134, 0.0151, 0.0139, 0.0130, 0.0124, 0.0139, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:00:24,468 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2549, 5.2849, 5.1605, 4.5429, 5.1005, 2.0470, 4.9011, 5.1782], device='cuda:0'), covar=tensor([0.0050, 0.0041, 0.0078, 0.0272, 0.0048, 0.1546, 0.0075, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0090, 0.0139, 0.0137, 0.0104, 0.0150, 0.0120, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:00:27,090 INFO [train.py:904] (0/8) Epoch 6, batch 1900, loss[loss=0.2227, simple_loss=0.2973, pruned_loss=0.07403, over 16880.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2886, pruned_loss=0.06696, over 3315061.33 frames. ], batch size: 83, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:00:51,247 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:00:52,606 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2968, 4.4293, 4.8141, 4.8616, 4.8666, 4.4496, 4.4481, 4.2474], device='cuda:0'), covar=tensor([0.0302, 0.0406, 0.0320, 0.0363, 0.0345, 0.0274, 0.0777, 0.0411], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0267, 0.0265, 0.0263, 0.0322, 0.0280, 0.0395, 0.0233], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 11:00:54,522 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.841e+02 2.667e+02 3.390e+02 4.181e+02 1.051e+03, threshold=6.780e+02, percent-clipped=5.0 2023-04-28 11:01:10,425 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6142, 4.7396, 5.1376, 5.2380, 5.1873, 4.7313, 4.7319, 4.4355], device='cuda:0'), covar=tensor([0.0262, 0.0347, 0.0319, 0.0327, 0.0358, 0.0249, 0.0735, 0.0381], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0267, 0.0266, 0.0263, 0.0322, 0.0280, 0.0396, 0.0233], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 11:01:26,552 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5467, 4.3149, 4.5175, 4.7451, 4.8820, 4.3401, 4.6255, 4.8344], device='cuda:0'), covar=tensor([0.1042, 0.0815, 0.1305, 0.0581, 0.0444, 0.0793, 0.1073, 0.0487], device='cuda:0'), in_proj_covar=tensor([0.0446, 0.0549, 0.0699, 0.0564, 0.0420, 0.0415, 0.0439, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:01:30,570 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:01:36,641 INFO [train.py:904] (0/8) Epoch 6, batch 1950, loss[loss=0.2002, simple_loss=0.2941, pruned_loss=0.05312, over 17067.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2876, pruned_loss=0.06578, over 3313975.03 frames. ], batch size: 55, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:01:57,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0736, 4.3765, 2.3080, 4.7685, 2.9272, 4.7429, 2.4170, 3.2195], device='cuda:0'), covar=tensor([0.0131, 0.0211, 0.1299, 0.0037, 0.0678, 0.0255, 0.1222, 0.0544], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0161, 0.0176, 0.0085, 0.0160, 0.0193, 0.0185, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:02:36,312 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 6, batch 2000, loss[loss=0.214, simple_loss=0.2961, pruned_loss=0.06594, over 17245.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2866, pruned_loss=0.06491, over 3321796.06 frames. ], batch size: 52, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:02:49,263 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=52752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:03:15,695 INFO [optim.py:368] (0/8) 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 11:03:48,768 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3849, 4.3231, 4.7666, 4.8464, 4.8585, 4.4908, 4.4251, 4.2046], device='cuda:0'), covar=tensor([0.0285, 0.0470, 0.0387, 0.0417, 0.0372, 0.0298, 0.0812, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0266, 0.0263, 0.0262, 0.0319, 0.0279, 0.0393, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 11:03:57,179 INFO [train.py:904] (0/8) Epoch 6, batch 2050, loss[loss=0.2568, simple_loss=0.3153, pruned_loss=0.09912, over 16294.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2865, pruned_loss=0.06503, over 3319163.53 frames. ], batch size: 165, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:03:57,773 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8258, 4.6831, 4.2054, 2.0164, 3.4090, 2.6478, 4.1136, 4.2108], device='cuda:0'), covar=tensor([0.0222, 0.0397, 0.0393, 0.1572, 0.0602, 0.0851, 0.0564, 0.0792], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0135, 0.0153, 0.0140, 0.0130, 0.0125, 0.0139, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:04:14,406 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=52813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:04:50,998 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0634, 5.0225, 5.5134, 5.5408, 5.5736, 5.1893, 5.1098, 4.7929], device='cuda:0'), covar=tensor([0.0227, 0.0310, 0.0310, 0.0453, 0.0313, 0.0212, 0.0677, 0.0273], device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0264, 0.0262, 0.0261, 0.0314, 0.0277, 0.0390, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 11:05:05,618 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 6, batch 2100, loss[loss=0.2306, simple_loss=0.3056, pruned_loss=0.0778, over 11789.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2874, pruned_loss=0.06634, over 3309170.69 frames. ], batch size: 248, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:05:16,652 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6151, 4.6521, 5.1573, 5.1678, 5.1766, 4.7255, 4.7395, 4.3244], device='cuda:0'), covar=tensor([0.0278, 0.0397, 0.0293, 0.0379, 0.0373, 0.0262, 0.0774, 0.0388], device='cuda:0'), in_proj_covar=tensor([0.0261, 0.0262, 0.0259, 0.0258, 0.0311, 0.0276, 0.0387, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 11:05:36,282 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.933e+02 3.490e+02 4.297e+02 7.985e+02, threshold=6.979e+02, percent-clipped=3.0 2023-04-28 11:06:12,922 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=52897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:06:18,019 INFO [train.py:904] (0/8) Epoch 6, batch 2150, loss[loss=0.2291, simple_loss=0.313, pruned_loss=0.07263, over 16725.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2894, pruned_loss=0.06716, over 3310070.09 frames. ], batch size: 57, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:06:33,075 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=52911.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 11:06:50,898 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2347, 2.0991, 2.6285, 3.2992, 2.9342, 3.5724, 2.2364, 3.4068], device='cuda:0'), covar=tensor([0.0082, 0.0224, 0.0132, 0.0104, 0.0119, 0.0075, 0.0199, 0.0081], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0148, 0.0133, 0.0135, 0.0139, 0.0100, 0.0143, 0.0085], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 11:07:14,997 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6935, 4.4078, 4.1317, 2.0695, 3.2707, 2.5005, 4.0601, 3.9350], device='cuda:0'), covar=tensor([0.0230, 0.0393, 0.0366, 0.1432, 0.0583, 0.0884, 0.0494, 0.0881], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0135, 0.0152, 0.0140, 0.0130, 0.0125, 0.0139, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:07:29,315 INFO [train.py:904] (0/8) Epoch 6, batch 2200, loss[loss=0.28, simple_loss=0.3339, pruned_loss=0.113, over 11701.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2899, pruned_loss=0.06776, over 3300927.50 frames. ], batch size: 247, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:07:52,259 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=52968.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:07:56,705 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.110e+02 3.315e+02 3.926e+02 4.640e+02 8.894e+02, threshold=7.853e+02, percent-clipped=3.0 2023-04-28 11:08:38,592 INFO [train.py:904] (0/8) Epoch 6, batch 2250, loss[loss=0.2013, simple_loss=0.275, pruned_loss=0.0638, over 16878.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2897, pruned_loss=0.06718, over 3310790.49 frames. ], batch size: 96, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:08:59,898 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53016.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:09:13,458 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.78 vs. limit=5.0 2023-04-28 11:09:21,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1868, 1.9555, 1.4927, 1.6887, 2.2608, 2.0491, 2.2543, 2.3887], device='cuda:0'), covar=tensor([0.0058, 0.0175, 0.0233, 0.0224, 0.0102, 0.0172, 0.0108, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0094, 0.0165, 0.0164, 0.0161, 0.0163, 0.0167, 0.0153, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:09:46,429 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8246, 4.1238, 4.5364, 3.2432, 3.9156, 4.3470, 3.8645, 2.6747], device='cuda:0'), covar=tensor([0.0273, 0.0025, 0.0017, 0.0185, 0.0042, 0.0032, 0.0041, 0.0259], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0063, 0.0061, 0.0115, 0.0063, 0.0072, 0.0066, 0.0109], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 11:09:47,165 INFO [train.py:904] (0/8) Epoch 6, batch 2300, loss[loss=0.187, simple_loss=0.2763, pruned_loss=0.04887, over 17108.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2903, pruned_loss=0.06695, over 3305980.71 frames. ], batch size: 49, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:09:54,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53056.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:10:15,631 INFO [optim.py:368] (0/8) 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 11:10:22,125 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0326, 5.4423, 5.5440, 5.4049, 5.3706, 5.9701, 5.6318, 5.3809], device='cuda:0'), covar=tensor([0.0667, 0.1629, 0.1359, 0.1743, 0.2509, 0.0788, 0.1010, 0.2000], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0430, 0.0421, 0.0367, 0.0487, 0.0452, 0.0347, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 11:10:52,186 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1901, 2.2242, 2.3618, 4.8001, 1.9905, 3.2639, 2.5677, 2.5108], device='cuda:0'), covar=tensor([0.0504, 0.2387, 0.1247, 0.0229, 0.3073, 0.1216, 0.1934, 0.2673], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0334, 0.0275, 0.0317, 0.0375, 0.0351, 0.0305, 0.0406], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:10:57,597 INFO [train.py:904] (0/8) Epoch 6, batch 2350, loss[loss=0.1905, simple_loss=0.2753, pruned_loss=0.05285, over 15863.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2899, pruned_loss=0.06654, over 3319363.75 frames. ], batch size: 35, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:10:58,287 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 11:11:08,093 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:11:20,139 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53117.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:12:08,723 INFO [train.py:904] (0/8) Epoch 6, batch 2400, loss[loss=0.2208, simple_loss=0.2935, pruned_loss=0.07411, over 16853.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2899, pruned_loss=0.06679, over 3326122.55 frames. ], batch size: 96, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:12:21,053 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7592, 3.0929, 2.8116, 4.9108, 4.2416, 4.4354, 1.5518, 3.0667], device='cuda:0'), covar=tensor([0.1365, 0.0571, 0.1047, 0.0083, 0.0286, 0.0331, 0.1401, 0.0757], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0144, 0.0167, 0.0096, 0.0191, 0.0187, 0.0161, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 11:12:36,946 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.873e+02 2.840e+02 3.422e+02 4.151e+02 8.672e+02, threshold=6.844e+02, percent-clipped=2.0 2023-04-28 11:13:19,228 INFO [train.py:904] (0/8) Epoch 6, batch 2450, loss[loss=0.2185, simple_loss=0.2825, pruned_loss=0.07725, over 16912.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2908, pruned_loss=0.06632, over 3330887.44 frames. ], batch size: 109, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:13:33,009 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53211.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:14:09,636 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 11:14:15,962 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:14:29,168 INFO [train.py:904] (0/8) Epoch 6, batch 2500, loss[loss=0.1823, simple_loss=0.2783, pruned_loss=0.04318, over 17056.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2899, pruned_loss=0.06617, over 3331879.67 frames. ], batch size: 53, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:14:39,972 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53259.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:14:57,078 INFO [optim.py:368] (0/8) 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 11:15:32,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 11:15:38,467 INFO [train.py:904] (0/8) Epoch 6, batch 2550, loss[loss=0.1925, simple_loss=0.2813, pruned_loss=0.05181, over 17110.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2909, pruned_loss=0.06655, over 3327384.31 frames. ], batch size: 48, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:15:40,102 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53302.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:43,210 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:15:44,735 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 11:15:58,250 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53315.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:16:48,727 INFO [train.py:904] (0/8) Epoch 6, batch 2600, loss[loss=0.2272, simple_loss=0.2936, pruned_loss=0.08045, over 16868.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2908, pruned_loss=0.0664, over 3315161.87 frames. ], batch size: 116, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:17:09,364 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53365.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:11,640 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=53367.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:14,178 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.49 vs. limit=5.0 2023-04-28 11:17:16,418 INFO [optim.py:368] (0/8) 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 11:17:24,031 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53376.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:17:59,457 INFO [train.py:904] (0/8) Epoch 6, batch 2650, loss[loss=0.202, simple_loss=0.2781, pruned_loss=0.06297, over 16191.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2919, pruned_loss=0.06642, over 3309646.93 frames. ], batch size: 164, lr: 1.16e-02, grad_scale: 8.0 2023-04-28 11:18:10,304 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:18:14,692 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=53428.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:18:48,273 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5249, 4.3394, 3.7372, 2.0118, 3.1343, 2.4675, 3.7775, 3.8988], device='cuda:0'), covar=tensor([0.0246, 0.0433, 0.0509, 0.1607, 0.0664, 0.0893, 0.0674, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0138, 0.0156, 0.0141, 0.0131, 0.0125, 0.0141, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:19:09,543 INFO [train.py:904] (0/8) Epoch 6, batch 2700, loss[loss=0.1947, simple_loss=0.2891, pruned_loss=0.05011, over 17257.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2919, pruned_loss=0.06582, over 3317624.05 frames. ], batch size: 52, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:19:12,886 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2167, 4.8618, 5.1762, 5.3944, 5.5573, 4.8631, 5.5010, 5.5029], device='cuda:0'), covar=tensor([0.0973, 0.0764, 0.1162, 0.0460, 0.0388, 0.0499, 0.0416, 0.0388], device='cuda:0'), in_proj_covar=tensor([0.0445, 0.0537, 0.0694, 0.0557, 0.0421, 0.0416, 0.0435, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:19:16,991 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:19:38,363 INFO [optim.py:368] (0/8) 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 11:20:13,235 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5166, 3.6096, 3.9899, 2.7804, 3.5819, 3.9317, 3.6609, 2.3284], device='cuda:0'), covar=tensor([0.0286, 0.0097, 0.0025, 0.0218, 0.0042, 0.0043, 0.0037, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0060, 0.0061, 0.0111, 0.0062, 0.0069, 0.0063, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 11:20:19,916 INFO [train.py:904] (0/8) Epoch 6, batch 2750, loss[loss=0.2193, simple_loss=0.2909, pruned_loss=0.07385, over 16758.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2915, pruned_loss=0.06492, over 3329673.14 frames. ], batch size: 124, lr: 1.16e-02, grad_scale: 4.0 2023-04-28 11:21:31,904 INFO [train.py:904] (0/8) Epoch 6, batch 2800, loss[loss=0.1628, simple_loss=0.2432, pruned_loss=0.04118, over 17000.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2915, pruned_loss=0.06489, over 3327341.18 frames. ], batch size: 41, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:21:53,323 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5635, 2.0242, 2.1947, 4.1037, 1.8469, 2.7271, 2.0737, 2.1800], device='cuda:0'), covar=tensor([0.0731, 0.2776, 0.1349, 0.0402, 0.3417, 0.1529, 0.2553, 0.2709], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0336, 0.0275, 0.0316, 0.0375, 0.0353, 0.0304, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:22:01,719 INFO [optim.py:368] (0/8) 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 11:22:04,705 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4224, 4.2107, 3.6936, 2.2075, 2.9526, 2.5163, 3.7208, 3.9814], device='cuda:0'), covar=tensor([0.0343, 0.0572, 0.0490, 0.1430, 0.0797, 0.0900, 0.0849, 0.0972], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0137, 0.0156, 0.0140, 0.0131, 0.0125, 0.0141, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:22:30,945 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5682, 4.3443, 4.5285, 4.7623, 4.9064, 4.4283, 4.7929, 4.8155], device='cuda:0'), covar=tensor([0.0958, 0.0768, 0.1192, 0.0485, 0.0436, 0.0726, 0.0734, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0445, 0.0535, 0.0692, 0.0554, 0.0417, 0.0414, 0.0431, 0.0465], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:22:37,076 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53597.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:22:43,368 INFO [train.py:904] (0/8) Epoch 6, batch 2850, loss[loss=0.2195, simple_loss=0.3002, pruned_loss=0.06944, over 16692.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2899, pruned_loss=0.06473, over 3334561.78 frames. ], batch size: 76, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:23:39,068 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-28 11:23:51,293 INFO [train.py:904] (0/8) Epoch 6, batch 2900, loss[loss=0.2073, simple_loss=0.2932, pruned_loss=0.06068, over 16694.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2885, pruned_loss=0.06407, over 3335279.10 frames. ], batch size: 62, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:24:04,497 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53660.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:19,312 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53671.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:24:20,249 INFO [optim.py:368] (0/8) 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 11:24:55,862 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 11:24:59,797 INFO [train.py:904] (0/8) Epoch 6, batch 2950, loss[loss=0.196, simple_loss=0.2687, pruned_loss=0.06166, over 16780.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2884, pruned_loss=0.06545, over 3334242.32 frames. ], batch size: 102, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:25:15,499 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53712.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:25:30,130 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=53723.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:25:40,701 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-28 11:26:08,577 INFO [train.py:904] (0/8) Epoch 6, batch 3000, loss[loss=0.1876, simple_loss=0.2851, pruned_loss=0.04508, over 17113.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2884, pruned_loss=0.06592, over 3341098.48 frames. ], batch size: 48, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:26:08,578 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 11:26:17,401 INFO [train.py:938] (0/8) Epoch 6, validation: loss=0.1514, simple_loss=0.258, pruned_loss=0.02246, over 944034.00 frames. 2023-04-28 11:26:17,402 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 11:26:29,472 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53760.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:26:45,970 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 6, batch 3050, loss[loss=0.1993, simple_loss=0.2881, pruned_loss=0.05521, over 17053.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2879, pruned_loss=0.06593, over 3335633.73 frames. ], batch size: 53, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:27:32,872 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6711, 2.5322, 2.2609, 2.2674, 2.9396, 2.7279, 3.6717, 3.2145], device='cuda:0'), covar=tensor([0.0035, 0.0182, 0.0193, 0.0226, 0.0122, 0.0172, 0.0094, 0.0112], device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0166, 0.0166, 0.0163, 0.0163, 0.0169, 0.0157, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:28:35,660 INFO [train.py:904] (0/8) Epoch 6, batch 3100, loss[loss=0.2457, simple_loss=0.3044, pruned_loss=0.09345, over 16873.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2873, pruned_loss=0.06624, over 3334622.51 frames. ], batch size: 116, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:29:05,910 INFO [optim.py:368] (0/8) 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 11:29:40,297 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53897.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:29:45,147 INFO [train.py:904] (0/8) Epoch 6, batch 3150, loss[loss=0.2116, simple_loss=0.2798, pruned_loss=0.07172, over 16206.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2864, pruned_loss=0.06589, over 3324742.62 frames. ], batch size: 165, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:30:17,334 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2099, 5.6573, 5.7864, 5.6607, 5.6940, 6.1717, 5.7832, 5.5575], device='cuda:0'), covar=tensor([0.0670, 0.1422, 0.1502, 0.1444, 0.2079, 0.0752, 0.1051, 0.2013], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0430, 0.0425, 0.0364, 0.0490, 0.0449, 0.0351, 0.0495], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 11:30:49,062 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=53945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:30:53,659 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8352, 4.4284, 4.4353, 3.2135, 3.8726, 4.4343, 3.9665, 2.9865], device='cuda:0'), covar=tensor([0.0289, 0.0021, 0.0028, 0.0201, 0.0046, 0.0034, 0.0037, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0061, 0.0062, 0.0112, 0.0063, 0.0070, 0.0065, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 11:30:56,672 INFO [train.py:904] (0/8) Epoch 6, batch 3200, loss[loss=0.2255, simple_loss=0.2981, pruned_loss=0.07645, over 16458.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2858, pruned_loss=0.0658, over 3322543.12 frames. ], batch size: 75, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:31:09,512 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53960.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:31:25,799 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=53971.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:31:26,626 INFO [optim.py:368] (0/8) 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 11:32:05,727 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-54000.pt 2023-04-28 11:32:09,687 INFO [train.py:904] (0/8) Epoch 6, batch 3250, loss[loss=0.2279, simple_loss=0.2913, pruned_loss=0.08228, over 16885.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2854, pruned_loss=0.06572, over 3326707.84 frames. ], batch size: 90, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:32:19,698 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54008.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:25,634 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5968, 4.5218, 4.4931, 3.9692, 4.5528, 1.7579, 4.3206, 4.3768], device='cuda:0'), covar=tensor([0.0072, 0.0057, 0.0103, 0.0280, 0.0056, 0.1754, 0.0089, 0.0118], device='cuda:0'), in_proj_covar=tensor([0.0108, 0.0096, 0.0148, 0.0144, 0.0111, 0.0154, 0.0128, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:32:35,405 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54019.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:32:40,153 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=54023.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:33:17,041 INFO [train.py:904] (0/8) Epoch 6, batch 3300, loss[loss=0.1752, simple_loss=0.2661, pruned_loss=0.04212, over 17122.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2856, pruned_loss=0.06495, over 3331542.00 frames. ], batch size: 48, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:33:45,160 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=54071.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:33:46,071 INFO [optim.py:368] (0/8) 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 11:34:26,275 INFO [train.py:904] (0/8) Epoch 6, batch 3350, loss[loss=0.1808, simple_loss=0.2598, pruned_loss=0.05092, over 16942.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2876, pruned_loss=0.0661, over 3316672.49 frames. ], batch size: 41, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:34:28,508 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1074, 5.4645, 4.7505, 5.4398, 4.9358, 4.6155, 5.0284, 5.4849], device='cuda:0'), covar=tensor([0.1490, 0.1228, 0.2054, 0.0716, 0.1190, 0.1179, 0.1223, 0.1404], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0571, 0.0472, 0.0358, 0.0352, 0.0354, 0.0457, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:35:28,237 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-04-28 11:35:34,492 INFO [train.py:904] (0/8) Epoch 6, batch 3400, loss[loss=0.1932, simple_loss=0.2851, pruned_loss=0.0507, over 17268.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2878, pruned_loss=0.0659, over 3304905.93 frames. ], batch size: 52, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:36:01,496 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0450, 3.1841, 1.7714, 3.2141, 2.3429, 3.2443, 1.8933, 2.5015], device='cuda:0'), covar=tensor([0.0186, 0.0346, 0.1441, 0.0141, 0.0728, 0.0451, 0.1287, 0.0642], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0166, 0.0179, 0.0091, 0.0165, 0.0204, 0.0189, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:36:05,330 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 6, batch 3450, loss[loss=0.1872, simple_loss=0.2802, pruned_loss=0.04704, over 17105.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2862, pruned_loss=0.065, over 3311318.76 frames. ], batch size: 49, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:36:57,052 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-04-28 11:37:58,614 INFO [train.py:904] (0/8) Epoch 6, batch 3500, loss[loss=0.2212, simple_loss=0.2845, pruned_loss=0.07896, over 16394.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2845, pruned_loss=0.06397, over 3313034.61 frames. ], batch size: 146, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:38:28,906 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.000e+02 2.700e+02 3.164e+02 3.995e+02 8.205e+02, threshold=6.329e+02, percent-clipped=2.0 2023-04-28 11:39:10,639 INFO [train.py:904] (0/8) Epoch 6, batch 3550, loss[loss=0.2241, simple_loss=0.2865, pruned_loss=0.08088, over 16719.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2836, pruned_loss=0.06401, over 3309242.20 frames. ], batch size: 124, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:39:29,681 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 11:39:48,067 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3902, 3.2640, 3.5752, 2.5572, 3.1993, 3.5819, 3.4330, 2.1796], device='cuda:0'), covar=tensor([0.0256, 0.0092, 0.0038, 0.0190, 0.0043, 0.0048, 0.0045, 0.0253], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0060, 0.0062, 0.0112, 0.0062, 0.0070, 0.0065, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 11:40:07,294 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5213, 4.0566, 4.0005, 1.9382, 4.2758, 4.2262, 3.1315, 3.3390], device='cuda:0'), covar=tensor([0.0722, 0.0104, 0.0139, 0.1094, 0.0055, 0.0079, 0.0346, 0.0308], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0093, 0.0084, 0.0138, 0.0072, 0.0084, 0.0116, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 11:40:21,764 INFO [train.py:904] (0/8) Epoch 6, batch 3600, loss[loss=0.2085, simple_loss=0.3, pruned_loss=0.05847, over 17076.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2824, pruned_loss=0.06324, over 3301749.96 frames. ], batch size: 53, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:40:51,132 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 3.018e+02 3.592e+02 4.498e+02 7.311e+02, threshold=7.184e+02, percent-clipped=7.0 2023-04-28 11:41:33,855 INFO [train.py:904] (0/8) Epoch 6, batch 3650, loss[loss=0.2025, simple_loss=0.268, pruned_loss=0.06848, over 16907.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2809, pruned_loss=0.06317, over 3310820.52 frames. ], batch size: 116, lr: 1.15e-02, grad_scale: 8.0 2023-04-28 11:42:46,527 INFO [train.py:904] (0/8) Epoch 6, batch 3700, loss[loss=0.2242, simple_loss=0.2904, pruned_loss=0.07897, over 16705.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2802, pruned_loss=0.06541, over 3298171.44 frames. ], batch size: 124, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:43:17,523 INFO [optim.py:368] (0/8) 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 11:43:29,645 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-04-28 11:43:33,695 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4217, 1.9630, 2.2159, 3.9842, 1.9926, 2.7072, 2.0894, 2.1448], device='cuda:0'), covar=tensor([0.0641, 0.2392, 0.1229, 0.0297, 0.2622, 0.1295, 0.2382, 0.2268], device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0342, 0.0280, 0.0324, 0.0382, 0.0362, 0.0311, 0.0412], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:43:56,060 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4544, 3.8251, 3.8211, 1.7703, 3.9503, 4.0039, 3.0919, 3.1085], device='cuda:0'), covar=tensor([0.0699, 0.0098, 0.0114, 0.1080, 0.0062, 0.0074, 0.0279, 0.0308], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0091, 0.0080, 0.0134, 0.0070, 0.0081, 0.0112, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 11:43:59,571 INFO [train.py:904] (0/8) Epoch 6, batch 3750, loss[loss=0.2218, simple_loss=0.2939, pruned_loss=0.07484, over 16610.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2819, pruned_loss=0.0671, over 3290623.21 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:45:13,206 INFO [train.py:904] (0/8) Epoch 6, batch 3800, loss[loss=0.2124, simple_loss=0.2789, pruned_loss=0.07291, over 16679.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2829, pruned_loss=0.06872, over 3285954.96 frames. ], batch size: 76, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:45:24,007 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-28 11:45:46,023 INFO [optim.py:368] (0/8) 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 11:46:27,930 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.44 vs. limit=5.0 2023-04-28 11:46:28,651 INFO [train.py:904] (0/8) Epoch 6, batch 3850, loss[loss=0.205, simple_loss=0.2726, pruned_loss=0.06873, over 16743.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2825, pruned_loss=0.06896, over 3289457.55 frames. ], batch size: 89, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:40,657 INFO [train.py:904] (0/8) Epoch 6, batch 3900, loss[loss=0.2079, simple_loss=0.2841, pruned_loss=0.06583, over 16320.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2824, pruned_loss=0.06979, over 3286699.84 frames. ], batch size: 35, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:47:59,659 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54664.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 11:48:10,870 INFO [optim.py:368] (0/8) 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 11:48:19,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6998, 2.5337, 2.3738, 3.6098, 3.0374, 3.6577, 1.4252, 2.8145], device='cuda:0'), covar=tensor([0.1280, 0.0569, 0.1055, 0.0100, 0.0237, 0.0354, 0.1417, 0.0696], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0143, 0.0165, 0.0097, 0.0190, 0.0187, 0.0158, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 11:48:53,477 INFO [train.py:904] (0/8) Epoch 6, batch 3950, loss[loss=0.2163, simple_loss=0.2811, pruned_loss=0.07571, over 16731.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2824, pruned_loss=0.07044, over 3273077.22 frames. ], batch size: 89, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:49:27,734 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54725.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 11:50:04,727 INFO [train.py:904] (0/8) Epoch 6, batch 4000, loss[loss=0.2097, simple_loss=0.2796, pruned_loss=0.06994, over 16930.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2816, pruned_loss=0.07026, over 3280139.34 frames. ], batch size: 109, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 11:50:12,668 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8453, 4.2977, 4.6232, 2.3444, 4.9049, 5.0489, 3.4619, 4.0423], device='cuda:0'), covar=tensor([0.0777, 0.0147, 0.0143, 0.1101, 0.0036, 0.0028, 0.0302, 0.0246], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0095, 0.0082, 0.0138, 0.0072, 0.0083, 0.0116, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 11:50:36,655 INFO [optim.py:368] (0/8) 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 11:51:17,349 INFO [train.py:904] (0/8) Epoch 6, batch 4050, loss[loss=0.1995, simple_loss=0.2689, pruned_loss=0.065, over 16652.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2815, pruned_loss=0.06883, over 3283608.25 frames. ], batch size: 57, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:28,303 INFO [train.py:904] (0/8) Epoch 6, batch 4100, loss[loss=0.2114, simple_loss=0.2972, pruned_loss=0.06278, over 16789.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.282, pruned_loss=0.06665, over 3290120.24 frames. ], batch size: 83, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:52:45,866 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=54862.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:53:01,858 INFO [optim.py:368] (0/8) 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 11:53:05,218 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-04-28 11:53:45,449 INFO [train.py:904] (0/8) Epoch 6, batch 4150, loss[loss=0.2202, simple_loss=0.3117, pruned_loss=0.0643, over 16860.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2897, pruned_loss=0.07005, over 3258019.50 frames. ], batch size: 96, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:53:58,114 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 11:54:19,549 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=54923.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 11:55:01,109 INFO [train.py:904] (0/8) Epoch 6, batch 4200, loss[loss=0.2436, simple_loss=0.331, pruned_loss=0.07806, over 16527.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.2979, pruned_loss=0.07239, over 3234494.16 frames. ], batch size: 68, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:55:33,723 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 3.214e+02 3.756e+02 4.661e+02 1.168e+03, threshold=7.512e+02, percent-clipped=6.0 2023-04-28 11:55:37,617 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7412, 5.0858, 4.8112, 4.8662, 4.5217, 4.3238, 4.5264, 5.1449], device='cuda:0'), covar=tensor([0.0602, 0.0588, 0.0718, 0.0465, 0.0590, 0.0837, 0.0709, 0.0563], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0524, 0.0438, 0.0335, 0.0324, 0.0336, 0.0428, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:56:15,276 INFO [train.py:904] (0/8) Epoch 6, batch 4250, loss[loss=0.2128, simple_loss=0.2879, pruned_loss=0.06885, over 12563.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3007, pruned_loss=0.07224, over 3221571.35 frames. ], batch size: 246, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:56:22,981 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7113, 4.6886, 4.5135, 3.9808, 4.6055, 1.7038, 4.3029, 4.4476], device='cuda:0'), covar=tensor([0.0053, 0.0048, 0.0098, 0.0231, 0.0060, 0.1794, 0.0095, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0089, 0.0136, 0.0133, 0.0103, 0.0145, 0.0119, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:56:44,007 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55020.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 11:57:09,660 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 11:57:29,397 INFO [train.py:904] (0/8) Epoch 6, batch 4300, loss[loss=0.2221, simple_loss=0.3058, pruned_loss=0.06926, over 16565.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3015, pruned_loss=0.0711, over 3215567.63 frames. ], batch size: 75, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:57:32,054 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-28 11:57:39,477 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 11:58:01,804 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.606e+02 3.247e+02 3.921e+02 6.703e+02, threshold=6.494e+02, percent-clipped=0.0 2023-04-28 11:58:43,135 INFO [train.py:904] (0/8) Epoch 6, batch 4350, loss[loss=0.2598, simple_loss=0.3298, pruned_loss=0.09488, over 15283.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3047, pruned_loss=0.07224, over 3212475.45 frames. ], batch size: 191, lr: 1.14e-02, grad_scale: 4.0 2023-04-28 11:58:45,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7953, 3.6208, 3.0027, 1.5835, 2.5892, 2.1671, 3.3087, 3.4920], device='cuda:0'), covar=tensor([0.0283, 0.0458, 0.0560, 0.1642, 0.0776, 0.0872, 0.0592, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0133, 0.0154, 0.0139, 0.0132, 0.0124, 0.0138, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 11:59:34,664 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0591, 5.0785, 4.8275, 4.3204, 4.9943, 1.8903, 4.7147, 4.8449], device='cuda:0'), covar=tensor([0.0037, 0.0031, 0.0070, 0.0233, 0.0037, 0.1669, 0.0061, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0088, 0.0134, 0.0133, 0.0100, 0.0145, 0.0118, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 11:59:52,984 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55148.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 11:59:56,596 INFO [train.py:904] (0/8) Epoch 6, batch 4400, loss[loss=0.2585, simple_loss=0.3409, pruned_loss=0.08801, over 16546.00 frames. ], tot_loss[loss=0.2271, simple_loss=0.3071, pruned_loss=0.07352, over 3220154.54 frames. ], batch size: 75, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:00:27,151 INFO [optim.py:368] (0/8) 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 12:00:49,166 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6070, 3.9213, 4.1628, 2.0031, 4.5047, 4.6119, 3.0627, 3.4177], device='cuda:0'), covar=tensor([0.0741, 0.0184, 0.0157, 0.1078, 0.0051, 0.0031, 0.0335, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0095, 0.0081, 0.0139, 0.0072, 0.0079, 0.0115, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 12:00:57,994 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7860, 1.6258, 2.1837, 2.7814, 2.7247, 3.1263, 1.7582, 2.8889], device='cuda:0'), covar=tensor([0.0091, 0.0246, 0.0153, 0.0126, 0.0100, 0.0068, 0.0216, 0.0049], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0150, 0.0132, 0.0132, 0.0137, 0.0099, 0.0144, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 12:01:06,618 INFO [train.py:904] (0/8) Epoch 6, batch 4450, loss[loss=0.2494, simple_loss=0.3285, pruned_loss=0.08516, over 16860.00 frames. ], tot_loss[loss=0.2298, simple_loss=0.3104, pruned_loss=0.07454, over 3220448.11 frames. ], batch size: 116, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:01:18,687 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55209.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:21,637 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:01:31,365 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55218.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:01:37,082 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5722, 5.4461, 5.3834, 5.2348, 4.9526, 5.3345, 5.1872, 4.9720], device='cuda:0'), covar=tensor([0.0342, 0.0137, 0.0134, 0.0125, 0.0740, 0.0166, 0.0162, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0200, 0.0220, 0.0197, 0.0252, 0.0219, 0.0157, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:01:46,654 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3685, 5.6433, 5.3446, 5.5246, 5.0694, 4.7842, 5.2764, 5.7752], device='cuda:0'), covar=tensor([0.0614, 0.0625, 0.0819, 0.0431, 0.0547, 0.0567, 0.0495, 0.0561], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0520, 0.0435, 0.0333, 0.0317, 0.0334, 0.0422, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:02:16,927 INFO [train.py:904] (0/8) Epoch 6, batch 4500, loss[loss=0.2453, simple_loss=0.3192, pruned_loss=0.08571, over 16681.00 frames. ], tot_loss[loss=0.2297, simple_loss=0.3103, pruned_loss=0.07457, over 3222719.30 frames. ], batch size: 62, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:02:47,076 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:02:48,406 INFO [optim.py:368] (0/8) 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 12:03:01,010 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-28 12:03:29,866 INFO [train.py:904] (0/8) Epoch 6, batch 4550, loss[loss=0.248, simple_loss=0.309, pruned_loss=0.09349, over 11866.00 frames. ], tot_loss[loss=0.2295, simple_loss=0.3099, pruned_loss=0.07457, over 3228369.15 frames. ], batch size: 246, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:03:33,367 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7242, 1.3612, 1.5929, 1.6755, 1.7736, 1.9657, 1.4160, 1.7510], device='cuda:0'), covar=tensor([0.0101, 0.0196, 0.0096, 0.0164, 0.0114, 0.0073, 0.0188, 0.0044], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0148, 0.0130, 0.0131, 0.0137, 0.0099, 0.0143, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 12:03:57,418 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55320.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:04:41,590 INFO [train.py:904] (0/8) Epoch 6, batch 4600, loss[loss=0.2072, simple_loss=0.2896, pruned_loss=0.06241, over 17035.00 frames. ], tot_loss[loss=0.2285, simple_loss=0.3099, pruned_loss=0.07361, over 3233517.61 frames. ], batch size: 55, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:05:07,047 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55368.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:05:13,325 INFO [optim.py:368] (0/8) 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 12:05:52,747 INFO [train.py:904] (0/8) Epoch 6, batch 4650, loss[loss=0.2336, simple_loss=0.3179, pruned_loss=0.07467, over 15418.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3083, pruned_loss=0.07327, over 3231341.33 frames. ], batch size: 190, lr: 1.14e-02, grad_scale: 8.0 2023-04-28 12:06:12,017 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6954, 3.5511, 3.7025, 3.6129, 3.6622, 4.1522, 3.8805, 3.5798], device='cuda:0'), covar=tensor([0.2114, 0.1855, 0.1533, 0.2230, 0.2807, 0.1552, 0.1265, 0.2496], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0396, 0.0393, 0.0347, 0.0454, 0.0417, 0.0324, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 12:07:03,019 INFO [train.py:904] (0/8) Epoch 6, batch 4700, loss[loss=0.217, simple_loss=0.2981, pruned_loss=0.06794, over 16922.00 frames. ], tot_loss[loss=0.2244, simple_loss=0.3054, pruned_loss=0.07171, over 3227443.37 frames. ], batch size: 109, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:07:34,100 INFO [optim.py:368] (0/8) 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 12:08:12,132 INFO [train.py:904] (0/8) Epoch 6, batch 4750, loss[loss=0.2049, simple_loss=0.2849, pruned_loss=0.06246, over 16823.00 frames. ], tot_loss[loss=0.2204, simple_loss=0.3013, pruned_loss=0.06977, over 3234877.55 frames. ], batch size: 39, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:08:17,382 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:08:36,505 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55518.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:09:22,945 INFO [train.py:904] (0/8) Epoch 6, batch 4800, loss[loss=0.209, simple_loss=0.2943, pruned_loss=0.06182, over 16714.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2979, pruned_loss=0.06807, over 3234777.53 frames. ], batch size: 134, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:09:45,467 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55566.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:46,600 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=55567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:09:53,993 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.469e+02 2.878e+02 3.568e+02 5.612e+02, threshold=5.756e+02, percent-clipped=0.0 2023-04-28 12:10:35,271 INFO [train.py:904] (0/8) Epoch 6, batch 4850, loss[loss=0.2181, simple_loss=0.3052, pruned_loss=0.06548, over 16755.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2999, pruned_loss=0.06805, over 3221874.34 frames. ], batch size: 124, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:11:48,114 INFO [train.py:904] (0/8) Epoch 6, batch 4900, loss[loss=0.2165, simple_loss=0.295, pruned_loss=0.06903, over 11923.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.2985, pruned_loss=0.06647, over 3216095.63 frames. ], batch size: 246, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:12:11,272 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9922, 5.8130, 5.9255, 5.5688, 5.7788, 6.2056, 5.8233, 5.6087], device='cuda:0'), covar=tensor([0.0716, 0.1227, 0.1105, 0.1765, 0.2268, 0.1051, 0.0957, 0.2240], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0387, 0.0384, 0.0340, 0.0449, 0.0418, 0.0322, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 12:12:19,107 INFO [optim.py:368] (0/8) 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 12:12:39,485 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5936, 4.9041, 4.5898, 4.6564, 4.3088, 4.2878, 4.3894, 4.9435], device='cuda:0'), covar=tensor([0.0788, 0.0683, 0.0993, 0.0521, 0.0698, 0.0907, 0.0735, 0.0758], device='cuda:0'), in_proj_covar=tensor([0.0394, 0.0521, 0.0441, 0.0336, 0.0322, 0.0337, 0.0425, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:12:59,379 INFO [train.py:904] (0/8) Epoch 6, batch 4950, loss[loss=0.227, simple_loss=0.313, pruned_loss=0.07047, over 16478.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.2987, pruned_loss=0.06598, over 3199081.64 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:13:36,266 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9471, 4.0214, 1.6958, 4.6675, 2.5741, 4.3483, 2.0585, 2.6964], device='cuda:0'), covar=tensor([0.0137, 0.0197, 0.1828, 0.0020, 0.0785, 0.0230, 0.1454, 0.0715], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0156, 0.0177, 0.0082, 0.0161, 0.0186, 0.0185, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:13:49,040 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 12:14:05,675 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0755, 3.5565, 3.8454, 1.3399, 4.0585, 4.0995, 2.8347, 2.6664], device='cuda:0'), covar=tensor([0.0950, 0.0153, 0.0103, 0.1472, 0.0043, 0.0047, 0.0390, 0.0535], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0096, 0.0082, 0.0142, 0.0073, 0.0080, 0.0119, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 12:14:05,689 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55748.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:14:08,603 INFO [train.py:904] (0/8) Epoch 6, batch 5000, loss[loss=0.2029, simple_loss=0.2894, pruned_loss=0.0582, over 16664.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3007, pruned_loss=0.06682, over 3203616.39 frames. ], batch size: 57, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:14:26,514 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7559, 4.0871, 2.9949, 2.2917, 3.2620, 2.2931, 4.3380, 4.2494], device='cuda:0'), covar=tensor([0.2105, 0.0589, 0.1420, 0.1496, 0.1851, 0.1447, 0.0425, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0251, 0.0270, 0.0250, 0.0283, 0.0201, 0.0247, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:14:38,598 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.723e+02 3.199e+02 3.755e+02 7.199e+02, threshold=6.398e+02, percent-clipped=5.0 2023-04-28 12:15:21,207 INFO [train.py:904] (0/8) Epoch 6, batch 5050, loss[loss=0.2174, simple_loss=0.3044, pruned_loss=0.06517, over 16290.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3016, pruned_loss=0.06723, over 3197237.39 frames. ], batch size: 165, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:15:25,521 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:15:31,676 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 12:15:33,296 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55809.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:00,079 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5434, 1.9658, 2.1366, 4.1218, 1.7636, 2.6562, 2.1751, 2.1540], device='cuda:0'), covar=tensor([0.0654, 0.2523, 0.1400, 0.0300, 0.3348, 0.1621, 0.2196, 0.2589], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0338, 0.0278, 0.0313, 0.0379, 0.0350, 0.0304, 0.0400], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:16:28,631 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=55848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:31,567 INFO [train.py:904] (0/8) Epoch 6, batch 5100, loss[loss=0.2134, simple_loss=0.2957, pruned_loss=0.0655, over 16582.00 frames. ], tot_loss[loss=0.2154, simple_loss=0.2992, pruned_loss=0.06586, over 3211977.18 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:16:32,964 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55852.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:16:54,837 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=55867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:17:04,481 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.683e+02 3.138e+02 3.930e+02 6.713e+02, threshold=6.277e+02, percent-clipped=2.0 2023-04-28 12:17:45,244 INFO [train.py:904] (0/8) Epoch 6, batch 5150, loss[loss=0.2344, simple_loss=0.3197, pruned_loss=0.0745, over 16533.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2995, pruned_loss=0.06467, over 3215315.18 frames. ], batch size: 75, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:17:57,061 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=55909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:05,650 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=55915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:18:31,875 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 12:18:58,946 INFO [train.py:904] (0/8) Epoch 6, batch 5200, loss[loss=0.2032, simple_loss=0.2875, pruned_loss=0.05942, over 15417.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2982, pruned_loss=0.06426, over 3205861.04 frames. ], batch size: 190, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:19:30,993 INFO [optim.py:368] (0/8) 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 12:20:11,266 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-56000.pt 2023-04-28 12:20:15,820 INFO [train.py:904] (0/8) Epoch 6, batch 5250, loss[loss=0.1882, simple_loss=0.2799, pruned_loss=0.0482, over 16671.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2958, pruned_loss=0.06416, over 3197836.35 frames. ], batch size: 89, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:20:20,374 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-28 12:21:28,024 INFO [train.py:904] (0/8) Epoch 6, batch 5300, loss[loss=0.2087, simple_loss=0.2819, pruned_loss=0.06773, over 12044.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2922, pruned_loss=0.06325, over 3182703.70 frames. ], batch size: 248, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:21:59,804 INFO [optim.py:368] (0/8) 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 12:22:40,529 INFO [train.py:904] (0/8) Epoch 6, batch 5350, loss[loss=0.1982, simple_loss=0.2987, pruned_loss=0.04883, over 16732.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2903, pruned_loss=0.06233, over 3191419.50 frames. ], batch size: 89, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:22:45,577 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:22:45,719 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9449, 3.4084, 3.2452, 2.1894, 2.9109, 3.2465, 3.2662, 1.8666], device='cuda:0'), covar=tensor([0.0353, 0.0017, 0.0027, 0.0233, 0.0049, 0.0046, 0.0031, 0.0308], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0055, 0.0060, 0.0115, 0.0062, 0.0071, 0.0065, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 12:23:40,712 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3591, 4.3727, 4.1794, 2.9472, 3.6634, 4.2009, 3.9417, 2.4586], device='cuda:0'), covar=tensor([0.0365, 0.0013, 0.0025, 0.0221, 0.0047, 0.0040, 0.0033, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0056, 0.0060, 0.0116, 0.0062, 0.0071, 0.0065, 0.0109], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 12:23:44,951 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56145.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:23:52,601 INFO [train.py:904] (0/8) Epoch 6, batch 5400, loss[loss=0.2575, simple_loss=0.3292, pruned_loss=0.09291, over 12486.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.294, pruned_loss=0.06385, over 3191485.01 frames. ], batch size: 247, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:24:26,637 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.771e+02 2.610e+02 3.124e+02 3.605e+02 7.209e+02, threshold=6.248e+02, percent-clipped=3.0 2023-04-28 12:25:08,999 INFO [train.py:904] (0/8) Epoch 6, batch 5450, loss[loss=0.2539, simple_loss=0.3278, pruned_loss=0.08998, over 17009.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2972, pruned_loss=0.06595, over 3182645.47 frames. ], batch size: 50, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:25:13,269 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7591, 2.5143, 2.3843, 3.6461, 2.6824, 3.6698, 1.5029, 2.8254], device='cuda:0'), covar=tensor([0.1156, 0.0524, 0.1013, 0.0110, 0.0204, 0.0331, 0.1370, 0.0650], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0143, 0.0170, 0.0094, 0.0188, 0.0186, 0.0162, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 12:25:14,388 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56204.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:17,623 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:25:41,355 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-04-28 12:26:21,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7682, 3.3566, 2.9286, 1.7014, 2.6288, 2.0413, 3.0447, 3.3045], device='cuda:0'), covar=tensor([0.0311, 0.0473, 0.0612, 0.1662, 0.0761, 0.0965, 0.0742, 0.0760], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0126, 0.0152, 0.0138, 0.0129, 0.0123, 0.0137, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:26:27,336 INFO [train.py:904] (0/8) Epoch 6, batch 5500, loss[loss=0.2329, simple_loss=0.3191, pruned_loss=0.07334, over 16654.00 frames. ], tot_loss[loss=0.2264, simple_loss=0.3067, pruned_loss=0.07305, over 3142736.08 frames. ], batch size: 76, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:26:55,074 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8941, 1.7002, 1.3425, 1.4612, 1.8512, 1.6405, 1.8025, 1.8803], device='cuda:0'), covar=tensor([0.0048, 0.0133, 0.0217, 0.0183, 0.0101, 0.0148, 0.0096, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0090, 0.0163, 0.0166, 0.0165, 0.0159, 0.0169, 0.0149, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:27:01,685 INFO [optim.py:368] (0/8) 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 12:27:32,088 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0409, 3.8597, 4.0183, 4.1939, 4.2978, 3.8696, 4.2815, 4.2796], device='cuda:0'), covar=tensor([0.0995, 0.0730, 0.1257, 0.0533, 0.0495, 0.1128, 0.0509, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0502, 0.0649, 0.0521, 0.0397, 0.0393, 0.0405, 0.0439], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:27:46,070 INFO [train.py:904] (0/8) Epoch 6, batch 5550, loss[loss=0.2823, simple_loss=0.3536, pruned_loss=0.1055, over 16686.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3153, pruned_loss=0.08002, over 3114848.77 frames. ], batch size: 124, lr: 1.13e-02, grad_scale: 4.0 2023-04-28 12:27:59,303 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5714, 2.6180, 1.6975, 2.6976, 2.1802, 2.7067, 1.9764, 2.3235], device='cuda:0'), covar=tensor([0.0193, 0.0408, 0.1204, 0.0090, 0.0632, 0.0496, 0.1058, 0.0571], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0158, 0.0179, 0.0082, 0.0165, 0.0190, 0.0190, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:28:10,344 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7226, 3.5518, 3.7198, 3.6103, 3.7569, 4.1604, 3.9506, 3.7270], device='cuda:0'), covar=tensor([0.1711, 0.2041, 0.1809, 0.2203, 0.2319, 0.1332, 0.1217, 0.2383], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0399, 0.0400, 0.0343, 0.0458, 0.0425, 0.0327, 0.0469], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 12:29:07,887 INFO [train.py:904] (0/8) Epoch 6, batch 5600, loss[loss=0.2341, simple_loss=0.3149, pruned_loss=0.07664, over 17208.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3223, pruned_loss=0.08635, over 3070337.54 frames. ], batch size: 45, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:29:45,256 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.502e+02 4.330e+02 5.114e+02 6.903e+02 1.749e+03, threshold=1.023e+03, percent-clipped=9.0 2023-04-28 12:30:29,973 INFO [train.py:904] (0/8) Epoch 6, batch 5650, loss[loss=0.2651, simple_loss=0.3444, pruned_loss=0.09295, over 16685.00 frames. ], tot_loss[loss=0.2566, simple_loss=0.329, pruned_loss=0.0921, over 3052926.68 frames. ], batch size: 89, lr: 1.13e-02, grad_scale: 8.0 2023-04-28 12:30:34,253 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56404.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:30:44,568 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 12:31:48,518 INFO [train.py:904] (0/8) Epoch 6, batch 5700, loss[loss=0.272, simple_loss=0.3482, pruned_loss=0.0979, over 16704.00 frames. ], tot_loss[loss=0.2581, simple_loss=0.3302, pruned_loss=0.09298, over 3072722.00 frames. ], batch size: 76, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:31:51,187 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:32:25,514 INFO [optim.py:368] (0/8) 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 12:33:08,900 INFO [train.py:904] (0/8) Epoch 6, batch 5750, loss[loss=0.2715, simple_loss=0.3298, pruned_loss=0.1066, over 11545.00 frames. ], tot_loss[loss=0.2629, simple_loss=0.3339, pruned_loss=0.09597, over 3030257.46 frames. ], batch size: 248, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:33:09,313 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:13,903 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:33:19,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9236, 1.7134, 1.4509, 1.5682, 1.9170, 1.6659, 1.7878, 1.8805], device='cuda:0'), covar=tensor([0.0053, 0.0134, 0.0186, 0.0176, 0.0094, 0.0135, 0.0111, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0158, 0.0162, 0.0158, 0.0153, 0.0163, 0.0143, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:33:44,535 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56523.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:33:49,425 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7019, 3.9488, 4.1287, 2.0108, 4.4363, 4.4023, 3.1988, 3.2555], device='cuda:0'), covar=tensor([0.0719, 0.0137, 0.0164, 0.1121, 0.0048, 0.0064, 0.0268, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0092, 0.0080, 0.0139, 0.0072, 0.0080, 0.0114, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 12:34:30,321 INFO [train.py:904] (0/8) Epoch 6, batch 5800, loss[loss=0.2206, simple_loss=0.312, pruned_loss=0.06454, over 16522.00 frames. ], tot_loss[loss=0.2607, simple_loss=0.3333, pruned_loss=0.09406, over 3027217.77 frames. ], batch size: 75, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:34:32,755 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:34:36,071 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1488, 3.2301, 3.2472, 1.5541, 3.4815, 3.4552, 2.7633, 2.6211], device='cuda:0'), covar=tensor([0.0790, 0.0144, 0.0158, 0.1220, 0.0069, 0.0109, 0.0305, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0092, 0.0080, 0.0139, 0.0072, 0.0080, 0.0113, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 12:35:01,177 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1315, 3.3031, 3.5170, 3.5027, 3.4880, 3.3016, 3.3058, 3.3617], device='cuda:0'), covar=tensor([0.0335, 0.0446, 0.0393, 0.0468, 0.0488, 0.0413, 0.0756, 0.0457], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0247, 0.0252, 0.0253, 0.0304, 0.0272, 0.0373, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 12:35:05,968 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.365e+02 3.759e+02 4.783e+02 6.283e+02 1.634e+03, threshold=9.566e+02, percent-clipped=2.0 2023-04-28 12:35:22,945 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56584.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 12:35:49,161 INFO [train.py:904] (0/8) Epoch 6, batch 5850, loss[loss=0.2421, simple_loss=0.3217, pruned_loss=0.08122, over 16731.00 frames. ], tot_loss[loss=0.2555, simple_loss=0.3295, pruned_loss=0.09072, over 3046216.36 frames. ], batch size: 83, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:35:56,132 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9681, 3.9172, 3.8913, 3.3573, 3.8877, 1.7182, 3.6363, 3.6636], device='cuda:0'), covar=tensor([0.0076, 0.0076, 0.0104, 0.0275, 0.0069, 0.1944, 0.0098, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0087, 0.0134, 0.0131, 0.0099, 0.0150, 0.0115, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:37:11,656 INFO [train.py:904] (0/8) Epoch 6, batch 5900, loss[loss=0.2454, simple_loss=0.3297, pruned_loss=0.08055, over 16756.00 frames. ], tot_loss[loss=0.2539, simple_loss=0.3282, pruned_loss=0.08976, over 3056213.20 frames. ], batch size: 83, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:37:52,140 INFO [optim.py:368] (0/8) 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 12:38:33,945 INFO [train.py:904] (0/8) Epoch 6, batch 5950, loss[loss=0.2632, simple_loss=0.3343, pruned_loss=0.09605, over 11698.00 frames. ], tot_loss[loss=0.252, simple_loss=0.3283, pruned_loss=0.08781, over 3056197.71 frames. ], batch size: 249, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:11,063 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6634, 2.6809, 1.7053, 2.7649, 2.1757, 2.7235, 1.9871, 2.3174], device='cuda:0'), covar=tensor([0.0193, 0.0322, 0.1189, 0.0091, 0.0577, 0.0464, 0.1034, 0.0524], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0152, 0.0174, 0.0081, 0.0160, 0.0185, 0.0184, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:39:29,587 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 12:39:52,048 INFO [train.py:904] (0/8) Epoch 6, batch 6000, loss[loss=0.2715, simple_loss=0.3387, pruned_loss=0.1021, over 16417.00 frames. ], tot_loss[loss=0.2494, simple_loss=0.3264, pruned_loss=0.08627, over 3088969.07 frames. ], batch size: 146, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:39:52,049 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 12:40:01,530 INFO [train.py:938] (0/8) Epoch 6, validation: loss=0.18, simple_loss=0.2922, pruned_loss=0.03386, over 944034.00 frames. 2023-04-28 12:40:01,530 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 12:40:27,223 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3400, 1.4091, 1.9173, 2.4203, 2.2946, 2.6416, 1.6173, 2.3621], device='cuda:0'), covar=tensor([0.0093, 0.0257, 0.0153, 0.0139, 0.0123, 0.0088, 0.0235, 0.0061], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0149, 0.0130, 0.0132, 0.0136, 0.0099, 0.0146, 0.0087], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 12:40:36,513 INFO [optim.py:368] (0/8) 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 12:41:19,700 INFO [train.py:904] (0/8) Epoch 6, batch 6050, loss[loss=0.239, simple_loss=0.3155, pruned_loss=0.08129, over 16658.00 frames. ], tot_loss[loss=0.2471, simple_loss=0.324, pruned_loss=0.08514, over 3103970.36 frames. ], batch size: 57, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:41:20,680 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=56801.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:41:23,998 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56803.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:41:25,537 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 12:41:33,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1880, 3.3403, 1.7102, 3.5190, 2.3442, 3.4522, 1.8907, 2.4990], device='cuda:0'), covar=tensor([0.0156, 0.0310, 0.1572, 0.0065, 0.0780, 0.0505, 0.1452, 0.0691], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0154, 0.0174, 0.0081, 0.0162, 0.0187, 0.0185, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:41:42,098 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4436, 3.4557, 2.6359, 2.1284, 2.4705, 2.0824, 3.5105, 3.5094], device='cuda:0'), covar=tensor([0.2351, 0.0682, 0.1342, 0.1801, 0.1992, 0.1594, 0.0442, 0.0712], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0247, 0.0268, 0.0251, 0.0284, 0.0202, 0.0248, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:42:08,541 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 12:42:15,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7021, 3.9028, 1.8724, 4.2839, 2.6160, 4.1579, 2.1855, 2.7486], device='cuda:0'), covar=tensor([0.0166, 0.0256, 0.1694, 0.0048, 0.0852, 0.0347, 0.1535, 0.0730], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0154, 0.0175, 0.0081, 0.0163, 0.0188, 0.0186, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:42:33,598 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=56849.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:42:36,674 INFO [train.py:904] (0/8) Epoch 6, batch 6100, loss[loss=0.2278, simple_loss=0.3102, pruned_loss=0.07266, over 16655.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3234, pruned_loss=0.08348, over 3126942.96 frames. ], batch size: 134, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:42:56,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7883, 3.1134, 2.5697, 4.7552, 3.9709, 4.3315, 1.6829, 2.7222], device='cuda:0'), covar=tensor([0.1187, 0.0486, 0.1064, 0.0082, 0.0290, 0.0303, 0.1211, 0.0876], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0146, 0.0170, 0.0096, 0.0200, 0.0191, 0.0165, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 12:42:58,704 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56864.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:43:12,443 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8869, 3.0278, 2.9989, 4.8425, 3.9579, 4.4940, 1.8608, 3.1920], device='cuda:0'), covar=tensor([0.1202, 0.0565, 0.0889, 0.0078, 0.0314, 0.0273, 0.1181, 0.0700], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0145, 0.0170, 0.0096, 0.0199, 0.0190, 0.0165, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 12:43:13,035 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.259e+02 3.305e+02 4.282e+02 5.206e+02 1.267e+03, threshold=8.564e+02, percent-clipped=3.0 2023-04-28 12:43:20,783 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=56879.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 12:43:46,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8926, 1.4977, 2.3476, 2.8189, 2.5988, 3.0652, 1.4696, 2.9193], device='cuda:0'), covar=tensor([0.0069, 0.0283, 0.0137, 0.0109, 0.0109, 0.0072, 0.0309, 0.0041], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0148, 0.0130, 0.0129, 0.0133, 0.0099, 0.0144, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 12:43:50,706 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 12:43:56,261 INFO [train.py:904] (0/8) Epoch 6, batch 6150, loss[loss=0.2312, simple_loss=0.3176, pruned_loss=0.07235, over 16816.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.322, pruned_loss=0.08346, over 3107957.54 frames. ], batch size: 102, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:44:08,074 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56908.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:12,440 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7981, 3.5149, 3.1183, 1.8342, 2.6658, 2.0952, 3.3035, 3.3633], device='cuda:0'), covar=tensor([0.0270, 0.0482, 0.0539, 0.1618, 0.0768, 0.0968, 0.0580, 0.0703], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0126, 0.0152, 0.0139, 0.0132, 0.0124, 0.0138, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:45:17,273 INFO [train.py:904] (0/8) Epoch 6, batch 6200, loss[loss=0.2438, simple_loss=0.3125, pruned_loss=0.08757, over 16623.00 frames. ], tot_loss[loss=0.2432, simple_loss=0.3202, pruned_loss=0.08307, over 3098941.90 frames. ], batch size: 62, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:45:45,660 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=56969.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:45:53,733 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.179e+02 3.479e+02 4.387e+02 5.634e+02 1.000e+03, threshold=8.774e+02, percent-clipped=2.0 2023-04-28 12:45:59,611 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2955, 1.8600, 2.1299, 3.7935, 1.8046, 2.5725, 2.1359, 1.9879], device='cuda:0'), covar=tensor([0.0693, 0.2610, 0.1410, 0.0361, 0.3280, 0.1423, 0.2279, 0.2491], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0339, 0.0278, 0.0315, 0.0385, 0.0349, 0.0304, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:46:22,658 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56993.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:25,942 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=56995.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:46:34,337 INFO [train.py:904] (0/8) Epoch 6, batch 6250, loss[loss=0.2098, simple_loss=0.2998, pruned_loss=0.05993, over 16787.00 frames. ], tot_loss[loss=0.2424, simple_loss=0.3196, pruned_loss=0.08261, over 3112122.21 frames. ], batch size: 76, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:46:45,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7805, 6.1593, 5.7807, 5.8604, 5.3372, 5.1954, 5.5662, 6.2113], device='cuda:0'), covar=tensor([0.0703, 0.0555, 0.0892, 0.0524, 0.0702, 0.0489, 0.0761, 0.0587], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0528, 0.0449, 0.0348, 0.0326, 0.0342, 0.0435, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:47:50,603 INFO [train.py:904] (0/8) Epoch 6, batch 6300, loss[loss=0.2062, simple_loss=0.2957, pruned_loss=0.05836, over 16825.00 frames. ], tot_loss[loss=0.2404, simple_loss=0.3187, pruned_loss=0.08107, over 3133499.11 frames. ], batch size: 102, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:47:56,066 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:47:59,213 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.167e+02 3.660e+02 4.420e+02 5.766e+02 1.393e+03, threshold=8.841e+02, percent-clipped=5.0 2023-04-28 12:48:30,440 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9409, 2.3330, 1.7691, 1.9928, 2.8035, 2.5093, 3.0247, 3.0565], device='cuda:0'), covar=tensor([0.0045, 0.0205, 0.0286, 0.0250, 0.0117, 0.0172, 0.0116, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0160, 0.0163, 0.0161, 0.0156, 0.0165, 0.0148, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:49:09,392 INFO [train.py:904] (0/8) Epoch 6, batch 6350, loss[loss=0.2089, simple_loss=0.2884, pruned_loss=0.06474, over 17132.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3193, pruned_loss=0.08196, over 3147128.37 frames. ], batch size: 48, lr: 1.12e-02, grad_scale: 4.0 2023-04-28 12:50:26,705 INFO [train.py:904] (0/8) Epoch 6, batch 6400, loss[loss=0.2405, simple_loss=0.3121, pruned_loss=0.08441, over 16508.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3204, pruned_loss=0.08425, over 3119639.91 frames. ], batch size: 75, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:50:38,661 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57159.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:50:42,192 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5803, 3.6151, 2.7785, 2.1564, 2.5595, 2.1645, 3.7046, 3.6377], device='cuda:0'), covar=tensor([0.2278, 0.0649, 0.1282, 0.1655, 0.1874, 0.1547, 0.0420, 0.0741], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0248, 0.0267, 0.0252, 0.0282, 0.0202, 0.0247, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:50:49,409 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3837, 2.9421, 2.5659, 2.2272, 2.2651, 2.1407, 2.8789, 2.9770], device='cuda:0'), covar=tensor([0.1808, 0.0669, 0.1106, 0.1500, 0.1766, 0.1532, 0.0462, 0.0831], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0248, 0.0267, 0.0252, 0.0282, 0.0201, 0.0247, 0.0261], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:51:01,323 INFO [optim.py:368] (0/8) 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 12:51:07,819 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57179.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:51:19,355 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57186.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:36,341 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57197.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:51:42,174 INFO [train.py:904] (0/8) Epoch 6, batch 6450, loss[loss=0.2356, simple_loss=0.3142, pruned_loss=0.07852, over 16687.00 frames. ], tot_loss[loss=0.2439, simple_loss=0.3201, pruned_loss=0.08385, over 3099785.23 frames. ], batch size: 134, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:51:55,766 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6579, 2.6691, 1.7998, 2.7470, 2.1770, 2.7215, 1.9873, 2.3735], device='cuda:0'), covar=tensor([0.0169, 0.0389, 0.1179, 0.0080, 0.0586, 0.0627, 0.1229, 0.0484], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0153, 0.0176, 0.0081, 0.0161, 0.0188, 0.0187, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:52:26,742 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57227.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 12:52:52,977 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57244.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:52:57,213 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57247.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:03,835 INFO [train.py:904] (0/8) Epoch 6, batch 6500, loss[loss=0.222, simple_loss=0.3012, pruned_loss=0.07146, over 16961.00 frames. ], tot_loss[loss=0.242, simple_loss=0.3178, pruned_loss=0.08314, over 3095267.71 frames. ], batch size: 41, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:53:15,554 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57258.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 12:53:25,371 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57264.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:53:25,514 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7862, 4.7420, 4.5739, 3.9423, 4.5679, 1.6306, 4.4070, 4.5123], device='cuda:0'), covar=tensor([0.0059, 0.0054, 0.0091, 0.0280, 0.0063, 0.2010, 0.0081, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0086, 0.0130, 0.0129, 0.0098, 0.0148, 0.0113, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 12:53:41,781 INFO [optim.py:368] (0/8) 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 12:54:09,013 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2780, 3.6293, 3.7534, 1.5409, 3.9169, 3.9541, 3.0462, 2.8817], device='cuda:0'), covar=tensor([0.0831, 0.0110, 0.0130, 0.1242, 0.0057, 0.0077, 0.0302, 0.0409], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0091, 0.0082, 0.0138, 0.0072, 0.0081, 0.0115, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 12:54:25,249 INFO [train.py:904] (0/8) Epoch 6, batch 6550, loss[loss=0.2526, simple_loss=0.3453, pruned_loss=0.07999, over 15297.00 frames. ], tot_loss[loss=0.2446, simple_loss=0.3213, pruned_loss=0.08396, over 3104371.58 frames. ], batch size: 190, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:54:31,753 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57305.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:55:05,123 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 12:55:40,101 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57349.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:55:42,179 INFO [train.py:904] (0/8) Epoch 6, batch 6600, loss[loss=0.3123, simple_loss=0.3605, pruned_loss=0.1321, over 11357.00 frames. ], tot_loss[loss=0.2464, simple_loss=0.3237, pruned_loss=0.08451, over 3102057.52 frames. ], batch size: 247, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:55:42,532 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57351.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:56:18,771 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.503e+02 3.780e+02 4.683e+02 5.878e+02 1.016e+03, threshold=9.365e+02, percent-clipped=5.0 2023-04-28 12:56:56,481 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57398.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:57:00,526 INFO [train.py:904] (0/8) Epoch 6, batch 6650, loss[loss=0.218, simple_loss=0.3047, pruned_loss=0.06563, over 16939.00 frames. ], tot_loss[loss=0.2469, simple_loss=0.3236, pruned_loss=0.08505, over 3108194.66 frames. ], batch size: 96, lr: 1.12e-02, grad_scale: 8.0 2023-04-28 12:58:15,923 INFO [train.py:904] (0/8) Epoch 6, batch 6700, loss[loss=0.2354, simple_loss=0.3069, pruned_loss=0.08194, over 16996.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3214, pruned_loss=0.08421, over 3113740.90 frames. ], batch size: 55, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:58:29,454 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:29,618 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57459.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:34,653 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57462.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:58:54,096 INFO [optim.py:368] (0/8) 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 12:58:56,554 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7831, 3.2874, 2.7441, 4.8904, 4.0769, 4.5230, 1.6613, 3.4303], device='cuda:0'), covar=tensor([0.1403, 0.0539, 0.1116, 0.0091, 0.0336, 0.0308, 0.1450, 0.0691], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0147, 0.0173, 0.0096, 0.0199, 0.0194, 0.0167, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 12:59:24,794 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 12:59:34,120 INFO [train.py:904] (0/8) Epoch 6, batch 6750, loss[loss=0.2407, simple_loss=0.3188, pruned_loss=0.08133, over 16916.00 frames. ], tot_loss[loss=0.2445, simple_loss=0.3206, pruned_loss=0.08425, over 3113057.53 frames. ], batch size: 109, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 12:59:43,068 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57507.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:07,998 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57523.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:36,588 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57542.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:00:49,728 INFO [train.py:904] (0/8) Epoch 6, batch 6800, loss[loss=0.2964, simple_loss=0.3552, pruned_loss=0.1188, over 11652.00 frames. ], tot_loss[loss=0.2437, simple_loss=0.3203, pruned_loss=0.08349, over 3126002.19 frames. ], batch size: 248, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:00:53,684 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57553.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 13:00:58,186 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:09,837 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57564.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:01:27,507 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.272e+02 3.624e+02 4.504e+02 5.898e+02 1.165e+03, threshold=9.008e+02, percent-clipped=3.0 2023-04-28 13:02:06,711 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57600.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:07,777 INFO [train.py:904] (0/8) Epoch 6, batch 6850, loss[loss=0.2174, simple_loss=0.313, pruned_loss=0.06089, over 17095.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3222, pruned_loss=0.08476, over 3115833.42 frames. ], batch size: 49, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:02:21,505 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2684, 4.3292, 4.3814, 4.2705, 4.2446, 4.9016, 4.4762, 4.1445], device='cuda:0'), covar=tensor([0.1162, 0.1638, 0.1440, 0.1647, 0.2496, 0.0950, 0.1200, 0.2364], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0400, 0.0405, 0.0352, 0.0460, 0.0434, 0.0334, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 13:02:24,687 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57612.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:02:44,893 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2519, 3.3261, 1.5087, 3.4498, 2.2949, 3.4656, 1.6674, 2.5340], device='cuda:0'), covar=tensor([0.0157, 0.0296, 0.1641, 0.0085, 0.0815, 0.0461, 0.1680, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0156, 0.0180, 0.0084, 0.0164, 0.0192, 0.0191, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 13:03:22,268 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:03:24,934 INFO [train.py:904] (0/8) Epoch 6, batch 6900, loss[loss=0.253, simple_loss=0.3288, pruned_loss=0.08865, over 16682.00 frames. ], tot_loss[loss=0.2472, simple_loss=0.3247, pruned_loss=0.08488, over 3115328.43 frames. ], batch size: 62, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:03:25,282 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:02,710 INFO [optim.py:368] (0/8) 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 13:04:20,075 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4339, 3.3915, 3.3924, 2.9180, 3.3582, 2.0478, 3.1322, 2.9133], device='cuda:0'), covar=tensor([0.0101, 0.0084, 0.0118, 0.0207, 0.0080, 0.1557, 0.0107, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0084, 0.0130, 0.0128, 0.0097, 0.0148, 0.0113, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:04:37,889 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:04:43,540 INFO [train.py:904] (0/8) Epoch 6, batch 6950, loss[loss=0.2794, simple_loss=0.3475, pruned_loss=0.1056, over 15302.00 frames. ], tot_loss[loss=0.2521, simple_loss=0.3281, pruned_loss=0.08801, over 3092521.08 frames. ], batch size: 191, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:05:25,694 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2728, 3.4048, 1.6760, 3.6269, 2.3390, 3.5532, 1.8659, 2.5281], device='cuda:0'), covar=tensor([0.0173, 0.0340, 0.1618, 0.0065, 0.0777, 0.0440, 0.1550, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0154, 0.0177, 0.0083, 0.0161, 0.0190, 0.0187, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 13:05:33,588 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=57733.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:05:46,123 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-28 13:06:01,326 INFO [train.py:904] (0/8) Epoch 6, batch 7000, loss[loss=0.2628, simple_loss=0.345, pruned_loss=0.09031, over 16307.00 frames. ], tot_loss[loss=0.2512, simple_loss=0.3282, pruned_loss=0.08706, over 3096073.92 frames. ], batch size: 165, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:06:05,963 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57754.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:06:38,243 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 3.231e+02 4.685e+02 6.535e+02 1.606e+03, threshold=9.370e+02, percent-clipped=8.0 2023-04-28 13:07:07,520 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=57794.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:07:17,562 INFO [train.py:904] (0/8) Epoch 6, batch 7050, loss[loss=0.2262, simple_loss=0.306, pruned_loss=0.0732, over 16744.00 frames. ], tot_loss[loss=0.2504, simple_loss=0.3284, pruned_loss=0.08617, over 3107731.88 frames. ], batch size: 57, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:07:44,462 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:20,250 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57842.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:34,178 INFO [train.py:904] (0/8) Epoch 6, batch 7100, loss[loss=0.2313, simple_loss=0.3114, pruned_loss=0.0756, over 16797.00 frames. ], tot_loss[loss=0.2502, simple_loss=0.3273, pruned_loss=0.08654, over 3091597.58 frames. ], batch size: 83, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:08:34,657 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=57851.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:08:37,900 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57853.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 13:09:10,527 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57890.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:48,249 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=57900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:09:49,606 INFO [train.py:904] (0/8) Epoch 6, batch 7150, loss[loss=0.2797, simple_loss=0.3341, pruned_loss=0.1127, over 11520.00 frames. ], tot_loss[loss=0.2488, simple_loss=0.325, pruned_loss=0.0863, over 3077954.82 frames. ], batch size: 248, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:09:49,971 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:11:00,231 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=57948.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:11:04,899 INFO [train.py:904] (0/8) Epoch 6, batch 7200, loss[loss=0.2175, simple_loss=0.2978, pruned_loss=0.06858, over 16689.00 frames. ], tot_loss[loss=0.2452, simple_loss=0.3221, pruned_loss=0.0841, over 3085571.19 frames. ], batch size: 62, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:11:41,804 INFO [optim.py:368] (0/8) 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 13:12:24,004 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-58000.pt 2023-04-28 13:12:28,080 INFO [train.py:904] (0/8) Epoch 6, batch 7250, loss[loss=0.2719, simple_loss=0.3258, pruned_loss=0.1091, over 11702.00 frames. ], tot_loss[loss=0.243, simple_loss=0.32, pruned_loss=0.083, over 3075647.82 frames. ], batch size: 248, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:12:43,007 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3654, 3.6140, 3.0304, 5.2127, 4.3375, 4.5816, 2.2055, 3.5465], device='cuda:0'), covar=tensor([0.1064, 0.0502, 0.1035, 0.0079, 0.0327, 0.0302, 0.1149, 0.0652], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0142, 0.0169, 0.0094, 0.0195, 0.0190, 0.0163, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 13:12:58,403 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4031, 4.5882, 4.6685, 4.5762, 4.5822, 5.0702, 4.6139, 4.4057], device='cuda:0'), covar=tensor([0.1111, 0.1508, 0.1364, 0.1533, 0.2175, 0.0947, 0.1370, 0.2392], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0397, 0.0402, 0.0350, 0.0457, 0.0433, 0.0331, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 13:13:04,467 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9520, 2.1961, 1.7830, 2.0397, 2.5763, 2.3590, 3.0022, 2.8833], device='cuda:0'), covar=tensor([0.0043, 0.0213, 0.0288, 0.0239, 0.0135, 0.0205, 0.0098, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0159, 0.0164, 0.0162, 0.0155, 0.0165, 0.0145, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:13:44,374 INFO [train.py:904] (0/8) Epoch 6, batch 7300, loss[loss=0.292, simple_loss=0.3529, pruned_loss=0.1156, over 11849.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3183, pruned_loss=0.08211, over 3072880.80 frames. ], batch size: 247, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:13:49,480 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58054.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:13:49,981 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-28 13:14:15,962 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 13:14:21,944 INFO [optim.py:368] (0/8) 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 13:14:33,077 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8025, 1.6121, 1.4424, 1.3732, 1.7322, 1.4897, 1.6112, 1.7610], device='cuda:0'), covar=tensor([0.0046, 0.0139, 0.0201, 0.0168, 0.0098, 0.0138, 0.0080, 0.0080], device='cuda:0'), in_proj_covar=tensor([0.0085, 0.0158, 0.0164, 0.0161, 0.0155, 0.0165, 0.0145, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:14:43,297 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58089.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:02,133 INFO [train.py:904] (0/8) Epoch 6, batch 7350, loss[loss=0.2113, simple_loss=0.2919, pruned_loss=0.06541, over 16548.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3185, pruned_loss=0.08211, over 3066924.57 frames. ], batch size: 62, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:15:03,737 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58102.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:17,802 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-28 13:15:28,216 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:15:38,254 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9900, 2.6784, 2.6316, 1.7455, 2.8027, 2.8113, 2.4550, 2.2952], device='cuda:0'), covar=tensor([0.0690, 0.0141, 0.0147, 0.0954, 0.0078, 0.0108, 0.0371, 0.0436], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0089, 0.0080, 0.0137, 0.0071, 0.0079, 0.0114, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 13:16:04,403 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58141.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:08,531 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:19,323 INFO [train.py:904] (0/8) Epoch 6, batch 7400, loss[loss=0.2587, simple_loss=0.3366, pruned_loss=0.09035, over 15401.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3198, pruned_loss=0.08281, over 3075222.72 frames. ], batch size: 190, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:16:19,774 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58151.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:43,125 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:16:57,166 INFO [optim.py:368] (0/8) 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 13:17:16,445 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3465, 3.4742, 1.8397, 3.7210, 2.4003, 3.6321, 1.9727, 2.5087], device='cuda:0'), covar=tensor([0.0173, 0.0303, 0.1530, 0.0073, 0.0764, 0.0477, 0.1406, 0.0731], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0152, 0.0179, 0.0084, 0.0161, 0.0187, 0.0188, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 13:17:34,284 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58199.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:37,073 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 13:17:37,744 INFO [train.py:904] (0/8) Epoch 6, batch 7450, loss[loss=0.2777, simple_loss=0.33, pruned_loss=0.1127, over 11388.00 frames. ], tot_loss[loss=0.2453, simple_loss=0.3218, pruned_loss=0.08439, over 3083469.31 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:17:39,871 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58202.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:17:40,118 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 13:17:45,749 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58205.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:18:02,351 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2220, 1.8915, 1.6145, 1.5988, 2.0215, 1.8354, 2.1203, 2.2070], device='cuda:0'), covar=tensor([0.0055, 0.0196, 0.0255, 0.0251, 0.0133, 0.0201, 0.0124, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0159, 0.0165, 0.0161, 0.0155, 0.0166, 0.0147, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:18:18,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 13:18:27,299 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 13:18:59,289 INFO [train.py:904] (0/8) Epoch 6, batch 7500, loss[loss=0.2238, simple_loss=0.3038, pruned_loss=0.07186, over 16687.00 frames. ], tot_loss[loss=0.2463, simple_loss=0.323, pruned_loss=0.08484, over 3057572.65 frames. ], batch size: 89, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:19:39,442 INFO [optim.py:368] (0/8) 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 13:20:16,977 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3995, 1.8916, 2.1337, 3.8784, 1.8378, 2.6278, 2.1359, 2.0581], device='cuda:0'), covar=tensor([0.0791, 0.2960, 0.1541, 0.0439, 0.3807, 0.1595, 0.2393, 0.2921], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0341, 0.0280, 0.0316, 0.0387, 0.0353, 0.0304, 0.0401], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:20:17,498 INFO [train.py:904] (0/8) Epoch 6, batch 7550, loss[loss=0.2312, simple_loss=0.3143, pruned_loss=0.07407, over 16769.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3213, pruned_loss=0.08448, over 3064364.23 frames. ], batch size: 89, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:20:38,067 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2989, 3.7286, 3.9535, 1.6288, 4.2264, 4.2122, 2.9671, 2.9571], device='cuda:0'), covar=tensor([0.0920, 0.0145, 0.0158, 0.1411, 0.0047, 0.0076, 0.0385, 0.0507], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0092, 0.0083, 0.0142, 0.0073, 0.0083, 0.0119, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 13:20:42,384 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3958, 1.4701, 1.9847, 2.4635, 2.3695, 2.8163, 1.5837, 2.4653], device='cuda:0'), covar=tensor([0.0099, 0.0275, 0.0175, 0.0153, 0.0140, 0.0082, 0.0260, 0.0067], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0147, 0.0128, 0.0130, 0.0134, 0.0098, 0.0145, 0.0085], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 13:21:33,636 INFO [train.py:904] (0/8) Epoch 6, batch 7600, loss[loss=0.2708, simple_loss=0.3276, pruned_loss=0.107, over 11484.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3201, pruned_loss=0.08428, over 3061443.60 frames. ], batch size: 248, lr: 1.11e-02, grad_scale: 8.0 2023-04-28 13:22:14,298 INFO [optim.py:368] (0/8) 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 13:22:33,722 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:22:52,239 INFO [train.py:904] (0/8) Epoch 6, batch 7650, loss[loss=0.2536, simple_loss=0.3258, pruned_loss=0.09066, over 15366.00 frames. ], tot_loss[loss=0.2448, simple_loss=0.3207, pruned_loss=0.08447, over 3080793.84 frames. ], batch size: 190, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:23:49,256 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58437.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:24:10,619 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 13:24:11,090 INFO [train.py:904] (0/8) Epoch 6, batch 7700, loss[loss=0.3149, simple_loss=0.3679, pruned_loss=0.131, over 11750.00 frames. ], tot_loss[loss=0.2442, simple_loss=0.32, pruned_loss=0.0842, over 3097134.81 frames. ], batch size: 246, lr: 1.11e-02, grad_scale: 4.0 2023-04-28 13:24:33,652 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-28 13:24:42,142 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 13:24:52,001 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.098e+02 3.672e+02 4.554e+02 5.723e+02 8.873e+02, threshold=9.107e+02, percent-clipped=1.0 2023-04-28 13:25:23,665 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58497.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:28,187 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=58500.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:25:29,696 INFO [train.py:904] (0/8) Epoch 6, batch 7750, loss[loss=0.213, simple_loss=0.2949, pruned_loss=0.06557, over 17023.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3202, pruned_loss=0.08425, over 3094932.80 frames. ], batch size: 55, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:26:13,837 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2491, 3.8348, 3.8789, 1.5790, 4.0368, 4.1639, 2.9728, 2.8953], device='cuda:0'), covar=tensor([0.0842, 0.0109, 0.0119, 0.1266, 0.0067, 0.0053, 0.0348, 0.0423], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0091, 0.0082, 0.0140, 0.0072, 0.0081, 0.0116, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 13:26:14,156 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 13:26:46,081 INFO [train.py:904] (0/8) Epoch 6, batch 7800, loss[loss=0.2433, simple_loss=0.3268, pruned_loss=0.0799, over 16400.00 frames. ], tot_loss[loss=0.2467, simple_loss=0.3221, pruned_loss=0.08567, over 3097214.15 frames. ], batch size: 68, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:27:26,212 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.190e+02 3.955e+02 4.761e+02 5.914e+02 1.211e+03, threshold=9.523e+02, percent-clipped=2.0 2023-04-28 13:28:01,033 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5435, 3.6631, 1.8346, 3.9832, 2.4144, 3.8769, 1.9945, 2.6036], device='cuda:0'), covar=tensor([0.0151, 0.0275, 0.1640, 0.0061, 0.0838, 0.0424, 0.1487, 0.0736], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0154, 0.0179, 0.0087, 0.0163, 0.0190, 0.0190, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 13:28:01,707 INFO [train.py:904] (0/8) Epoch 6, batch 7850, loss[loss=0.2201, simple_loss=0.3095, pruned_loss=0.06531, over 16834.00 frames. ], tot_loss[loss=0.2475, simple_loss=0.3234, pruned_loss=0.08575, over 3096889.17 frames. ], batch size: 102, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:28:04,171 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4109, 5.8328, 5.5054, 5.6071, 5.0336, 4.9315, 5.2704, 5.8946], device='cuda:0'), covar=tensor([0.0627, 0.0550, 0.0867, 0.0471, 0.0701, 0.0651, 0.0652, 0.0622], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0519, 0.0447, 0.0339, 0.0321, 0.0345, 0.0426, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:28:50,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0197, 3.1236, 1.5904, 3.2688, 2.1753, 3.2218, 1.7795, 2.4108], device='cuda:0'), covar=tensor([0.0188, 0.0330, 0.1510, 0.0083, 0.0771, 0.0453, 0.1434, 0.0697], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0155, 0.0180, 0.0088, 0.0164, 0.0192, 0.0192, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 13:29:14,711 INFO [train.py:904] (0/8) Epoch 6, batch 7900, loss[loss=0.2309, simple_loss=0.3168, pruned_loss=0.07246, over 16571.00 frames. ], tot_loss[loss=0.2462, simple_loss=0.3226, pruned_loss=0.08489, over 3110159.45 frames. ], batch size: 68, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:29:53,308 INFO [optim.py:368] (0/8) 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 13:30:02,345 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58681.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:30:31,235 INFO [train.py:904] (0/8) Epoch 6, batch 7950, loss[loss=0.286, simple_loss=0.3472, pruned_loss=0.1124, over 11899.00 frames. ], tot_loss[loss=0.2465, simple_loss=0.3226, pruned_loss=0.08517, over 3102973.17 frames. ], batch size: 248, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:31:33,551 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58742.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:31:46,446 INFO [train.py:904] (0/8) Epoch 6, batch 8000, loss[loss=0.2151, simple_loss=0.2978, pruned_loss=0.0662, over 16251.00 frames. ], tot_loss[loss=0.2468, simple_loss=0.3228, pruned_loss=0.0854, over 3094864.07 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:32:24,377 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 13:32:57,291 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:02,700 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=58800.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:33:03,408 INFO [train.py:904] (0/8) Epoch 6, batch 8050, loss[loss=0.2447, simple_loss=0.3179, pruned_loss=0.08576, over 16675.00 frames. ], tot_loss[loss=0.2461, simple_loss=0.3223, pruned_loss=0.0849, over 3093188.42 frames. ], batch size: 57, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:33:58,005 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58836.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:12,441 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58845.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:16,879 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=58848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:34:20,759 INFO [train.py:904] (0/8) Epoch 6, batch 8100, loss[loss=0.2744, simple_loss=0.3414, pruned_loss=0.1037, over 15248.00 frames. ], tot_loss[loss=0.2443, simple_loss=0.3212, pruned_loss=0.0837, over 3103883.67 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:03,721 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.302e+02 3.498e+02 4.445e+02 5.357e+02 1.535e+03, threshold=8.891e+02, percent-clipped=10.0 2023-04-28 13:35:38,308 INFO [train.py:904] (0/8) Epoch 6, batch 8150, loss[loss=0.2156, simple_loss=0.2967, pruned_loss=0.06724, over 16259.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3197, pruned_loss=0.08354, over 3097284.41 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:35:46,338 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=58906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:36:28,127 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8312, 1.6689, 1.4907, 1.4812, 1.7846, 1.6686, 1.6184, 1.9168], device='cuda:0'), covar=tensor([0.0056, 0.0141, 0.0203, 0.0208, 0.0106, 0.0154, 0.0104, 0.0119], device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0162, 0.0166, 0.0164, 0.0160, 0.0167, 0.0150, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:36:54,145 INFO [train.py:904] (0/8) Epoch 6, batch 8200, loss[loss=0.2213, simple_loss=0.3027, pruned_loss=0.06998, over 16286.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3165, pruned_loss=0.08248, over 3098802.64 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:37:21,742 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=58967.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:37:40,740 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.295e+02 3.597e+02 4.719e+02 5.797e+02 1.131e+03, threshold=9.438e+02, percent-clipped=6.0 2023-04-28 13:38:17,125 INFO [train.py:904] (0/8) Epoch 6, batch 8250, loss[loss=0.2148, simple_loss=0.3003, pruned_loss=0.0646, over 15336.00 frames. ], tot_loss[loss=0.2379, simple_loss=0.3153, pruned_loss=0.08027, over 3068423.91 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:38:24,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7837, 1.6487, 2.2760, 2.8832, 2.5542, 3.0555, 1.8552, 2.8131], device='cuda:0'), covar=tensor([0.0076, 0.0256, 0.0143, 0.0102, 0.0128, 0.0079, 0.0238, 0.0075], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0148, 0.0129, 0.0127, 0.0135, 0.0099, 0.0145, 0.0084], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 13:38:26,235 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-28 13:38:46,156 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8871, 3.9376, 2.1444, 4.4391, 2.6602, 4.2989, 2.3083, 3.0404], device='cuda:0'), covar=tensor([0.0121, 0.0188, 0.1229, 0.0042, 0.0656, 0.0345, 0.1322, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0151, 0.0175, 0.0084, 0.0158, 0.0186, 0.0188, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 13:39:16,136 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:39:37,614 INFO [train.py:904] (0/8) Epoch 6, batch 8300, loss[loss=0.2227, simple_loss=0.3094, pruned_loss=0.068, over 16772.00 frames. ], tot_loss[loss=0.2326, simple_loss=0.3121, pruned_loss=0.07659, over 3071191.70 frames. ], batch size: 124, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:40:22,347 INFO [optim.py:368] (0/8) 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 13:40:59,224 INFO [train.py:904] (0/8) Epoch 6, batch 8350, loss[loss=0.2183, simple_loss=0.3088, pruned_loss=0.06393, over 15164.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3106, pruned_loss=0.07386, over 3072653.56 frames. ], batch size: 190, lr: 1.10e-02, grad_scale: 2.0 2023-04-28 13:41:38,094 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-28 13:41:49,267 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59131.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:42:21,180 INFO [train.py:904] (0/8) Epoch 6, batch 8400, loss[loss=0.2257, simple_loss=0.3187, pruned_loss=0.06635, over 16784.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3074, pruned_loss=0.07141, over 3064075.31 frames. ], batch size: 124, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:42:40,777 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3947, 3.0078, 2.6692, 2.2626, 2.2577, 2.1867, 2.9049, 2.9796], device='cuda:0'), covar=tensor([0.1953, 0.0603, 0.1062, 0.1492, 0.1947, 0.1568, 0.0326, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0236, 0.0260, 0.0243, 0.0263, 0.0196, 0.0239, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:42:42,356 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:43:05,301 INFO [optim.py:368] (0/8) 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 13:43:29,376 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-28 13:43:40,735 INFO [train.py:904] (0/8) Epoch 6, batch 8450, loss[loss=0.1862, simple_loss=0.2804, pruned_loss=0.04595, over 16428.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3045, pruned_loss=0.06874, over 3069164.14 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:44:01,282 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1541, 2.9206, 2.7167, 1.9405, 2.5705, 2.1212, 2.7305, 2.8283], device='cuda:0'), covar=tensor([0.0337, 0.0581, 0.0453, 0.1460, 0.0723, 0.0958, 0.0639, 0.0739], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0125, 0.0149, 0.0138, 0.0130, 0.0122, 0.0135, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 13:44:19,005 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:44:19,429 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-28 13:45:00,577 INFO [train.py:904] (0/8) Epoch 6, batch 8500, loss[loss=0.2004, simple_loss=0.2887, pruned_loss=0.05603, over 16360.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.3001, pruned_loss=0.06562, over 3066642.73 frames. ], batch size: 146, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:45:19,668 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.033e+02 2.693e+02 3.373e+02 4.229e+02 1.183e+03, threshold=6.745e+02, percent-clipped=1.0 2023-04-28 13:46:25,538 INFO [train.py:904] (0/8) Epoch 6, batch 8550, loss[loss=0.2077, simple_loss=0.297, pruned_loss=0.05915, over 15399.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2976, pruned_loss=0.06397, over 3076915.89 frames. ], batch size: 191, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:47:38,309 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:47:41,056 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2715, 1.8752, 2.0838, 3.6417, 1.8391, 2.4922, 2.1144, 1.9969], device='cuda:0'), covar=tensor([0.0579, 0.2546, 0.1433, 0.0341, 0.3388, 0.1466, 0.2345, 0.2750], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0326, 0.0271, 0.0298, 0.0376, 0.0340, 0.0298, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:48:07,970 INFO [train.py:904] (0/8) Epoch 6, batch 8600, loss[loss=0.2016, simple_loss=0.2919, pruned_loss=0.05564, over 17050.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2978, pruned_loss=0.06316, over 3056691.31 frames. ], batch size: 53, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:48:12,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9532, 1.9850, 2.2312, 3.2502, 1.9863, 2.3908, 2.2026, 1.9316], device='cuda:0'), covar=tensor([0.0645, 0.2373, 0.1240, 0.0384, 0.3081, 0.1457, 0.2073, 0.2887], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0327, 0.0272, 0.0299, 0.0376, 0.0340, 0.0298, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:48:20,532 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8677, 3.9766, 1.8357, 4.3139, 2.4702, 4.1929, 1.9301, 2.7462], device='cuda:0'), covar=tensor([0.0124, 0.0214, 0.1663, 0.0047, 0.0837, 0.0299, 0.1666, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0146, 0.0174, 0.0081, 0.0155, 0.0179, 0.0185, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-28 13:49:03,383 INFO [optim.py:368] (0/8) 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 13:49:16,115 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59385.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:49:46,457 INFO [train.py:904] (0/8) Epoch 6, batch 8650, loss[loss=0.1927, simple_loss=0.2746, pruned_loss=0.05543, over 12037.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2955, pruned_loss=0.06119, over 3067194.48 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:50:16,207 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4303, 1.9165, 1.9653, 3.9573, 1.7901, 2.5353, 2.0554, 2.0448], device='cuda:0'), covar=tensor([0.0606, 0.2689, 0.1629, 0.0286, 0.3495, 0.1482, 0.2544, 0.2789], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0326, 0.0272, 0.0298, 0.0375, 0.0338, 0.0298, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:50:55,079 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59431.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:50:55,413 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 13:51:31,169 INFO [train.py:904] (0/8) Epoch 6, batch 8700, loss[loss=0.2207, simple_loss=0.2989, pruned_loss=0.07127, over 12317.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2925, pruned_loss=0.05968, over 3078120.34 frames. ], batch size: 247, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:51:59,966 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2215, 3.3561, 3.4076, 1.6862, 3.5530, 3.6171, 2.7538, 2.7557], device='cuda:0'), covar=tensor([0.0702, 0.0134, 0.0135, 0.1151, 0.0063, 0.0064, 0.0340, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0087, 0.0078, 0.0136, 0.0066, 0.0078, 0.0112, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 13:52:01,582 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3432, 4.6364, 4.3969, 4.3916, 4.0818, 4.0611, 4.2361, 4.6553], device='cuda:0'), covar=tensor([0.0807, 0.0760, 0.0963, 0.0528, 0.0660, 0.1190, 0.0784, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0385, 0.0505, 0.0427, 0.0328, 0.0310, 0.0336, 0.0415, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:52:21,221 INFO [optim.py:368] (0/8) 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 13:52:22,283 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59479.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:05,581 INFO [train.py:904] (0/8) Epoch 6, batch 8750, loss[loss=0.2366, simple_loss=0.3258, pruned_loss=0.07373, over 16246.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2926, pruned_loss=0.05921, over 3094263.72 frames. ], batch size: 165, lr: 1.10e-02, grad_scale: 4.0 2023-04-28 13:53:08,776 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59502.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:53:54,451 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59520.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:54:55,925 INFO [train.py:904] (0/8) Epoch 6, batch 8800, loss[loss=0.1886, simple_loss=0.2862, pruned_loss=0.04555, over 16667.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2911, pruned_loss=0.05801, over 3096477.91 frames. ], batch size: 89, lr: 1.10e-02, grad_scale: 8.0 2023-04-28 13:55:17,778 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59562.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:19,840 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59563.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:55:41,479 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-28 13:55:52,198 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.676e+02 2.649e+02 3.204e+02 3.980e+02 7.645e+02, threshold=6.408e+02, percent-clipped=1.0 2023-04-28 13:55:53,284 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5091, 2.6626, 2.2356, 3.9781, 3.0206, 3.9433, 1.1034, 2.6855], device='cuda:0'), covar=tensor([0.1458, 0.0628, 0.1217, 0.0100, 0.0159, 0.0351, 0.1698, 0.0839], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0140, 0.0165, 0.0092, 0.0173, 0.0185, 0.0163, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 13:56:37,985 INFO [train.py:904] (0/8) Epoch 6, batch 8850, loss[loss=0.2181, simple_loss=0.3085, pruned_loss=0.06384, over 16894.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2937, pruned_loss=0.05766, over 3083939.13 frames. ], batch size: 116, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:56:56,206 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59610.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 13:57:42,916 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4583, 1.9741, 1.6668, 1.6641, 2.2184, 2.0483, 2.2854, 2.3654], device='cuda:0'), covar=tensor([0.0033, 0.0205, 0.0229, 0.0225, 0.0111, 0.0169, 0.0093, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0082, 0.0161, 0.0161, 0.0160, 0.0156, 0.0161, 0.0139, 0.0141], device='cuda:0'), out_proj_covar=tensor([9.5169e-05, 1.8760e-04, 1.8217e-04, 1.8241e-04, 1.8171e-04, 1.8664e-04, 1.5545e-04, 1.6276e-04], device='cuda:0') 2023-04-28 13:57:56,244 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2714, 1.8732, 1.9686, 3.7558, 1.8739, 2.4771, 2.0856, 1.9636], device='cuda:0'), covar=tensor([0.0655, 0.2735, 0.1622, 0.0326, 0.3338, 0.1583, 0.2415, 0.3035], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0329, 0.0275, 0.0301, 0.0382, 0.0342, 0.0301, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 13:58:21,082 INFO [train.py:904] (0/8) Epoch 6, batch 8900, loss[loss=0.2114, simple_loss=0.3, pruned_loss=0.06138, over 16753.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2935, pruned_loss=0.05669, over 3084631.09 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 13:59:22,392 INFO [optim.py:368] (0/8) 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 14:00:23,145 INFO [train.py:904] (0/8) Epoch 6, batch 8950, loss[loss=0.1887, simple_loss=0.2934, pruned_loss=0.04195, over 16573.00 frames. ], tot_loss[loss=0.203, simple_loss=0.293, pruned_loss=0.05646, over 3095277.41 frames. ], batch size: 68, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:00:24,682 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 14:02:12,310 INFO [train.py:904] (0/8) Epoch 6, batch 9000, loss[loss=0.1666, simple_loss=0.2574, pruned_loss=0.03788, over 16808.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2892, pruned_loss=0.05482, over 3086725.99 frames. ], batch size: 83, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:02:12,312 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 14:02:22,191 INFO [train.py:938] (0/8) Epoch 6, validation: loss=0.1682, simple_loss=0.2716, pruned_loss=0.03235, over 944034.00 frames. 2023-04-28 14:02:22,193 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 14:03:21,585 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.787e+02 3.475e+02 4.351e+02 1.064e+03, threshold=6.950e+02, percent-clipped=4.0 2023-04-28 14:03:43,414 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4634, 4.2219, 4.4291, 4.6105, 4.7279, 4.2344, 4.7456, 4.7391], device='cuda:0'), covar=tensor([0.1038, 0.0764, 0.1050, 0.0485, 0.0479, 0.0763, 0.0374, 0.0414], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0466, 0.0588, 0.0477, 0.0364, 0.0358, 0.0375, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 14:03:43,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3883, 3.7548, 3.8508, 1.5690, 4.0893, 4.0519, 3.0160, 3.0085], device='cuda:0'), covar=tensor([0.0770, 0.0122, 0.0146, 0.1303, 0.0040, 0.0079, 0.0310, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0087, 0.0078, 0.0137, 0.0066, 0.0079, 0.0112, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 14:04:06,174 INFO [train.py:904] (0/8) Epoch 6, batch 9050, loss[loss=0.2158, simple_loss=0.2969, pruned_loss=0.06738, over 12935.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2893, pruned_loss=0.05556, over 3075378.82 frames. ], batch size: 250, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:04:44,579 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=59820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:45,895 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:05:50,589 INFO [train.py:904] (0/8) Epoch 6, batch 9100, loss[loss=0.1968, simple_loss=0.2936, pruned_loss=0.05005, over 16662.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2892, pruned_loss=0.05601, over 3087461.72 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:06:03,679 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=59858.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:24,267 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=59868.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:06:51,201 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 14:06:54,446 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0123, 3.5114, 3.4735, 2.1045, 3.3492, 3.5203, 3.4308, 1.8011], device='cuda:0'), covar=tensor([0.0384, 0.0022, 0.0041, 0.0314, 0.0047, 0.0046, 0.0044, 0.0377], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0053, 0.0057, 0.0114, 0.0060, 0.0070, 0.0062, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 14:06:55,734 INFO [optim.py:368] (0/8) 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 14:07:15,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5753, 3.7868, 4.0523, 1.8809, 4.2885, 4.2563, 3.1640, 3.1438], device='cuda:0'), covar=tensor([0.0610, 0.0122, 0.0104, 0.1002, 0.0031, 0.0047, 0.0256, 0.0319], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0086, 0.0076, 0.0136, 0.0065, 0.0078, 0.0111, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 14:07:19,857 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=59889.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:07:42,764 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1327, 4.1623, 4.2556, 4.2539, 4.2839, 4.7765, 4.4616, 4.1638], device='cuda:0'), covar=tensor([0.1313, 0.1474, 0.1279, 0.1881, 0.2386, 0.1019, 0.0949, 0.2254], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0385, 0.0392, 0.0334, 0.0442, 0.0424, 0.0319, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 14:07:47,008 INFO [train.py:904] (0/8) Epoch 6, batch 9150, loss[loss=0.1876, simple_loss=0.2806, pruned_loss=0.04729, over 16313.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2902, pruned_loss=0.0564, over 3073426.62 frames. ], batch size: 166, lr: 1.09e-02, grad_scale: 4.0 2023-04-28 14:08:06,768 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:09:29,916 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=59950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:09:31,024 INFO [train.py:904] (0/8) Epoch 6, batch 9200, loss[loss=0.1988, simple_loss=0.2878, pruned_loss=0.05488, over 16344.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2853, pruned_loss=0.05509, over 3079076.29 frames. ], batch size: 146, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:10:10,183 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3701, 4.3563, 4.8341, 4.8115, 4.7622, 4.4344, 4.4620, 4.2104], device='cuda:0'), covar=tensor([0.0224, 0.0380, 0.0268, 0.0331, 0.0382, 0.0326, 0.0708, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0237, 0.0246, 0.0236, 0.0285, 0.0260, 0.0350, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-28 14:10:22,282 INFO [optim.py:368] (0/8) 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 14:10:25,491 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7308, 3.8925, 3.9973, 1.8058, 4.2891, 4.2934, 3.1960, 3.1262], device='cuda:0'), covar=tensor([0.0646, 0.0151, 0.0158, 0.1172, 0.0034, 0.0043, 0.0298, 0.0396], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0087, 0.0077, 0.0135, 0.0065, 0.0077, 0.0112, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 14:11:04,751 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-60000.pt 2023-04-28 14:11:09,068 INFO [train.py:904] (0/8) Epoch 6, batch 9250, loss[loss=0.1887, simple_loss=0.2819, pruned_loss=0.04778, over 15346.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2846, pruned_loss=0.05495, over 3041543.34 frames. ], batch size: 191, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:12:34,007 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60040.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:12:58,490 INFO [train.py:904] (0/8) Epoch 6, batch 9300, loss[loss=0.1895, simple_loss=0.2758, pruned_loss=0.05157, over 16396.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2829, pruned_loss=0.05369, over 3058205.95 frames. ], batch size: 146, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:13:01,053 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60052.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:33,153 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60065.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:13:35,602 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0914, 3.5179, 3.4111, 2.4605, 3.2249, 3.3129, 3.2747, 1.7746], device='cuda:0'), covar=tensor([0.0323, 0.0016, 0.0028, 0.0232, 0.0051, 0.0057, 0.0040, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0053, 0.0056, 0.0114, 0.0061, 0.0071, 0.0062, 0.0109], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 14:14:01,616 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:14:05,092 INFO [optim.py:368] (0/8) 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 14:14:44,863 INFO [train.py:904] (0/8) Epoch 6, batch 9350, loss[loss=0.2084, simple_loss=0.2937, pruned_loss=0.06156, over 15512.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2834, pruned_loss=0.05425, over 3052596.08 frames. ], batch size: 192, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:14:45,757 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60101.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 14:15:12,254 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60113.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 14:15:36,689 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:16:03,914 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60139.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:16:25,957 INFO [train.py:904] (0/8) Epoch 6, batch 9400, loss[loss=0.1867, simple_loss=0.2895, pruned_loss=0.04193, over 16171.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2835, pruned_loss=0.05348, over 3073221.31 frames. ], batch size: 165, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:16:40,806 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60158.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:16:53,299 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 14:16:55,821 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.63 vs. limit=5.0 2023-04-28 14:17:25,100 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 3.031e+02 3.474e+02 4.387e+02 1.018e+03, threshold=6.948e+02, percent-clipped=2.0 2023-04-28 14:17:51,753 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7458, 2.7203, 1.5838, 2.8439, 2.0581, 2.7863, 1.7355, 2.3733], device='cuda:0'), covar=tensor([0.0220, 0.0443, 0.1554, 0.0136, 0.0721, 0.0543, 0.1598, 0.0571], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0147, 0.0174, 0.0082, 0.0154, 0.0178, 0.0186, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 14:18:05,560 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:09,620 INFO [train.py:904] (0/8) Epoch 6, batch 9450, loss[loss=0.1891, simple_loss=0.2824, pruned_loss=0.04792, over 16684.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2859, pruned_loss=0.05364, over 3082003.35 frames. ], batch size: 134, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:18:14,584 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60206.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:18:29,104 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 14:19:39,965 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:19:50,597 INFO [train.py:904] (0/8) Epoch 6, batch 9500, loss[loss=0.2019, simple_loss=0.2909, pruned_loss=0.05648, over 16364.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.284, pruned_loss=0.05265, over 3095006.91 frames. ], batch size: 146, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:19:59,579 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9112, 3.0029, 3.0201, 2.0837, 2.8854, 2.9649, 3.0189, 1.6946], device='cuda:0'), covar=tensor([0.0352, 0.0022, 0.0041, 0.0276, 0.0058, 0.0059, 0.0044, 0.0360], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0053, 0.0056, 0.0113, 0.0061, 0.0071, 0.0062, 0.0110], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 14:20:11,547 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60259.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:20:18,273 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 14:20:21,134 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7903, 5.1766, 4.9079, 4.9286, 4.6135, 4.5892, 4.6519, 5.2146], device='cuda:0'), covar=tensor([0.0723, 0.0754, 0.0944, 0.0536, 0.0613, 0.0730, 0.0806, 0.0724], device='cuda:0'), in_proj_covar=tensor([0.0389, 0.0507, 0.0419, 0.0330, 0.0311, 0.0337, 0.0416, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 14:20:51,249 INFO [optim.py:368] (0/8) 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 14:21:39,653 INFO [train.py:904] (0/8) Epoch 6, batch 9550, loss[loss=0.2259, simple_loss=0.3147, pruned_loss=0.06858, over 16146.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2827, pruned_loss=0.05266, over 3070047.57 frames. ], batch size: 165, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:22:27,426 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-04-28 14:23:21,510 INFO [train.py:904] (0/8) Epoch 6, batch 9600, loss[loss=0.2139, simple_loss=0.2865, pruned_loss=0.07059, over 12164.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2841, pruned_loss=0.05381, over 3059400.66 frames. ], batch size: 247, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:24:17,216 INFO [optim.py:368] (0/8) 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 14:24:56,965 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60396.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 14:25:09,710 INFO [train.py:904] (0/8) Epoch 6, batch 9650, loss[loss=0.2075, simple_loss=0.2926, pruned_loss=0.06122, over 16880.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2865, pruned_loss=0.05456, over 3061631.15 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:25:28,086 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60408.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 14:25:57,601 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60421.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:22,101 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60434.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:26:58,010 INFO [train.py:904] (0/8) Epoch 6, batch 9700, loss[loss=0.2016, simple_loss=0.2801, pruned_loss=0.06153, over 12515.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2854, pruned_loss=0.05434, over 3052226.81 frames. ], batch size: 250, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:27:34,390 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5543, 1.9211, 1.6378, 1.6036, 2.3284, 2.0105, 2.4044, 2.4480], device='cuda:0'), covar=tensor([0.0037, 0.0230, 0.0278, 0.0299, 0.0133, 0.0221, 0.0105, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0081, 0.0161, 0.0162, 0.0161, 0.0158, 0.0161, 0.0139, 0.0142], device='cuda:0'), out_proj_covar=tensor([9.2792e-05, 1.8692e-04, 1.8195e-04, 1.8278e-04, 1.8328e-04, 1.8570e-04, 1.5400e-04, 1.6369e-04], device='cuda:0') 2023-04-28 14:27:59,429 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 6, batch 9750, loss[loss=0.1953, simple_loss=0.2867, pruned_loss=0.05196, over 16434.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2836, pruned_loss=0.05403, over 3061297.01 frames. ], batch size: 146, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:28:47,812 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60504.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:29:10,971 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-28 14:30:11,580 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:19,577 INFO [train.py:904] (0/8) Epoch 6, batch 9800, loss[loss=0.2169, simple_loss=0.3092, pruned_loss=0.06231, over 16751.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2839, pruned_loss=0.05316, over 3072355.09 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:30:21,864 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60552.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:30:21,970 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9547, 3.7410, 3.9679, 4.1122, 4.2349, 3.7691, 4.2146, 4.2037], device='cuda:0'), covar=tensor([0.0892, 0.0914, 0.1069, 0.0500, 0.0433, 0.1200, 0.0446, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0477, 0.0592, 0.0481, 0.0366, 0.0361, 0.0380, 0.0410], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 14:30:27,287 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=60554.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:31:14,373 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.779e+02 2.802e+02 3.388e+02 4.163e+02 9.547e+02, threshold=6.775e+02, percent-clipped=1.0 2023-04-28 14:31:48,143 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60593.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:32:03,905 INFO [train.py:904] (0/8) Epoch 6, batch 9850, loss[loss=0.2141, simple_loss=0.3112, pruned_loss=0.05849, over 16858.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2855, pruned_loss=0.05296, over 3075876.30 frames. ], batch size: 124, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:33:56,852 INFO [train.py:904] (0/8) Epoch 6, batch 9900, loss[loss=0.2054, simple_loss=0.2833, pruned_loss=0.06373, over 12230.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2864, pruned_loss=0.05324, over 3057602.86 frames. ], batch size: 246, lr: 1.09e-02, grad_scale: 8.0 2023-04-28 14:35:04,553 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.757e+02 3.371e+02 4.013e+02 6.843e+02, threshold=6.741e+02, percent-clipped=1.0 2023-04-28 14:35:42,402 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:35:52,850 INFO [train.py:904] (0/8) Epoch 6, batch 9950, loss[loss=0.1958, simple_loss=0.2883, pruned_loss=0.05169, over 16575.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2873, pruned_loss=0.05349, over 3029592.73 frames. ], batch size: 134, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:36:09,272 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60708.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:34,721 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=60718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:36:42,416 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60721.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:13,198 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:37,956 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60744.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:37:44,065 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5725, 3.5439, 3.4603, 3.0293, 3.4433, 1.9457, 3.2137, 3.0072], device='cuda:0'), covar=tensor([0.0095, 0.0082, 0.0115, 0.0160, 0.0074, 0.1875, 0.0100, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0095, 0.0080, 0.0126, 0.0114, 0.0094, 0.0150, 0.0110, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 14:37:50,332 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9321, 1.7668, 2.1602, 3.2343, 2.8793, 3.3566, 2.0288, 3.2187], device='cuda:0'), covar=tensor([0.0122, 0.0266, 0.0194, 0.0084, 0.0115, 0.0079, 0.0243, 0.0058], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0148, 0.0131, 0.0129, 0.0135, 0.0095, 0.0144, 0.0081], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 14:37:54,654 INFO [train.py:904] (0/8) Epoch 6, batch 10000, loss[loss=0.2031, simple_loss=0.2783, pruned_loss=0.06394, over 12872.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2857, pruned_loss=0.05317, over 3053128.32 frames. ], batch size: 248, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:38:06,063 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60756.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:15,890 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1470, 4.1076, 4.5464, 4.4932, 4.5376, 4.1745, 4.2439, 4.0857], device='cuda:0'), covar=tensor([0.0232, 0.0409, 0.0401, 0.0439, 0.0319, 0.0304, 0.0597, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0241, 0.0247, 0.0238, 0.0281, 0.0262, 0.0349, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-28 14:38:29,854 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60769.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:50,593 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=60779.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:38:51,225 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 2.599e+02 3.122e+02 4.143e+02 8.746e+02, threshold=6.243e+02, percent-clipped=4.0 2023-04-28 14:38:56,776 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60782.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:39:33,422 INFO [train.py:904] (0/8) Epoch 6, batch 10050, loss[loss=0.1999, simple_loss=0.2858, pruned_loss=0.05698, over 12200.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2863, pruned_loss=0.05337, over 3054884.61 frames. ], batch size: 248, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:40:19,640 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 14:40:55,536 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-28 14:41:05,274 INFO [train.py:904] (0/8) Epoch 6, batch 10100, loss[loss=0.1995, simple_loss=0.2849, pruned_loss=0.05707, over 15269.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2868, pruned_loss=0.05354, over 3041701.62 frames. ], batch size: 191, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:41:11,015 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=60854.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:41:20,053 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7040, 3.3957, 3.2147, 1.8451, 2.8584, 2.1816, 3.1414, 3.1542], device='cuda:0'), covar=tensor([0.0258, 0.0440, 0.0413, 0.1511, 0.0623, 0.0910, 0.0668, 0.0731], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0118, 0.0150, 0.0138, 0.0130, 0.0123, 0.0133, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 14:42:01,270 INFO [optim.py:368] (0/8) 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 14:42:18,001 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-28 14:42:22,736 INFO [train.py:904] (0/8) Epoch 6, batch 10150, loss[loss=0.2164, simple_loss=0.2837, pruned_loss=0.07449, over 12114.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2859, pruned_loss=0.05378, over 3035597.95 frames. ], batch size: 246, lr: 1.08e-02, grad_scale: 8.0 2023-04-28 14:42:24,872 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-6.pt 2023-04-28 14:42:48,319 INFO [train.py:904] (0/8) Epoch 7, batch 0, loss[loss=0.2127, simple_loss=0.2939, pruned_loss=0.06573, over 17227.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2939, pruned_loss=0.06573, over 17227.00 frames. ], batch size: 44, lr: 1.02e-02, grad_scale: 8.0 2023-04-28 14:42:48,320 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 14:42:55,783 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 14:42:55,996 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=60902.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:44:05,514 INFO [train.py:904] (0/8) Epoch 7, batch 50, loss[loss=0.2209, simple_loss=0.3041, pruned_loss=0.06881, over 16771.00 frames. ], tot_loss[loss=0.2359, simple_loss=0.3093, pruned_loss=0.08126, over 743673.18 frames. ], batch size: 57, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:44:49,463 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 3.058e+02 3.620e+02 4.703e+02 1.122e+03, threshold=7.241e+02, percent-clipped=6.0 2023-04-28 14:45:15,342 INFO [train.py:904] (0/8) Epoch 7, batch 100, loss[loss=0.2189, simple_loss=0.3029, pruned_loss=0.06742, over 16689.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3028, pruned_loss=0.07753, over 1316162.68 frames. ], batch size: 57, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:45:15,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6197, 4.0293, 4.2877, 3.0025, 3.9055, 4.0640, 4.0473, 2.6590], device='cuda:0'), covar=tensor([0.0295, 0.0042, 0.0022, 0.0219, 0.0040, 0.0048, 0.0034, 0.0250], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0057, 0.0059, 0.0117, 0.0062, 0.0073, 0.0065, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 14:45:23,157 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.12 vs. limit=5.0 2023-04-28 14:46:07,938 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2979, 5.1510, 5.0351, 4.7955, 4.5961, 5.0027, 5.1344, 4.6314], device='cuda:0'), covar=tensor([0.0503, 0.0290, 0.0260, 0.0221, 0.1046, 0.0303, 0.0239, 0.0535], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0215, 0.0227, 0.0201, 0.0256, 0.0232, 0.0161, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 14:46:24,666 INFO [train.py:904] (0/8) Epoch 7, batch 150, loss[loss=0.1812, simple_loss=0.2723, pruned_loss=0.04499, over 17250.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.2982, pruned_loss=0.07377, over 1763307.56 frames. ], batch size: 45, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:46:37,286 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61061.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:46:56,386 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61074.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:47:07,166 INFO [optim.py:368] (0/8) 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 14:47:34,207 INFO [train.py:904] (0/8) Epoch 7, batch 200, loss[loss=0.189, simple_loss=0.2777, pruned_loss=0.05012, over 17117.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.2972, pruned_loss=0.07348, over 2096859.09 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:48:01,640 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:48:44,076 INFO [train.py:904] (0/8) Epoch 7, batch 250, loss[loss=0.242, simple_loss=0.3022, pruned_loss=0.09087, over 11892.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.2945, pruned_loss=0.07241, over 2369047.78 frames. ], batch size: 246, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:49:24,069 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0934, 4.7471, 5.0507, 5.2803, 5.4968, 4.8155, 5.3716, 5.4006], device='cuda:0'), covar=tensor([0.1212, 0.0941, 0.1393, 0.0550, 0.0419, 0.0579, 0.0460, 0.0410], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0525, 0.0659, 0.0525, 0.0399, 0.0396, 0.0420, 0.0451], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 14:49:24,842 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.754e+02 3.071e+02 3.589e+02 4.451e+02 1.172e+03, threshold=7.179e+02, percent-clipped=5.0 2023-04-28 14:49:52,018 INFO [train.py:904] (0/8) Epoch 7, batch 300, loss[loss=0.1654, simple_loss=0.2472, pruned_loss=0.04183, over 16978.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.2898, pruned_loss=0.06888, over 2579896.83 frames. ], batch size: 41, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:50:36,757 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8245, 2.7506, 3.0817, 2.0730, 2.8922, 2.9848, 3.0196, 1.7249], device='cuda:0'), covar=tensor([0.0349, 0.0080, 0.0042, 0.0245, 0.0057, 0.0080, 0.0049, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0060, 0.0060, 0.0115, 0.0062, 0.0075, 0.0065, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 14:50:59,339 INFO [train.py:904] (0/8) Epoch 7, batch 350, loss[loss=0.2087, simple_loss=0.2729, pruned_loss=0.07219, over 16787.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2867, pruned_loss=0.06759, over 2739431.34 frames. ], batch size: 83, lr: 1.01e-02, grad_scale: 1.0 2023-04-28 14:51:33,882 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 14:51:35,271 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9564, 1.7375, 2.3105, 2.7722, 2.5523, 3.2802, 2.2369, 3.0445], device='cuda:0'), covar=tensor([0.0114, 0.0255, 0.0192, 0.0183, 0.0149, 0.0115, 0.0235, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0150, 0.0134, 0.0133, 0.0137, 0.0100, 0.0146, 0.0084], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 14:51:41,730 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.762e+02 2.825e+02 3.324e+02 4.016e+02 8.512e+02, threshold=6.649e+02, percent-clipped=2.0 2023-04-28 14:52:07,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7775, 4.0425, 2.1172, 4.3548, 2.8555, 4.3930, 2.2821, 2.9860], device='cuda:0'), covar=tensor([0.0155, 0.0249, 0.1420, 0.0093, 0.0657, 0.0295, 0.1225, 0.0571], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0156, 0.0179, 0.0090, 0.0158, 0.0192, 0.0188, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 14:52:08,571 INFO [train.py:904] (0/8) Epoch 7, batch 400, loss[loss=0.2213, simple_loss=0.3028, pruned_loss=0.06993, over 16711.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2848, pruned_loss=0.06692, over 2869093.58 frames. ], batch size: 62, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:53:17,214 INFO [train.py:904] (0/8) Epoch 7, batch 450, loss[loss=0.2045, simple_loss=0.292, pruned_loss=0.05845, over 17214.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2833, pruned_loss=0.06607, over 2970028.72 frames. ], batch size: 45, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:53:48,664 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:54:00,441 INFO [optim.py:368] (0/8) 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 14:54:00,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4013, 3.8013, 3.4802, 2.1245, 2.9945, 2.3744, 3.7568, 3.6870], device='cuda:0'), covar=tensor([0.0189, 0.0482, 0.0526, 0.1401, 0.0634, 0.0910, 0.0429, 0.0689], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0126, 0.0152, 0.0139, 0.0131, 0.0123, 0.0135, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 14:54:27,758 INFO [train.py:904] (0/8) Epoch 7, batch 500, loss[loss=0.2286, simple_loss=0.2938, pruned_loss=0.08169, over 11918.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2817, pruned_loss=0.0654, over 3038444.65 frames. ], batch size: 246, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:54:47,712 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=61417.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:54:54,991 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 14:55:35,831 INFO [train.py:904] (0/8) Epoch 7, batch 550, loss[loss=0.2198, simple_loss=0.2817, pruned_loss=0.07898, over 16825.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.281, pruned_loss=0.06442, over 3105444.26 frames. ], batch size: 102, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:55:58,781 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-28 14:56:17,397 INFO [optim.py:368] (0/8) 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 14:56:43,957 INFO [train.py:904] (0/8) Epoch 7, batch 600, loss[loss=0.1835, simple_loss=0.2734, pruned_loss=0.04683, over 17129.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2814, pruned_loss=0.06527, over 3149338.56 frames. ], batch size: 48, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:56:51,373 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7291, 4.7122, 4.6129, 4.4545, 4.1971, 4.6675, 4.5847, 4.3338], device='cuda:0'), covar=tensor([0.0582, 0.0444, 0.0291, 0.0247, 0.0906, 0.0455, 0.0489, 0.0619], device='cuda:0'), in_proj_covar=tensor([0.0208, 0.0232, 0.0244, 0.0217, 0.0282, 0.0249, 0.0173, 0.0280], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 14:57:03,802 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2921, 4.3032, 4.8210, 4.8049, 4.7861, 4.4035, 4.4594, 4.2945], device='cuda:0'), covar=tensor([0.0306, 0.0486, 0.0316, 0.0354, 0.0390, 0.0322, 0.0829, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0278, 0.0274, 0.0279, 0.0265, 0.0323, 0.0293, 0.0396, 0.0241], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 14:57:53,638 INFO [train.py:904] (0/8) Epoch 7, batch 650, loss[loss=0.1775, simple_loss=0.2565, pruned_loss=0.04928, over 16501.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2797, pruned_loss=0.06416, over 3191901.37 frames. ], batch size: 68, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:58:35,765 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.778e+02 2.890e+02 3.388e+02 4.314e+02 1.492e+03, threshold=6.776e+02, percent-clipped=7.0 2023-04-28 14:59:01,945 INFO [train.py:904] (0/8) Epoch 7, batch 700, loss[loss=0.2333, simple_loss=0.2941, pruned_loss=0.0862, over 16921.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2789, pruned_loss=0.06323, over 3230854.47 frames. ], batch size: 109, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 14:59:54,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1370, 1.3438, 1.8035, 2.1318, 2.1672, 2.2105, 1.4907, 2.1536], device='cuda:0'), covar=tensor([0.0114, 0.0254, 0.0136, 0.0140, 0.0120, 0.0104, 0.0223, 0.0064], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0152, 0.0135, 0.0136, 0.0139, 0.0101, 0.0146, 0.0086], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 15:00:08,133 INFO [train.py:904] (0/8) Epoch 7, batch 750, loss[loss=0.1922, simple_loss=0.2796, pruned_loss=0.05243, over 17116.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.279, pruned_loss=0.06276, over 3252273.88 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 2.0 2023-04-28 15:00:51,557 INFO [optim.py:368] (0/8) 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 15:01:16,778 INFO [train.py:904] (0/8) Epoch 7, batch 800, loss[loss=0.2267, simple_loss=0.2906, pruned_loss=0.08141, over 16471.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2794, pruned_loss=0.06306, over 3262688.32 frames. ], batch size: 146, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:01:37,380 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=61717.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:01:39,765 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1689, 4.1028, 4.0449, 3.9135, 3.7564, 4.1056, 3.7540, 3.7798], device='cuda:0'), covar=tensor([0.0480, 0.0391, 0.0219, 0.0194, 0.0703, 0.0318, 0.0814, 0.0547], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0240, 0.0249, 0.0221, 0.0288, 0.0255, 0.0177, 0.0287], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:01:57,043 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:01:59,346 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0382, 4.9941, 4.7528, 4.2074, 4.8451, 1.7568, 4.6158, 4.8442], device='cuda:0'), covar=tensor([0.0063, 0.0047, 0.0124, 0.0293, 0.0066, 0.2008, 0.0098, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0090, 0.0143, 0.0134, 0.0107, 0.0158, 0.0124, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:02:11,452 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7674, 4.0117, 4.3989, 3.2016, 3.8518, 4.1896, 3.9123, 2.8242], device='cuda:0'), covar=tensor([0.0257, 0.0035, 0.0019, 0.0188, 0.0045, 0.0052, 0.0039, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0061, 0.0061, 0.0116, 0.0064, 0.0077, 0.0066, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:02:12,633 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61743.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:02:24,048 INFO [train.py:904] (0/8) Epoch 7, batch 850, loss[loss=0.2098, simple_loss=0.2782, pruned_loss=0.07067, over 16709.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2784, pruned_loss=0.06241, over 3282176.88 frames. ], batch size: 134, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:02:42,575 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=61765.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:06,806 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.892e+02 2.876e+02 3.503e+02 4.271e+02 9.516e+02, threshold=7.005e+02, percent-clipped=8.0 2023-04-28 15:03:19,421 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:03:33,689 INFO [train.py:904] (0/8) Epoch 7, batch 900, loss[loss=0.1843, simple_loss=0.2729, pruned_loss=0.04781, over 17109.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.277, pruned_loss=0.06104, over 3293240.96 frames. ], batch size: 49, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:03:36,361 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61804.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:04:12,644 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2115, 1.9751, 1.4502, 1.7236, 2.3654, 2.1614, 2.3833, 2.4916], device='cuda:0'), covar=tensor([0.0094, 0.0234, 0.0325, 0.0296, 0.0122, 0.0212, 0.0160, 0.0154], device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0174, 0.0173, 0.0172, 0.0168, 0.0175, 0.0163, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:04:40,620 INFO [train.py:904] (0/8) Epoch 7, batch 950, loss[loss=0.1868, simple_loss=0.2784, pruned_loss=0.04759, over 17045.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2766, pruned_loss=0.06045, over 3306017.41 frames. ], batch size: 53, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:05:20,759 INFO [optim.py:368] (0/8) 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 15:05:46,848 INFO [train.py:904] (0/8) Epoch 7, batch 1000, loss[loss=0.2096, simple_loss=0.2687, pruned_loss=0.07527, over 16725.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2762, pruned_loss=0.06051, over 3304797.30 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:05:55,628 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 15:06:00,366 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 15:06:05,310 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=61914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:06:30,764 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 15:06:56,262 INFO [train.py:904] (0/8) Epoch 7, batch 1050, loss[loss=0.2253, simple_loss=0.2874, pruned_loss=0.08161, over 16428.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2759, pruned_loss=0.06077, over 3302833.41 frames. ], batch size: 146, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:07:28,197 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=61975.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:07:37,641 INFO [optim.py:368] (0/8) 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 15:07:55,342 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3796, 3.6715, 3.7201, 1.9207, 3.8733, 3.8577, 3.0672, 2.9336], device='cuda:0'), covar=tensor([0.0730, 0.0109, 0.0113, 0.0988, 0.0064, 0.0106, 0.0339, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0095, 0.0083, 0.0140, 0.0072, 0.0088, 0.0118, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 15:08:01,510 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-62000.pt 2023-04-28 15:08:07,103 INFO [train.py:904] (0/8) Epoch 7, batch 1100, loss[loss=0.1708, simple_loss=0.2447, pruned_loss=0.04845, over 16227.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2742, pruned_loss=0.05969, over 3306634.69 frames. ], batch size: 36, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:08:27,005 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0431, 5.4584, 5.5320, 5.4497, 5.4497, 6.0026, 5.6368, 5.4425], device='cuda:0'), covar=tensor([0.0772, 0.1796, 0.2014, 0.1722, 0.2661, 0.0964, 0.1205, 0.2195], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0447, 0.0451, 0.0380, 0.0513, 0.0480, 0.0360, 0.0512], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:09:15,542 INFO [train.py:904] (0/8) Epoch 7, batch 1150, loss[loss=0.181, simple_loss=0.2741, pruned_loss=0.04399, over 17091.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2736, pruned_loss=0.05944, over 3306077.33 frames. ], batch size: 47, lr: 1.01e-02, grad_scale: 4.0 2023-04-28 15:09:57,387 INFO [optim.py:368] (0/8) 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 15:10:04,076 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:18,796 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62099.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:10:22,594 INFO [train.py:904] (0/8) Epoch 7, batch 1200, loss[loss=0.1973, simple_loss=0.2724, pruned_loss=0.06112, over 15597.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2726, pruned_loss=0.05854, over 3313285.27 frames. ], batch size: 190, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:10:51,377 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62122.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:11:11,205 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5500, 4.5369, 4.5706, 4.6359, 4.4972, 5.1602, 4.8106, 4.4956], device='cuda:0'), covar=tensor([0.1262, 0.1773, 0.1908, 0.2007, 0.2874, 0.1087, 0.1166, 0.2652], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0450, 0.0453, 0.0381, 0.0516, 0.0480, 0.0357, 0.0514], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:11:25,388 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3073, 3.5861, 3.6735, 1.8057, 3.8211, 3.8037, 2.8947, 2.7782], device='cuda:0'), covar=tensor([0.0720, 0.0130, 0.0155, 0.1020, 0.0060, 0.0108, 0.0384, 0.0370], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0096, 0.0085, 0.0141, 0.0072, 0.0090, 0.0120, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 15:11:30,507 INFO [train.py:904] (0/8) Epoch 7, batch 1250, loss[loss=0.224, simple_loss=0.2824, pruned_loss=0.08282, over 16818.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2728, pruned_loss=0.05936, over 3318557.03 frames. ], batch size: 124, lr: 1.01e-02, grad_scale: 8.0 2023-04-28 15:12:13,503 INFO [optim.py:368] (0/8) 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 15:12:13,860 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62183.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:12:29,949 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-28 15:12:39,196 INFO [train.py:904] (0/8) Epoch 7, batch 1300, loss[loss=0.2141, simple_loss=0.2773, pruned_loss=0.07545, over 16400.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2728, pruned_loss=0.05957, over 3313677.51 frames. ], batch size: 146, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:13:49,061 INFO [train.py:904] (0/8) Epoch 7, batch 1350, loss[loss=0.1607, simple_loss=0.248, pruned_loss=0.03671, over 17242.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2729, pruned_loss=0.05917, over 3315157.55 frames. ], batch size: 45, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:13:50,908 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6158, 3.5500, 2.7908, 2.3218, 2.6715, 2.1891, 3.6184, 3.4693], device='cuda:0'), covar=tensor([0.2178, 0.0672, 0.1223, 0.1869, 0.2077, 0.1559, 0.0432, 0.0913], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0256, 0.0272, 0.0255, 0.0281, 0.0210, 0.0250, 0.0274], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:14:12,893 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0315, 4.6788, 4.9901, 5.2302, 5.4096, 4.7345, 5.3976, 5.3685], device='cuda:0'), covar=tensor([0.1032, 0.0979, 0.1353, 0.0523, 0.0433, 0.0645, 0.0387, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0569, 0.0723, 0.0574, 0.0436, 0.0428, 0.0458, 0.0494], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:14:15,372 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:19,059 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:14:31,595 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.487e+02 3.145e+02 3.874e+02 9.778e+02, threshold=6.290e+02, percent-clipped=3.0 2023-04-28 15:14:56,761 INFO [train.py:904] (0/8) Epoch 7, batch 1400, loss[loss=0.1906, simple_loss=0.2772, pruned_loss=0.05196, over 16728.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2728, pruned_loss=0.05862, over 3323544.55 frames. ], batch size: 57, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:15:42,068 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:16:04,314 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2778, 4.8966, 5.2180, 5.4208, 5.6208, 4.9471, 5.5357, 5.4899], device='cuda:0'), covar=tensor([0.0931, 0.0867, 0.1260, 0.0470, 0.0328, 0.0483, 0.0394, 0.0416], device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0571, 0.0728, 0.0575, 0.0437, 0.0429, 0.0459, 0.0496], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:16:06,307 INFO [train.py:904] (0/8) Epoch 7, batch 1450, loss[loss=0.1886, simple_loss=0.2609, pruned_loss=0.05818, over 15457.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2726, pruned_loss=0.05833, over 3319822.50 frames. ], batch size: 191, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:16:22,007 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2617, 5.7052, 5.4422, 5.5123, 5.0047, 5.0371, 5.1309, 5.8053], device='cuda:0'), covar=tensor([0.1094, 0.1011, 0.1179, 0.0718, 0.0833, 0.0712, 0.0937, 0.0851], device='cuda:0'), in_proj_covar=tensor([0.0459, 0.0596, 0.0497, 0.0396, 0.0370, 0.0385, 0.0490, 0.0433], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:16:47,646 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.994e+02 2.891e+02 3.326e+02 4.338e+02 8.815e+02, threshold=6.653e+02, percent-clipped=8.0 2023-04-28 15:16:56,013 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:00,014 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62391.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:10,971 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:17:14,294 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 15:17:14,671 INFO [train.py:904] (0/8) Epoch 7, batch 1500, loss[loss=0.2288, simple_loss=0.2941, pruned_loss=0.08171, over 16268.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2724, pruned_loss=0.05854, over 3318404.10 frames. ], batch size: 165, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:17:26,271 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8505, 4.8643, 4.7756, 4.1552, 4.7377, 1.9727, 4.5368, 4.7530], device='cuda:0'), covar=tensor([0.0100, 0.0070, 0.0119, 0.0351, 0.0086, 0.2026, 0.0117, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0094, 0.0146, 0.0140, 0.0112, 0.0160, 0.0130, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:18:01,113 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:15,914 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62447.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:23,654 INFO [train.py:904] (0/8) Epoch 7, batch 1550, loss[loss=0.2226, simple_loss=0.2819, pruned_loss=0.0817, over 16842.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2744, pruned_loss=0.06003, over 3319570.09 frames. ], batch size: 109, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:18:24,199 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62452.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:18:31,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.78 vs. limit=5.0 2023-04-28 15:18:59,846 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62478.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:19:01,451 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-28 15:19:05,636 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 15:19:06,020 INFO [optim.py:368] (0/8) 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 15:19:32,104 INFO [train.py:904] (0/8) Epoch 7, batch 1600, loss[loss=0.2174, simple_loss=0.3045, pruned_loss=0.06516, over 16663.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2771, pruned_loss=0.06099, over 3318742.58 frames. ], batch size: 62, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:19:35,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9846, 5.0054, 5.6524, 5.5861, 5.5610, 5.1102, 5.0938, 4.9101], device='cuda:0'), covar=tensor([0.0238, 0.0451, 0.0252, 0.0324, 0.0331, 0.0286, 0.0743, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0287, 0.0294, 0.0277, 0.0336, 0.0306, 0.0408, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 15:20:01,678 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6377, 2.2386, 2.4106, 4.3989, 2.0954, 2.9305, 2.3232, 2.4944], device='cuda:0'), covar=tensor([0.0719, 0.2641, 0.1475, 0.0320, 0.3028, 0.1580, 0.2335, 0.2542], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0347, 0.0290, 0.0319, 0.0381, 0.0377, 0.0316, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:20:31,547 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62545.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:20:40,969 INFO [train.py:904] (0/8) Epoch 7, batch 1650, loss[loss=0.2141, simple_loss=0.2808, pruned_loss=0.0737, over 16416.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.279, pruned_loss=0.06169, over 3321064.04 frames. ], batch size: 146, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:20:47,323 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.79 vs. limit=2.0 2023-04-28 15:21:04,900 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62570.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:21:22,702 INFO [optim.py:368] (0/8) 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 15:21:25,354 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:21:49,316 INFO [train.py:904] (0/8) Epoch 7, batch 1700, loss[loss=0.3038, simple_loss=0.3545, pruned_loss=0.1265, over 12191.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2817, pruned_loss=0.0632, over 3308093.97 frames. ], batch size: 246, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:21:55,144 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62606.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:12,361 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62618.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:27,853 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:51,298 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:22:58,556 INFO [train.py:904] (0/8) Epoch 7, batch 1750, loss[loss=0.2162, simple_loss=0.3069, pruned_loss=0.06275, over 17259.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2826, pruned_loss=0.06358, over 3298493.49 frames. ], batch size: 52, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:23:18,041 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62666.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:23:21,081 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1536, 4.7191, 4.5141, 5.2748, 5.3923, 4.7680, 5.2556, 5.3890], device='cuda:0'), covar=tensor([0.1048, 0.0931, 0.2411, 0.0744, 0.0750, 0.0706, 0.0802, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0569, 0.0722, 0.0572, 0.0440, 0.0429, 0.0457, 0.0493], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:23:41,547 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.869e+02 2.853e+02 3.319e+02 4.250e+02 7.879e+02, threshold=6.638e+02, percent-clipped=1.0 2023-04-28 15:24:08,276 INFO [train.py:904] (0/8) Epoch 7, batch 1800, loss[loss=0.2217, simple_loss=0.2924, pruned_loss=0.07554, over 16683.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2837, pruned_loss=0.0636, over 3298251.99 frames. ], batch size: 134, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:24:42,385 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:24:51,487 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-28 15:25:10,920 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:17,222 INFO [train.py:904] (0/8) Epoch 7, batch 1850, loss[loss=0.2467, simple_loss=0.324, pruned_loss=0.08465, over 15631.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.285, pruned_loss=0.06408, over 3304820.88 frames. ], batch size: 190, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:25:45,788 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=62773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:52,396 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:25:59,966 INFO [optim.py:368] (0/8) 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 15:26:26,228 INFO [train.py:904] (0/8) Epoch 7, batch 1900, loss[loss=0.2023, simple_loss=0.2944, pruned_loss=0.05513, over 17087.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2826, pruned_loss=0.06269, over 3308256.98 frames. ], batch size: 53, lr: 1.00e-02, grad_scale: 8.0 2023-04-28 15:26:38,877 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6337, 4.7355, 4.8293, 4.8154, 4.6539, 5.3345, 4.9276, 4.6294], device='cuda:0'), covar=tensor([0.1216, 0.1591, 0.1614, 0.1815, 0.2879, 0.1089, 0.1381, 0.2777], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0457, 0.0454, 0.0385, 0.0518, 0.0489, 0.0369, 0.0519], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:26:45,043 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6501, 1.2422, 1.5262, 1.7019, 1.7072, 1.9231, 1.4078, 1.7827], device='cuda:0'), covar=tensor([0.0127, 0.0227, 0.0118, 0.0174, 0.0139, 0.0097, 0.0216, 0.0073], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0156, 0.0141, 0.0141, 0.0145, 0.0102, 0.0150, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 15:27:00,820 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62826.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:11,565 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=62834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:27:36,784 INFO [train.py:904] (0/8) Epoch 7, batch 1950, loss[loss=0.1688, simple_loss=0.2559, pruned_loss=0.04087, over 17172.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.281, pruned_loss=0.0614, over 3318342.34 frames. ], batch size: 46, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:27:58,143 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-04-28 15:28:19,259 INFO [optim.py:368] (0/8) 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 15:28:44,336 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62901.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:28:45,197 INFO [train.py:904] (0/8) Epoch 7, batch 2000, loss[loss=0.2174, simple_loss=0.2844, pruned_loss=0.07516, over 16245.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2809, pruned_loss=0.0608, over 3325258.45 frames. ], batch size: 165, lr: 9.99e-03, grad_scale: 8.0 2023-04-28 15:28:53,373 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2203, 4.5786, 4.3561, 4.4076, 4.0383, 4.0357, 4.1779, 4.5964], device='cuda:0'), covar=tensor([0.0995, 0.0809, 0.0964, 0.0576, 0.0666, 0.1475, 0.0817, 0.0841], device='cuda:0'), in_proj_covar=tensor([0.0458, 0.0594, 0.0498, 0.0393, 0.0369, 0.0379, 0.0486, 0.0432], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:29:23,820 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=62929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:39,606 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=62941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:29:55,698 INFO [train.py:904] (0/8) Epoch 7, batch 2050, loss[loss=0.2136, simple_loss=0.3029, pruned_loss=0.06213, over 17039.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2816, pruned_loss=0.06106, over 3319819.02 frames. ], batch size: 53, lr: 9.99e-03, grad_scale: 16.0 2023-04-28 15:30:23,969 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8216, 4.0433, 3.0390, 2.4471, 2.9167, 2.4088, 4.1183, 3.8278], device='cuda:0'), covar=tensor([0.2040, 0.0550, 0.1204, 0.1727, 0.2143, 0.1578, 0.0391, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0259, 0.0273, 0.0258, 0.0285, 0.0211, 0.0253, 0.0278], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:30:31,323 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=62977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:30:39,211 INFO [optim.py:368] (0/8) 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 15:31:05,152 INFO [train.py:904] (0/8) Epoch 7, batch 2100, loss[loss=0.1911, simple_loss=0.2848, pruned_loss=0.04869, over 17099.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2828, pruned_loss=0.06222, over 3318117.19 frames. ], batch size: 53, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:31:33,222 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63022.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:09,392 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63047.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:32:09,548 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2815, 2.0200, 1.5797, 1.8008, 2.4162, 2.2203, 2.3971, 2.5172], device='cuda:0'), covar=tensor([0.0078, 0.0208, 0.0270, 0.0253, 0.0114, 0.0187, 0.0139, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0180, 0.0176, 0.0179, 0.0177, 0.0183, 0.0176, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:32:16,876 INFO [train.py:904] (0/8) Epoch 7, batch 2150, loss[loss=0.1795, simple_loss=0.2709, pruned_loss=0.0441, over 17206.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2829, pruned_loss=0.06239, over 3311080.70 frames. ], batch size: 46, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:33:00,562 INFO [optim.py:368] (0/8) 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 15:33:17,103 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:33:25,766 INFO [train.py:904] (0/8) Epoch 7, batch 2200, loss[loss=0.1853, simple_loss=0.2609, pruned_loss=0.0548, over 15902.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2838, pruned_loss=0.06296, over 3319458.79 frames. ], batch size: 35, lr: 9.98e-03, grad_scale: 8.0 2023-04-28 15:34:03,308 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:34:33,879 INFO [train.py:904] (0/8) Epoch 7, batch 2250, loss[loss=0.2195, simple_loss=0.3107, pruned_loss=0.06417, over 16693.00 frames. ], tot_loss[loss=0.205, simple_loss=0.284, pruned_loss=0.06299, over 3323684.85 frames. ], batch size: 57, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:35:18,527 INFO [optim.py:368] (0/8) 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 15:35:22,100 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9134, 3.2697, 2.4412, 4.3375, 3.9425, 4.1181, 1.5652, 2.8337], device='cuda:0'), covar=tensor([0.0958, 0.0351, 0.0908, 0.0097, 0.0172, 0.0408, 0.1036, 0.0657], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0146, 0.0169, 0.0103, 0.0197, 0.0199, 0.0165, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 15:35:42,327 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63201.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:35:43,099 INFO [train.py:904] (0/8) Epoch 7, batch 2300, loss[loss=0.2092, simple_loss=0.2813, pruned_loss=0.06858, over 16762.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2845, pruned_loss=0.06382, over 3316959.82 frames. ], batch size: 83, lr: 9.97e-03, grad_scale: 8.0 2023-04-28 15:36:10,177 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63221.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:36,742 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:47,592 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63249.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:36:51,524 INFO [train.py:904] (0/8) Epoch 7, batch 2350, loss[loss=0.2416, simple_loss=0.3061, pruned_loss=0.08851, over 16146.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2848, pruned_loss=0.06361, over 3323268.73 frames. ], batch size: 164, lr: 9.96e-03, grad_scale: 4.0 2023-04-28 15:37:33,972 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:37:38,005 INFO [optim.py:368] (0/8) 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 15:37:43,613 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 7, batch 2400, loss[loss=0.1767, simple_loss=0.2623, pruned_loss=0.04553, over 17198.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2859, pruned_loss=0.06411, over 3325177.36 frames. ], batch size: 43, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:38:27,612 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63322.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:10,370 INFO [train.py:904] (0/8) Epoch 7, batch 2450, loss[loss=0.2372, simple_loss=0.308, pruned_loss=0.08316, over 16413.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2863, pruned_loss=0.06344, over 3327522.60 frames. ], batch size: 146, lr: 9.96e-03, grad_scale: 8.0 2023-04-28 15:39:34,205 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63370.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:39:52,705 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8277, 4.1388, 2.1188, 4.3855, 2.7134, 4.3995, 2.4183, 3.0708], device='cuda:0'), covar=tensor([0.0158, 0.0208, 0.1312, 0.0106, 0.0694, 0.0323, 0.1199, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0164, 0.0181, 0.0100, 0.0161, 0.0205, 0.0189, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 15:39:55,102 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.805e+02 3.565e+02 4.384e+02 1.263e+03, threshold=7.129e+02, percent-clipped=5.0 2023-04-28 15:39:55,841 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 15:40:19,265 INFO [train.py:904] (0/8) Epoch 7, batch 2500, loss[loss=0.2386, simple_loss=0.3249, pruned_loss=0.0762, over 16019.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.286, pruned_loss=0.06292, over 3323761.82 frames. ], batch size: 35, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:40:43,683 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:46,514 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63422.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:55,021 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:40:58,951 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.69 vs. limit=2.0 2023-04-28 15:41:26,931 INFO [train.py:904] (0/8) Epoch 7, batch 2550, loss[loss=0.1805, simple_loss=0.2688, pruned_loss=0.04606, over 16837.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2854, pruned_loss=0.06222, over 3329927.98 frames. ], batch size: 42, lr: 9.95e-03, grad_scale: 4.0 2023-04-28 15:42:01,826 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:08,260 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63481.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:42:11,091 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 3.238e+02 3.870e+02 4.521e+02 8.304e+02, threshold=7.740e+02, percent-clipped=2.0 2023-04-28 15:42:18,587 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 15:42:35,906 INFO [train.py:904] (0/8) Epoch 7, batch 2600, loss[loss=0.1953, simple_loss=0.2691, pruned_loss=0.06078, over 16872.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2857, pruned_loss=0.06225, over 3323091.18 frames. ], batch size: 116, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:43:42,058 INFO [train.py:904] (0/8) Epoch 7, batch 2650, loss[loss=0.1909, simple_loss=0.2854, pruned_loss=0.04817, over 17118.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2863, pruned_loss=0.06201, over 3321122.16 frames. ], batch size: 48, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:44:05,512 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63569.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:15,591 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63577.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:27,535 INFO [optim.py:368] (0/8) 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 15:44:27,949 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63586.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:44:31,767 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3422, 2.3113, 1.9214, 2.1561, 2.8473, 2.5413, 3.2461, 3.0649], device='cuda:0'), covar=tensor([0.0047, 0.0259, 0.0318, 0.0285, 0.0140, 0.0232, 0.0160, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0175, 0.0172, 0.0175, 0.0172, 0.0177, 0.0173, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:44:48,754 INFO [train.py:904] (0/8) Epoch 7, batch 2700, loss[loss=0.2022, simple_loss=0.2905, pruned_loss=0.05694, over 16552.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2868, pruned_loss=0.06159, over 3326678.21 frames. ], batch size: 68, lr: 9.94e-03, grad_scale: 4.0 2023-04-28 15:45:05,661 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9632, 4.6709, 4.9447, 5.2288, 5.3588, 4.7040, 5.3220, 5.3715], device='cuda:0'), covar=tensor([0.1239, 0.1030, 0.1615, 0.0534, 0.0526, 0.0686, 0.0474, 0.0412], device='cuda:0'), in_proj_covar=tensor([0.0476, 0.0583, 0.0740, 0.0581, 0.0448, 0.0447, 0.0458, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:45:14,501 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2411, 3.2240, 3.5994, 2.3142, 3.3185, 3.6439, 3.4378, 1.9234], device='cuda:0'), covar=tensor([0.0335, 0.0114, 0.0039, 0.0271, 0.0058, 0.0056, 0.0058, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0063, 0.0061, 0.0116, 0.0065, 0.0076, 0.0067, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:45:28,045 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7537, 4.1540, 4.2613, 3.0682, 3.6678, 4.2064, 3.9692, 2.5906], device='cuda:0'), covar=tensor([0.0299, 0.0035, 0.0035, 0.0234, 0.0066, 0.0047, 0.0041, 0.0288], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0063, 0.0061, 0.0116, 0.0065, 0.0076, 0.0067, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:45:28,089 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63630.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:44,451 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-28 15:45:51,353 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63647.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:45:58,007 INFO [train.py:904] (0/8) Epoch 7, batch 2750, loss[loss=0.1876, simple_loss=0.2875, pruned_loss=0.04388, over 17238.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2865, pruned_loss=0.06106, over 3320986.64 frames. ], batch size: 52, lr: 9.93e-03, grad_scale: 4.0 2023-04-28 15:46:36,377 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3439, 5.2495, 5.0943, 4.7983, 4.7153, 5.2310, 5.1298, 4.7890], device='cuda:0'), covar=tensor([0.0432, 0.0372, 0.0245, 0.0246, 0.0930, 0.0298, 0.0286, 0.0587], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0256, 0.0262, 0.0236, 0.0299, 0.0264, 0.0186, 0.0300], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:46:45,849 INFO [optim.py:368] (0/8) 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 15:47:08,535 INFO [train.py:904] (0/8) Epoch 7, batch 2800, loss[loss=0.2405, simple_loss=0.3055, pruned_loss=0.08775, over 12400.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.286, pruned_loss=0.06087, over 3324551.18 frames. ], batch size: 246, lr: 9.93e-03, grad_scale: 8.0 2023-04-28 15:47:49,146 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-28 15:48:12,854 INFO [train.py:904] (0/8) Epoch 7, batch 2850, loss[loss=0.2024, simple_loss=0.2647, pruned_loss=0.07011, over 16691.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2859, pruned_loss=0.0613, over 3328972.45 frames. ], batch size: 134, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:48:47,365 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63776.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:48:50,125 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63778.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:49:02,039 INFO [optim.py:368] (0/8) 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 15:49:09,451 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3883, 3.9521, 3.3680, 2.0325, 2.9990, 2.5077, 3.7549, 3.8408], device='cuda:0'), covar=tensor([0.0223, 0.0531, 0.0592, 0.1518, 0.0655, 0.0887, 0.0439, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0143, 0.0159, 0.0144, 0.0136, 0.0126, 0.0141, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 15:49:23,199 INFO [train.py:904] (0/8) Epoch 7, batch 2900, loss[loss=0.1875, simple_loss=0.2815, pruned_loss=0.04679, over 17089.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2851, pruned_loss=0.06153, over 3330175.35 frames. ], batch size: 49, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:49:35,851 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63810.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:50:22,303 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 7, batch 2950, loss[loss=0.2324, simple_loss=0.2948, pruned_loss=0.08504, over 16891.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2841, pruned_loss=0.0623, over 3322894.87 frames. ], batch size: 109, lr: 9.92e-03, grad_scale: 8.0 2023-04-28 15:50:56,598 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 15:50:58,332 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63871.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:05,331 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=63877.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:51:17,102 INFO [optim.py:368] (0/8) 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 15:51:38,108 INFO [train.py:904] (0/8) Epoch 7, batch 3000, loss[loss=0.1886, simple_loss=0.2782, pruned_loss=0.04949, over 17185.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2842, pruned_loss=0.06212, over 3331692.96 frames. ], batch size: 46, lr: 9.91e-03, grad_scale: 8.0 2023-04-28 15:51:38,109 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 15:51:46,842 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 15:51:54,322 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63906.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:06,748 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=63916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:19,427 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:19,446 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63925.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:42,095 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=63942.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:52:53,886 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6016, 3.8851, 2.1283, 4.0727, 2.6197, 4.1071, 2.3521, 3.0064], device='cuda:0'), covar=tensor([0.0150, 0.0219, 0.1271, 0.0103, 0.0686, 0.0310, 0.1110, 0.0521], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0163, 0.0179, 0.0100, 0.0162, 0.0205, 0.0189, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 15:52:56,543 INFO [train.py:904] (0/8) Epoch 7, batch 3050, loss[loss=0.2068, simple_loss=0.281, pruned_loss=0.06627, over 16437.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2849, pruned_loss=0.06266, over 3325293.73 frames. ], batch size: 75, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:53:30,244 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=63977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:53:43,708 INFO [optim.py:368] (0/8) 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 15:54:00,926 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-64000.pt 2023-04-28 15:54:06,489 INFO [train.py:904] (0/8) Epoch 7, batch 3100, loss[loss=0.1778, simple_loss=0.2698, pruned_loss=0.04292, over 17046.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2841, pruned_loss=0.06215, over 3326150.75 frames. ], batch size: 50, lr: 9.91e-03, grad_scale: 4.0 2023-04-28 15:54:35,896 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64023.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:03,753 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1504, 1.6980, 2.3731, 2.9436, 2.5606, 3.3280, 1.9327, 3.2662], device='cuda:0'), covar=tensor([0.0109, 0.0292, 0.0187, 0.0162, 0.0165, 0.0098, 0.0266, 0.0080], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0157, 0.0141, 0.0143, 0.0147, 0.0105, 0.0149, 0.0096], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 15:55:16,385 INFO [train.py:904] (0/8) Epoch 7, batch 3150, loss[loss=0.2199, simple_loss=0.2894, pruned_loss=0.07518, over 16790.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2839, pruned_loss=0.06183, over 3327480.97 frames. ], batch size: 102, lr: 9.90e-03, grad_scale: 4.0 2023-04-28 15:55:36,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3259, 4.1258, 4.3513, 4.5678, 4.6377, 4.2359, 4.3458, 4.6452], device='cuda:0'), covar=tensor([0.1031, 0.0798, 0.1161, 0.0528, 0.0500, 0.0935, 0.1334, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0597, 0.0764, 0.0602, 0.0459, 0.0463, 0.0474, 0.0519], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 15:55:49,708 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64076.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:55:52,155 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64078.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:00,146 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64084.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:03,686 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.615e+02 2.613e+02 3.331e+02 4.706e+02 9.304e+02, threshold=6.662e+02, percent-clipped=4.0 2023-04-28 15:56:24,292 INFO [train.py:904] (0/8) Epoch 7, batch 3200, loss[loss=0.2173, simple_loss=0.2958, pruned_loss=0.06933, over 12373.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2825, pruned_loss=0.06149, over 3322025.84 frames. ], batch size: 246, lr: 9.90e-03, grad_scale: 8.0 2023-04-28 15:56:29,029 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0580, 4.2306, 4.4229, 2.1849, 4.7220, 4.7146, 3.1439, 3.7014], device='cuda:0'), covar=tensor([0.0618, 0.0136, 0.0184, 0.1024, 0.0048, 0.0065, 0.0347, 0.0310], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0099, 0.0086, 0.0139, 0.0073, 0.0092, 0.0121, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 15:56:54,785 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64124.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:56:57,817 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64126.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:57:00,613 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-28 15:57:33,487 INFO [train.py:904] (0/8) Epoch 7, batch 3250, loss[loss=0.2114, simple_loss=0.2781, pruned_loss=0.07236, over 16832.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2815, pruned_loss=0.0608, over 3329249.93 frames. ], batch size: 90, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:57:53,167 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64166.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:57:57,501 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-28 15:58:21,283 INFO [optim.py:368] (0/8) 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 15:58:42,129 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64201.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:58:43,055 INFO [train.py:904] (0/8) Epoch 7, batch 3300, loss[loss=0.1556, simple_loss=0.2426, pruned_loss=0.03434, over 16855.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2828, pruned_loss=0.0616, over 3332025.77 frames. ], batch size: 42, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 15:59:07,156 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6390, 4.3310, 4.4138, 2.9771, 3.6980, 4.2884, 4.0467, 2.6439], device='cuda:0'), covar=tensor([0.0303, 0.0022, 0.0024, 0.0230, 0.0058, 0.0042, 0.0034, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0062, 0.0062, 0.0117, 0.0067, 0.0078, 0.0069, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:59:16,277 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:19,001 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5026, 3.3908, 2.7066, 2.1719, 2.4260, 2.0664, 3.4226, 3.2755], device='cuda:0'), covar=tensor([0.2073, 0.0613, 0.1176, 0.1947, 0.2012, 0.1593, 0.0472, 0.0982], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0258, 0.0273, 0.0259, 0.0289, 0.0211, 0.0255, 0.0282], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 15:59:25,471 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:40,480 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64242.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 15:59:51,969 INFO [train.py:904] (0/8) Epoch 7, batch 3350, loss[loss=0.2326, simple_loss=0.2888, pruned_loss=0.08817, over 16936.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2834, pruned_loss=0.06169, over 3328818.16 frames. ], batch size: 109, lr: 9.89e-03, grad_scale: 8.0 2023-04-28 16:00:20,486 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:22,691 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:42,051 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.854e+02 2.871e+02 3.440e+02 4.131e+02 8.469e+02, threshold=6.880e+02, percent-clipped=3.0 2023-04-28 16:00:45,939 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64290.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:00:49,553 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:01:02,162 INFO [train.py:904] (0/8) Epoch 7, batch 3400, loss[loss=0.1715, simple_loss=0.2617, pruned_loss=0.04064, over 17145.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2833, pruned_loss=0.0616, over 3330484.77 frames. ], batch size: 47, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:01:06,751 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2970, 5.7021, 5.3926, 5.4827, 5.0807, 4.9068, 5.1786, 5.7831], device='cuda:0'), covar=tensor([0.1019, 0.0766, 0.0940, 0.0479, 0.0821, 0.0639, 0.0735, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0472, 0.0612, 0.0508, 0.0399, 0.0384, 0.0394, 0.0501, 0.0447], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:01:55,876 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8033, 4.3516, 2.4459, 4.6849, 2.8754, 4.6482, 2.4858, 3.2636], device='cuda:0'), covar=tensor([0.0196, 0.0213, 0.1238, 0.0072, 0.0715, 0.0269, 0.1243, 0.0562], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0162, 0.0177, 0.0100, 0.0161, 0.0204, 0.0189, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 16:02:02,952 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64345.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:02:10,863 INFO [train.py:904] (0/8) Epoch 7, batch 3450, loss[loss=0.2235, simple_loss=0.2979, pruned_loss=0.07457, over 11743.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.282, pruned_loss=0.06167, over 3315764.46 frames. ], batch size: 246, lr: 9.88e-03, grad_scale: 8.0 2023-04-28 16:02:49,996 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64379.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:02:51,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3165, 5.2680, 5.1560, 4.6094, 5.1392, 2.1668, 4.8424, 5.2055], device='cuda:0'), covar=tensor([0.0060, 0.0046, 0.0102, 0.0279, 0.0061, 0.1636, 0.0091, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0114, 0.0099, 0.0150, 0.0146, 0.0116, 0.0158, 0.0134, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:03:01,787 INFO [optim.py:368] (0/8) 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 16:03:21,655 INFO [train.py:904] (0/8) Epoch 7, batch 3500, loss[loss=0.2413, simple_loss=0.3074, pruned_loss=0.08759, over 16708.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2804, pruned_loss=0.06051, over 3314159.25 frames. ], batch size: 134, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:03:28,715 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64406.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:30,075 INFO [train.py:904] (0/8) Epoch 7, batch 3550, loss[loss=0.167, simple_loss=0.259, pruned_loss=0.03745, over 17241.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2795, pruned_loss=0.06027, over 3305127.34 frames. ], batch size: 45, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:04:48,156 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64465.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:04:49,888 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:18,415 INFO [optim.py:368] (0/8) 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 16:05:21,016 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-28 16:05:33,043 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9617, 3.9902, 3.8445, 3.6887, 3.4963, 3.9688, 3.6200, 3.6949], device='cuda:0'), covar=tensor([0.0529, 0.0439, 0.0306, 0.0252, 0.0872, 0.0362, 0.1008, 0.0528], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0260, 0.0268, 0.0240, 0.0305, 0.0268, 0.0186, 0.0304], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 16:05:37,852 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64501.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:05:39,315 INFO [train.py:904] (0/8) Epoch 7, batch 3600, loss[loss=0.2213, simple_loss=0.2842, pruned_loss=0.07923, over 16845.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2779, pruned_loss=0.05975, over 3303887.76 frames. ], batch size: 116, lr: 9.87e-03, grad_scale: 8.0 2023-04-28 16:05:56,512 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64514.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:14,242 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64526.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:18,460 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9163, 5.1005, 4.8267, 4.7201, 3.9248, 5.0550, 5.0063, 4.4908], device='cuda:0'), covar=tensor([0.0752, 0.0456, 0.0407, 0.0267, 0.2039, 0.0362, 0.0286, 0.0666], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0258, 0.0266, 0.0239, 0.0304, 0.0266, 0.0185, 0.0301], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 16:06:20,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4506, 5.8203, 5.5492, 5.5887, 5.1736, 4.8968, 5.2909, 5.9322], device='cuda:0'), covar=tensor([0.0947, 0.0777, 0.1011, 0.0574, 0.0723, 0.0676, 0.0844, 0.0750], device='cuda:0'), in_proj_covar=tensor([0.0470, 0.0610, 0.0507, 0.0397, 0.0379, 0.0393, 0.0497, 0.0444], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:06:45,519 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64549.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:06:48,849 INFO [train.py:904] (0/8) Epoch 7, batch 3650, loss[loss=0.1876, simple_loss=0.2718, pruned_loss=0.05169, over 16835.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2774, pruned_loss=0.06015, over 3296182.59 frames. ], batch size: 42, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:06:51,329 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7642, 3.2416, 2.8461, 1.8173, 2.5971, 2.0965, 3.3300, 3.2244], device='cuda:0'), covar=tensor([0.0245, 0.0609, 0.0542, 0.1528, 0.0702, 0.0873, 0.0508, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0140, 0.0155, 0.0140, 0.0132, 0.0124, 0.0138, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 16:07:18,653 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64572.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:40,347 INFO [optim.py:368] (0/8) 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 16:07:42,465 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64588.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:44,471 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64589.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:07:50,093 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6481, 4.6209, 4.6191, 4.1366, 4.5514, 1.8314, 4.4065, 4.4119], device='cuda:0'), covar=tensor([0.0094, 0.0064, 0.0126, 0.0259, 0.0068, 0.1869, 0.0097, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0099, 0.0150, 0.0147, 0.0117, 0.0158, 0.0134, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:08:02,005 INFO [train.py:904] (0/8) Epoch 7, batch 3700, loss[loss=0.1948, simple_loss=0.2648, pruned_loss=0.06238, over 16256.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2758, pruned_loss=0.06176, over 3276857.02 frames. ], batch size: 165, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:08:27,964 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:09,886 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:12,340 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:09:14,084 INFO [train.py:904] (0/8) Epoch 7, batch 3750, loss[loss=0.2176, simple_loss=0.2815, pruned_loss=0.07689, over 16757.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2771, pruned_loss=0.06353, over 3266427.14 frames. ], batch size: 124, lr: 9.86e-03, grad_scale: 8.0 2023-04-28 16:09:38,412 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-28 16:09:55,052 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64679.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:04,728 INFO [optim.py:368] (0/8) 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 16:10:24,144 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 16:10:24,868 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64701.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:25,833 INFO [train.py:904] (0/8) Epoch 7, batch 3800, loss[loss=0.2118, simple_loss=0.2934, pruned_loss=0.06512, over 16657.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2795, pruned_loss=0.06583, over 3255989.78 frames. ], batch size: 57, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:10:27,298 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64703.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:38,597 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64709.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:10:44,385 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2237, 5.2172, 5.0369, 4.8202, 4.5994, 5.1291, 5.0047, 4.8219], device='cuda:0'), covar=tensor([0.0406, 0.0196, 0.0167, 0.0172, 0.0913, 0.0205, 0.0228, 0.0498], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0252, 0.0257, 0.0232, 0.0294, 0.0259, 0.0180, 0.0294], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 16:10:46,097 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-28 16:11:05,639 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64727.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:31,487 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64745.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:39,997 INFO [train.py:904] (0/8) Epoch 7, batch 3850, loss[loss=0.2063, simple_loss=0.2819, pruned_loss=0.06539, over 15607.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2797, pruned_loss=0.06622, over 3259961.21 frames. ], batch size: 191, lr: 9.85e-03, grad_scale: 4.0 2023-04-28 16:11:40,527 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=64752.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:11:51,009 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 16:11:58,634 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64764.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:12:33,101 INFO [optim.py:368] (0/8) 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 16:12:39,503 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1453, 1.8770, 2.1862, 3.6328, 1.9461, 2.4049, 2.0557, 2.0171], device='cuda:0'), covar=tensor([0.0767, 0.2726, 0.1498, 0.0375, 0.2929, 0.1692, 0.2562, 0.2652], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0359, 0.0299, 0.0325, 0.0389, 0.0393, 0.0322, 0.0425], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:12:52,613 INFO [train.py:904] (0/8) Epoch 7, batch 3900, loss[loss=0.1973, simple_loss=0.2696, pruned_loss=0.0625, over 16398.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.279, pruned_loss=0.06664, over 3265147.22 frames. ], batch size: 35, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:13:00,365 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64806.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:08,309 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=64813.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:13:20,519 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64821.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:14:02,899 INFO [train.py:904] (0/8) Epoch 7, batch 3950, loss[loss=0.1788, simple_loss=0.2452, pruned_loss=0.05618, over 16489.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2787, pruned_loss=0.06722, over 3256003.64 frames. ], batch size: 75, lr: 9.84e-03, grad_scale: 4.0 2023-04-28 16:14:53,036 INFO [optim.py:368] (0/8) 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 16:14:53,489 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=64888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:14:58,556 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-28 16:15:10,993 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7947, 5.1968, 5.3803, 5.2290, 5.2685, 5.8015, 5.3356, 5.0729], device='cuda:0'), covar=tensor([0.0848, 0.1258, 0.1132, 0.1445, 0.2106, 0.0855, 0.0959, 0.1834], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0449, 0.0446, 0.0372, 0.0505, 0.0474, 0.0357, 0.0505], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 16:15:12,242 INFO [train.py:904] (0/8) Epoch 7, batch 4000, loss[loss=0.1854, simple_loss=0.2647, pruned_loss=0.05309, over 16621.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2779, pruned_loss=0.06712, over 3267121.34 frames. ], batch size: 62, lr: 9.84e-03, grad_scale: 8.0 2023-04-28 16:15:16,368 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9655, 1.7998, 2.2661, 2.8103, 2.6724, 2.8384, 1.4948, 2.9142], device='cuda:0'), covar=tensor([0.0084, 0.0245, 0.0161, 0.0111, 0.0120, 0.0098, 0.0278, 0.0054], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0154, 0.0139, 0.0141, 0.0146, 0.0105, 0.0148, 0.0095], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 16:16:01,468 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=64936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:11,947 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7114, 4.7132, 5.2165, 5.1150, 5.1555, 4.7587, 4.7487, 4.4737], device='cuda:0'), covar=tensor([0.0241, 0.0533, 0.0256, 0.0422, 0.0424, 0.0271, 0.0765, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0294, 0.0291, 0.0282, 0.0339, 0.0307, 0.0414, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 16:16:16,965 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=64945.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:16:24,740 INFO [train.py:904] (0/8) Epoch 7, batch 4050, loss[loss=0.188, simple_loss=0.2683, pruned_loss=0.05385, over 16899.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2768, pruned_loss=0.06509, over 3271074.48 frames. ], batch size: 109, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:16:51,241 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3107, 4.0381, 4.0061, 2.7262, 3.5696, 3.9700, 3.8345, 2.4117], device='cuda:0'), covar=tensor([0.0355, 0.0023, 0.0021, 0.0221, 0.0039, 0.0058, 0.0026, 0.0253], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0059, 0.0060, 0.0114, 0.0064, 0.0074, 0.0066, 0.0110], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 16:16:59,743 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 16:17:35,694 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65001.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:17:36,416 INFO [train.py:904] (0/8) Epoch 7, batch 4100, loss[loss=0.2036, simple_loss=0.2827, pruned_loss=0.06224, over 16669.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2773, pruned_loss=0.06335, over 3275671.41 frames. ], batch size: 62, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:17:39,350 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65004.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:00,461 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8224, 5.2293, 5.4759, 5.3340, 5.3059, 5.8651, 5.4748, 5.1938], device='cuda:0'), covar=tensor([0.0716, 0.1376, 0.1200, 0.1307, 0.2090, 0.0820, 0.0987, 0.1769], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0448, 0.0444, 0.0372, 0.0504, 0.0473, 0.0356, 0.0503], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 16:18:16,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3130, 3.1986, 3.2433, 3.4510, 3.4854, 3.1383, 3.3905, 3.5555], device='cuda:0'), covar=tensor([0.0800, 0.0738, 0.1157, 0.0522, 0.0586, 0.3097, 0.1121, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0449, 0.0547, 0.0692, 0.0560, 0.0424, 0.0424, 0.0434, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:18:27,619 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-04-28 16:18:29,118 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65037.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:18:29,183 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9365, 2.4898, 2.2882, 3.0975, 2.5286, 3.2338, 1.7637, 2.7315], device='cuda:0'), covar=tensor([0.1069, 0.0459, 0.1000, 0.0115, 0.0193, 0.0344, 0.1197, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0148, 0.0171, 0.0108, 0.0203, 0.0199, 0.0168, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 16:18:34,840 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-28 16:18:46,546 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:51,083 INFO [train.py:904] (0/8) Epoch 7, batch 4150, loss[loss=0.2925, simple_loss=0.3527, pruned_loss=0.1162, over 11612.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2856, pruned_loss=0.06711, over 3230856.50 frames. ], batch size: 246, lr: 9.83e-03, grad_scale: 8.0 2023-04-28 16:18:56,798 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:18:59,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8955, 4.7625, 4.6738, 3.6473, 4.6803, 1.6775, 4.4557, 4.5300], device='cuda:0'), covar=tensor([0.0060, 0.0063, 0.0110, 0.0434, 0.0074, 0.2185, 0.0111, 0.0171], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0096, 0.0145, 0.0143, 0.0113, 0.0154, 0.0130, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:19:01,901 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65059.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:19:39,624 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8699, 1.8313, 2.2692, 2.9019, 2.7468, 3.2765, 1.7749, 3.0108], device='cuda:0'), covar=tensor([0.0101, 0.0256, 0.0160, 0.0127, 0.0113, 0.0061, 0.0251, 0.0068], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0158, 0.0141, 0.0142, 0.0149, 0.0107, 0.0152, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 16:19:44,055 INFO [optim.py:368] (0/8) 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 16:20:03,718 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65101.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:04,514 INFO [train.py:904] (0/8) Epoch 7, batch 4200, loss[loss=0.2535, simple_loss=0.3468, pruned_loss=0.08007, over 16758.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2934, pruned_loss=0.06939, over 3193708.58 frames. ], batch size: 124, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:20:13,271 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65108.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:24,684 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:20:32,087 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65121.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:21:16,736 INFO [train.py:904] (0/8) Epoch 7, batch 4250, loss[loss=0.1961, simple_loss=0.2834, pruned_loss=0.05445, over 17249.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2956, pruned_loss=0.06914, over 3177210.45 frames. ], batch size: 45, lr: 9.82e-03, grad_scale: 8.0 2023-04-28 16:21:21,533 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 16:21:42,350 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:21:58,310 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-28 16:22:08,031 INFO [optim.py:368] (0/8) 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 16:22:11,441 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-28 16:22:29,073 INFO [train.py:904] (0/8) Epoch 7, batch 4300, loss[loss=0.2187, simple_loss=0.3097, pruned_loss=0.0638, over 16686.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2967, pruned_loss=0.06829, over 3177800.13 frames. ], batch size: 83, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:22:34,484 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5694, 2.8922, 2.4215, 4.1154, 3.1076, 3.8450, 1.3653, 2.7967], device='cuda:0'), covar=tensor([0.1500, 0.0598, 0.1272, 0.0136, 0.0319, 0.0364, 0.1803, 0.0874], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0146, 0.0169, 0.0105, 0.0199, 0.0195, 0.0166, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 16:23:31,788 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65245.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:23:41,838 INFO [train.py:904] (0/8) Epoch 7, batch 4350, loss[loss=0.2236, simple_loss=0.3044, pruned_loss=0.07138, over 16263.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3002, pruned_loss=0.0692, over 3194913.35 frames. ], batch size: 35, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:30,740 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65285.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:33,881 INFO [optim.py:368] (0/8) 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 16:24:42,174 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:24:54,960 INFO [train.py:904] (0/8) Epoch 7, batch 4400, loss[loss=0.2346, simple_loss=0.3095, pruned_loss=0.07984, over 11665.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3019, pruned_loss=0.06979, over 3203856.40 frames. ], batch size: 246, lr: 9.81e-03, grad_scale: 8.0 2023-04-28 16:24:57,792 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65304.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:25:13,655 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0909, 2.3495, 1.9286, 2.2340, 2.7768, 2.4892, 3.0987, 3.0250], device='cuda:0'), covar=tensor([0.0044, 0.0193, 0.0267, 0.0217, 0.0102, 0.0181, 0.0085, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0170, 0.0171, 0.0170, 0.0165, 0.0171, 0.0163, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:25:39,696 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65332.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:25:59,704 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:07,814 INFO [train.py:904] (0/8) Epoch 7, batch 4450, loss[loss=0.222, simple_loss=0.3087, pruned_loss=0.06768, over 16259.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3051, pruned_loss=0.07045, over 3212428.60 frames. ], batch size: 165, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:26:08,113 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65352.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:17,482 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65359.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:26:58,771 INFO [optim.py:368] (0/8) 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 16:27:19,292 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65401.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:20,147 INFO [train.py:904] (0/8) Epoch 7, batch 4500, loss[loss=0.2311, simple_loss=0.308, pruned_loss=0.07708, over 16449.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3052, pruned_loss=0.07088, over 3209957.43 frames. ], batch size: 35, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:27:26,480 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:28,764 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65408.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:27:32,344 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65411.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:21,712 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4524, 3.4977, 3.2697, 3.0827, 3.0459, 3.3306, 3.2311, 3.0778], device='cuda:0'), covar=tensor([0.0411, 0.0253, 0.0195, 0.0181, 0.0522, 0.0237, 0.1162, 0.0365], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0229, 0.0237, 0.0211, 0.0268, 0.0235, 0.0167, 0.0267], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:28:27,212 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65449.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:31,286 INFO [train.py:904] (0/8) Epoch 7, batch 4550, loss[loss=0.2315, simple_loss=0.3142, pruned_loss=0.0744, over 16187.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3058, pruned_loss=0.0713, over 3212624.80 frames. ], batch size: 165, lr: 9.80e-03, grad_scale: 8.0 2023-04-28 16:28:37,648 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65456.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:28:43,521 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6739, 4.6069, 4.3533, 3.7283, 4.5171, 1.6133, 4.2708, 4.3085], device='cuda:0'), covar=tensor([0.0048, 0.0039, 0.0103, 0.0291, 0.0052, 0.1999, 0.0083, 0.0137], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0092, 0.0140, 0.0138, 0.0108, 0.0152, 0.0124, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:29:09,803 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9101, 4.8442, 5.4290, 5.3840, 5.4229, 4.9117, 4.9148, 4.4136], device='cuda:0'), covar=tensor([0.0208, 0.0272, 0.0226, 0.0301, 0.0286, 0.0240, 0.0610, 0.0369], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0269, 0.0268, 0.0263, 0.0317, 0.0286, 0.0384, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 16:29:21,415 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.690e+02 2.206e+02 2.596e+02 3.051e+02 6.081e+02, threshold=5.193e+02, percent-clipped=1.0 2023-04-28 16:29:41,307 INFO [train.py:904] (0/8) Epoch 7, batch 4600, loss[loss=0.2118, simple_loss=0.3002, pruned_loss=0.06166, over 16692.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.3062, pruned_loss=0.07062, over 3230024.92 frames. ], batch size: 134, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:30:24,213 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65532.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:30:50,583 INFO [train.py:904] (0/8) Epoch 7, batch 4650, loss[loss=0.2078, simple_loss=0.2931, pruned_loss=0.0613, over 16223.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3044, pruned_loss=0.06983, over 3233099.47 frames. ], batch size: 165, lr: 9.79e-03, grad_scale: 8.0 2023-04-28 16:31:41,970 INFO [optim.py:368] (0/8) 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 16:31:49,751 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65593.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:32:03,153 INFO [train.py:904] (0/8) Epoch 7, batch 4700, loss[loss=0.2098, simple_loss=0.2903, pruned_loss=0.06467, over 16575.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3021, pruned_loss=0.06866, over 3237926.18 frames. ], batch size: 62, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:32:45,764 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65632.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:32:56,274 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65641.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:33:12,633 INFO [train.py:904] (0/8) Epoch 7, batch 4750, loss[loss=0.1919, simple_loss=0.2687, pruned_loss=0.05759, over 16779.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.297, pruned_loss=0.06593, over 3256403.33 frames. ], batch size: 39, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:33:53,099 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:34:03,334 INFO [optim.py:368] (0/8) 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 16:34:22,796 INFO [train.py:904] (0/8) Epoch 7, batch 4800, loss[loss=0.2122, simple_loss=0.3016, pruned_loss=0.06137, over 16673.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2931, pruned_loss=0.06391, over 3254819.55 frames. ], batch size: 134, lr: 9.78e-03, grad_scale: 8.0 2023-04-28 16:34:36,407 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65711.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:35:36,927 INFO [train.py:904] (0/8) Epoch 7, batch 4850, loss[loss=0.205, simple_loss=0.2949, pruned_loss=0.05757, over 16219.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2944, pruned_loss=0.06445, over 3223067.18 frames. ], batch size: 165, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:35:45,500 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65759.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:36:04,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7245, 4.6717, 4.6762, 3.9453, 4.6259, 1.6735, 4.3921, 4.5559], device='cuda:0'), covar=tensor([0.0059, 0.0063, 0.0078, 0.0380, 0.0060, 0.1968, 0.0089, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0101, 0.0090, 0.0136, 0.0135, 0.0105, 0.0149, 0.0120, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:36:25,141 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 16:36:28,553 INFO [optim.py:368] (0/8) 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 16:36:48,895 INFO [train.py:904] (0/8) Epoch 7, batch 4900, loss[loss=0.2016, simple_loss=0.2883, pruned_loss=0.05743, over 16347.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.295, pruned_loss=0.06402, over 3192562.71 frames. ], batch size: 146, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:37:05,464 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 16:37:25,804 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2959, 3.9158, 3.6041, 1.9470, 2.9498, 2.4632, 3.5255, 3.7867], device='cuda:0'), covar=tensor([0.0170, 0.0397, 0.0454, 0.1530, 0.0710, 0.0802, 0.0540, 0.0626], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0134, 0.0155, 0.0140, 0.0133, 0.0125, 0.0137, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 16:37:43,356 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=65838.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:38:02,375 INFO [train.py:904] (0/8) Epoch 7, batch 4950, loss[loss=0.2346, simple_loss=0.315, pruned_loss=0.07713, over 12018.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2949, pruned_loss=0.06354, over 3192789.61 frames. ], batch size: 246, lr: 9.77e-03, grad_scale: 8.0 2023-04-28 16:38:28,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3743, 4.2850, 4.2442, 3.3729, 4.2207, 1.4496, 4.0445, 4.1113], device='cuda:0'), covar=tensor([0.0074, 0.0075, 0.0109, 0.0425, 0.0087, 0.2083, 0.0101, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0091, 0.0138, 0.0136, 0.0106, 0.0151, 0.0121, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:38:53,492 INFO [optim.py:368] (0/8) 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 16:38:53,901 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=65888.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 16:39:08,353 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=65899.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:39:12,695 INFO [train.py:904] (0/8) Epoch 7, batch 5000, loss[loss=0.2395, simple_loss=0.3313, pruned_loss=0.07384, over 16870.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2966, pruned_loss=0.06408, over 3197258.00 frames. ], batch size: 109, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:39:32,243 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2594, 3.5188, 3.8356, 1.5815, 3.9946, 4.0370, 2.8707, 2.8062], device='cuda:0'), covar=tensor([0.0786, 0.0144, 0.0099, 0.1255, 0.0045, 0.0048, 0.0344, 0.0440], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0096, 0.0082, 0.0139, 0.0070, 0.0085, 0.0118, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 16:39:40,850 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7194, 1.4877, 2.0132, 2.5547, 2.5895, 2.8681, 1.5535, 2.7905], device='cuda:0'), covar=tensor([0.0105, 0.0317, 0.0215, 0.0192, 0.0140, 0.0107, 0.0307, 0.0076], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0160, 0.0145, 0.0146, 0.0152, 0.0108, 0.0155, 0.0097], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 16:40:01,384 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 16:40:09,147 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=65941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:40:19,399 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9502, 4.8510, 5.3792, 5.3220, 5.3499, 4.9056, 4.8839, 4.6046], device='cuda:0'), covar=tensor([0.0217, 0.0449, 0.0253, 0.0373, 0.0361, 0.0242, 0.0751, 0.0328], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0265, 0.0265, 0.0262, 0.0316, 0.0286, 0.0383, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 16:40:23,716 INFO [train.py:904] (0/8) Epoch 7, batch 5050, loss[loss=0.2214, simple_loss=0.3165, pruned_loss=0.06318, over 16945.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.297, pruned_loss=0.06404, over 3193668.74 frames. ], batch size: 96, lr: 9.76e-03, grad_scale: 8.0 2023-04-28 16:40:36,531 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6469, 4.0601, 4.4023, 2.0288, 4.6255, 4.5987, 3.0852, 3.3897], device='cuda:0'), covar=tensor([0.0718, 0.0117, 0.0092, 0.1061, 0.0035, 0.0032, 0.0326, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0096, 0.0081, 0.0139, 0.0070, 0.0085, 0.0119, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 16:41:14,580 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.911e+02 2.609e+02 2.967e+02 3.441e+02 6.265e+02, threshold=5.933e+02, percent-clipped=0.0 2023-04-28 16:41:16,134 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=65989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:41:20,679 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2681, 3.3211, 3.5949, 3.5653, 3.5632, 3.3694, 3.3691, 3.4426], device='cuda:0'), covar=tensor([0.0306, 0.0511, 0.0349, 0.0420, 0.0424, 0.0351, 0.0764, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0266, 0.0265, 0.0264, 0.0318, 0.0287, 0.0384, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 16:41:29,851 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-66000.pt 2023-04-28 16:41:34,777 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-28 16:41:35,933 INFO [train.py:904] (0/8) Epoch 7, batch 5100, loss[loss=0.2023, simple_loss=0.2892, pruned_loss=0.05771, over 16741.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2953, pruned_loss=0.06335, over 3186688.31 frames. ], batch size: 124, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:41:52,019 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6007, 4.3355, 4.6514, 4.8055, 4.9451, 4.5033, 4.9301, 4.8914], device='cuda:0'), covar=tensor([0.1086, 0.0887, 0.1086, 0.0456, 0.0346, 0.0630, 0.0378, 0.0491], device='cuda:0'), in_proj_covar=tensor([0.0441, 0.0538, 0.0673, 0.0546, 0.0408, 0.0413, 0.0416, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:42:06,686 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5112, 3.9809, 4.3153, 1.9375, 4.5546, 4.4974, 3.2056, 3.1595], device='cuda:0'), covar=tensor([0.0774, 0.0138, 0.0104, 0.1165, 0.0034, 0.0039, 0.0277, 0.0437], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0096, 0.0081, 0.0139, 0.0070, 0.0085, 0.0119, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 16:42:46,404 INFO [train.py:904] (0/8) Epoch 7, batch 5150, loss[loss=0.1895, simple_loss=0.2787, pruned_loss=0.05014, over 17068.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2958, pruned_loss=0.06289, over 3181692.03 frames. ], batch size: 53, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:43:37,177 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 7, batch 5200, loss[loss=0.1949, simple_loss=0.2796, pruned_loss=0.05513, over 16390.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2948, pruned_loss=0.06271, over 3181574.30 frames. ], batch size: 146, lr: 9.75e-03, grad_scale: 8.0 2023-04-28 16:44:56,221 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:45:06,330 INFO [train.py:904] (0/8) Epoch 7, batch 5250, loss[loss=0.2413, simple_loss=0.3073, pruned_loss=0.08765, over 12168.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2915, pruned_loss=0.06218, over 3201441.02 frames. ], batch size: 246, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:45:19,001 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66160.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:45:39,760 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0867, 3.2615, 3.3726, 1.5802, 3.5746, 3.5779, 2.7732, 2.5720], device='cuda:0'), covar=tensor([0.0798, 0.0147, 0.0122, 0.1167, 0.0050, 0.0067, 0.0362, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0096, 0.0081, 0.0139, 0.0070, 0.0084, 0.0119, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 16:45:59,717 INFO [optim.py:368] (0/8) 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 16:46:00,038 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66188.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:46:06,946 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66194.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:46:18,395 INFO [train.py:904] (0/8) Epoch 7, batch 5300, loss[loss=0.2012, simple_loss=0.2794, pruned_loss=0.06152, over 16452.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2882, pruned_loss=0.06086, over 3202234.30 frames. ], batch size: 68, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:46:23,393 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66205.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:46:45,725 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66221.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:47:04,762 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0616, 3.5332, 3.2419, 1.8483, 2.9946, 2.3587, 3.3973, 3.4415], device='cuda:0'), covar=tensor([0.0202, 0.0521, 0.0583, 0.1586, 0.0658, 0.0849, 0.0564, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0132, 0.0156, 0.0141, 0.0133, 0.0124, 0.0138, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 16:47:06,143 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66236.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:47:22,732 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0497, 3.3332, 3.6114, 3.5765, 3.5198, 3.3084, 3.3339, 3.3815], device='cuda:0'), covar=tensor([0.0414, 0.0545, 0.0326, 0.0372, 0.0486, 0.0389, 0.0771, 0.0429], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0274, 0.0276, 0.0269, 0.0327, 0.0297, 0.0398, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 16:47:25,958 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-28 16:47:28,954 INFO [train.py:904] (0/8) Epoch 7, batch 5350, loss[loss=0.2235, simple_loss=0.3181, pruned_loss=0.06446, over 16175.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2864, pruned_loss=0.05969, over 3211323.76 frames. ], batch size: 165, lr: 9.74e-03, grad_scale: 8.0 2023-04-28 16:47:36,898 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66258.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:47:38,451 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-28 16:48:20,490 INFO [optim.py:368] (0/8) 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 16:48:40,532 INFO [train.py:904] (0/8) Epoch 7, batch 5400, loss[loss=0.2367, simple_loss=0.3155, pruned_loss=0.07891, over 16579.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2894, pruned_loss=0.06105, over 3186502.52 frames. ], batch size: 57, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:49:00,402 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6403, 2.6660, 1.7075, 2.7515, 2.0603, 2.7150, 1.9568, 2.3775], device='cuda:0'), covar=tensor([0.0200, 0.0356, 0.1220, 0.0097, 0.0669, 0.0458, 0.1076, 0.0527], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0158, 0.0180, 0.0094, 0.0165, 0.0194, 0.0191, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 16:49:05,301 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66319.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:49:31,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8832, 2.1430, 2.4215, 4.5131, 1.9354, 2.9275, 2.3516, 2.5218], device='cuda:0'), covar=tensor([0.0645, 0.2640, 0.1560, 0.0296, 0.3323, 0.1587, 0.2355, 0.2502], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0349, 0.0294, 0.0319, 0.0388, 0.0382, 0.0314, 0.0414], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:49:56,008 INFO [train.py:904] (0/8) Epoch 7, batch 5450, loss[loss=0.2597, simple_loss=0.3175, pruned_loss=0.101, over 11741.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.293, pruned_loss=0.06296, over 3177841.93 frames. ], batch size: 248, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:50:49,089 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6685, 2.5858, 2.3498, 3.8259, 2.6565, 3.8663, 1.2937, 2.7595], device='cuda:0'), covar=tensor([0.1377, 0.0659, 0.1189, 0.0168, 0.0290, 0.0357, 0.1623, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0148, 0.0170, 0.0104, 0.0197, 0.0195, 0.0167, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 16:50:50,983 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 3.043e+02 3.833e+02 5.336e+02 9.922e+02, threshold=7.666e+02, percent-clipped=19.0 2023-04-28 16:51:12,562 INFO [train.py:904] (0/8) Epoch 7, batch 5500, loss[loss=0.2597, simple_loss=0.3409, pruned_loss=0.08927, over 16902.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3017, pruned_loss=0.06856, over 3158080.93 frames. ], batch size: 96, lr: 9.73e-03, grad_scale: 8.0 2023-04-28 16:52:29,860 INFO [train.py:904] (0/8) Epoch 7, batch 5550, loss[loss=0.3288, simple_loss=0.3769, pruned_loss=0.1403, over 11400.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3105, pruned_loss=0.07616, over 3098663.45 frames. ], batch size: 248, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:53:27,054 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.385e+02 3.933e+02 4.760e+02 6.147e+02 1.111e+03, threshold=9.520e+02, percent-clipped=8.0 2023-04-28 16:53:35,772 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66494.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:53:47,087 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66500.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:53:49,160 INFO [train.py:904] (0/8) Epoch 7, batch 5600, loss[loss=0.2304, simple_loss=0.3169, pruned_loss=0.07192, over 17121.00 frames. ], tot_loss[loss=0.2405, simple_loss=0.3172, pruned_loss=0.0819, over 3058753.12 frames. ], batch size: 48, lr: 9.72e-03, grad_scale: 8.0 2023-04-28 16:54:11,836 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66516.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 16:54:20,958 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66521.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:54:53,081 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4575, 2.4565, 1.9559, 2.2145, 2.8440, 2.5362, 3.2799, 3.1570], device='cuda:0'), covar=tensor([0.0039, 0.0251, 0.0307, 0.0265, 0.0143, 0.0222, 0.0113, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0170, 0.0170, 0.0168, 0.0166, 0.0173, 0.0162, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:54:54,204 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66542.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 16:55:08,336 INFO [train.py:904] (0/8) Epoch 7, batch 5650, loss[loss=0.2351, simple_loss=0.3098, pruned_loss=0.08023, over 16781.00 frames. ], tot_loss[loss=0.2481, simple_loss=0.3225, pruned_loss=0.08688, over 3054758.59 frames. ], batch size: 83, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:55:53,643 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66582.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 16:56:03,091 INFO [optim.py:368] (0/8) 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 16:56:23,913 INFO [train.py:904] (0/8) Epoch 7, batch 5700, loss[loss=0.3144, simple_loss=0.3594, pruned_loss=0.1347, over 11259.00 frames. ], tot_loss[loss=0.2505, simple_loss=0.3244, pruned_loss=0.08836, over 3051752.20 frames. ], batch size: 248, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:56:42,496 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66614.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 16:57:24,625 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0293, 2.1826, 2.3119, 2.8757, 2.3546, 3.2006, 1.6968, 2.7581], device='cuda:0'), covar=tensor([0.1088, 0.0534, 0.0933, 0.0128, 0.0225, 0.0389, 0.1259, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0146, 0.0169, 0.0103, 0.0196, 0.0194, 0.0166, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 16:57:29,137 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4663, 3.9006, 4.1525, 1.8794, 4.3413, 4.3789, 3.1617, 3.2911], device='cuda:0'), covar=tensor([0.0831, 0.0143, 0.0144, 0.1174, 0.0048, 0.0057, 0.0302, 0.0386], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0094, 0.0080, 0.0138, 0.0070, 0.0085, 0.0117, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 16:57:41,031 INFO [train.py:904] (0/8) Epoch 7, batch 5750, loss[loss=0.2208, simple_loss=0.3088, pruned_loss=0.06639, over 16737.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.327, pruned_loss=0.0898, over 3041225.67 frames. ], batch size: 83, lr: 9.71e-03, grad_scale: 8.0 2023-04-28 16:58:35,267 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7069, 2.2104, 2.2410, 4.5535, 2.0086, 2.9950, 2.3382, 2.4628], device='cuda:0'), covar=tensor([0.0692, 0.2706, 0.1659, 0.0247, 0.3313, 0.1445, 0.2432, 0.2615], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0351, 0.0295, 0.0321, 0.0390, 0.0381, 0.0315, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:58:39,458 INFO [optim.py:368] (0/8) 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 16:59:02,140 INFO [train.py:904] (0/8) Epoch 7, batch 5800, loss[loss=0.2906, simple_loss=0.3638, pruned_loss=0.1087, over 15385.00 frames. ], tot_loss[loss=0.2533, simple_loss=0.3277, pruned_loss=0.08948, over 3027400.48 frames. ], batch size: 191, lr: 9.70e-03, grad_scale: 16.0 2023-04-28 16:59:23,072 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9080, 5.2318, 4.9125, 4.9702, 4.6190, 4.5284, 4.6218, 5.3132], device='cuda:0'), covar=tensor([0.0839, 0.0666, 0.1007, 0.0536, 0.0702, 0.0785, 0.0854, 0.0708], device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0550, 0.0473, 0.0361, 0.0346, 0.0372, 0.0458, 0.0405], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 16:59:48,394 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66731.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:00:19,935 INFO [train.py:904] (0/8) Epoch 7, batch 5850, loss[loss=0.2288, simple_loss=0.3076, pruned_loss=0.07499, over 16611.00 frames. ], tot_loss[loss=0.2487, simple_loss=0.3244, pruned_loss=0.08648, over 3056159.92 frames. ], batch size: 62, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:01:20,004 INFO [optim.py:368] (0/8) 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 17:01:25,729 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=66792.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 17:01:39,037 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66800.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:01:41,108 INFO [train.py:904] (0/8) Epoch 7, batch 5900, loss[loss=0.2254, simple_loss=0.3155, pruned_loss=0.06764, over 16996.00 frames. ], tot_loss[loss=0.2477, simple_loss=0.3235, pruned_loss=0.08594, over 3050458.87 frames. ], batch size: 41, lr: 9.70e-03, grad_scale: 8.0 2023-04-28 17:02:07,251 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66816.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:02:40,556 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2359, 3.5633, 3.6901, 1.5625, 3.8649, 3.8973, 3.0247, 2.8456], device='cuda:0'), covar=tensor([0.0791, 0.0146, 0.0159, 0.1279, 0.0061, 0.0102, 0.0303, 0.0402], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0097, 0.0082, 0.0140, 0.0072, 0.0088, 0.0119, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 17:02:51,277 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9482, 3.9934, 2.1334, 4.5321, 2.7557, 4.4289, 2.2778, 3.0247], device='cuda:0'), covar=tensor([0.0136, 0.0268, 0.1450, 0.0055, 0.0671, 0.0291, 0.1333, 0.0591], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0157, 0.0179, 0.0095, 0.0163, 0.0192, 0.0188, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 17:02:56,756 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66848.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:03:02,864 INFO [train.py:904] (0/8) Epoch 7, batch 5950, loss[loss=0.2274, simple_loss=0.3137, pruned_loss=0.07053, over 16745.00 frames. ], tot_loss[loss=0.2455, simple_loss=0.3233, pruned_loss=0.08379, over 3076508.95 frames. ], batch size: 89, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:03:22,516 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66864.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:03:41,707 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=66877.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:03:45,566 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 17:04:00,341 INFO [optim.py:368] (0/8) 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 17:04:22,266 INFO [train.py:904] (0/8) Epoch 7, batch 6000, loss[loss=0.2063, simple_loss=0.2861, pruned_loss=0.06328, over 16478.00 frames. ], tot_loss[loss=0.244, simple_loss=0.3222, pruned_loss=0.08294, over 3091725.88 frames. ], batch size: 68, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:04:22,267 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 17:04:32,875 INFO [train.py:938] (0/8) Epoch 7, validation: loss=0.1758, simple_loss=0.2891, pruned_loss=0.03127, over 944034.00 frames. 2023-04-28 17:04:32,876 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 17:04:40,066 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-28 17:04:51,368 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=66914.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:05:53,515 INFO [train.py:904] (0/8) Epoch 7, batch 6050, loss[loss=0.2738, simple_loss=0.3272, pruned_loss=0.1102, over 11289.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3199, pruned_loss=0.08171, over 3091572.91 frames. ], batch size: 246, lr: 9.69e-03, grad_scale: 8.0 2023-04-28 17:06:02,063 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=66957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:09,560 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=66962.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:06:36,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8236, 3.2972, 2.7847, 4.6541, 3.6065, 4.3127, 1.6934, 2.9270], device='cuda:0'), covar=tensor([0.1336, 0.0523, 0.0952, 0.0105, 0.0363, 0.0301, 0.1389, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0148, 0.0170, 0.0104, 0.0198, 0.0196, 0.0167, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 17:06:46,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9714, 3.2120, 3.3367, 1.6071, 3.4757, 3.5301, 2.7992, 2.6400], device='cuda:0'), covar=tensor([0.0875, 0.0163, 0.0133, 0.1180, 0.0068, 0.0104, 0.0357, 0.0459], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0096, 0.0081, 0.0139, 0.0071, 0.0086, 0.0118, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 17:06:50,897 INFO [optim.py:368] (0/8) 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 17:07:12,279 INFO [train.py:904] (0/8) Epoch 7, batch 6100, loss[loss=0.2448, simple_loss=0.324, pruned_loss=0.08287, over 16793.00 frames. ], tot_loss[loss=0.2392, simple_loss=0.3185, pruned_loss=0.0799, over 3106729.52 frames. ], batch size: 62, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:07:39,920 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67018.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:08:32,448 INFO [train.py:904] (0/8) Epoch 7, batch 6150, loss[loss=0.2273, simple_loss=0.3083, pruned_loss=0.07317, over 15426.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3168, pruned_loss=0.07942, over 3103209.91 frames. ], batch size: 191, lr: 9.68e-03, grad_scale: 8.0 2023-04-28 17:09:23,270 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67083.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:09:29,199 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67087.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:09:30,669 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3597, 3.2629, 3.3557, 3.4588, 3.4855, 3.2415, 3.4823, 3.5385], device='cuda:0'), covar=tensor([0.0906, 0.0701, 0.1088, 0.0486, 0.0593, 0.1792, 0.0726, 0.0572], device='cuda:0'), in_proj_covar=tensor([0.0438, 0.0535, 0.0671, 0.0538, 0.0413, 0.0409, 0.0421, 0.0469], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:09:31,360 INFO [optim.py:368] (0/8) 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 17:09:51,776 INFO [train.py:904] (0/8) Epoch 7, batch 6200, loss[loss=0.2721, simple_loss=0.3472, pruned_loss=0.09847, over 16743.00 frames. ], tot_loss[loss=0.2365, simple_loss=0.3152, pruned_loss=0.07894, over 3114207.97 frames. ], batch size: 134, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:02,473 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67144.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:11:14,132 INFO [train.py:904] (0/8) Epoch 7, batch 6250, loss[loss=0.2096, simple_loss=0.293, pruned_loss=0.06314, over 17096.00 frames. ], tot_loss[loss=0.2377, simple_loss=0.3158, pruned_loss=0.07986, over 3088146.15 frames. ], batch size: 49, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:11:52,570 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67177.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:12:06,836 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-04-28 17:12:11,782 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.341e+02 3.226e+02 4.153e+02 4.983e+02 8.969e+02, threshold=8.306e+02, percent-clipped=3.0 2023-04-28 17:12:30,908 INFO [train.py:904] (0/8) Epoch 7, batch 6300, loss[loss=0.2525, simple_loss=0.3288, pruned_loss=0.08813, over 16924.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3155, pruned_loss=0.07896, over 3097674.73 frames. ], batch size: 109, lr: 9.67e-03, grad_scale: 4.0 2023-04-28 17:13:03,888 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4355, 3.4062, 3.3752, 2.7863, 3.3228, 2.1791, 3.0570, 2.7601], device='cuda:0'), covar=tensor([0.0105, 0.0084, 0.0117, 0.0195, 0.0065, 0.1490, 0.0094, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0092, 0.0141, 0.0138, 0.0106, 0.0156, 0.0124, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:13:08,176 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67225.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:13:49,878 INFO [train.py:904] (0/8) Epoch 7, batch 6350, loss[loss=0.2368, simple_loss=0.3107, pruned_loss=0.08145, over 16651.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3165, pruned_loss=0.08066, over 3089593.72 frames. ], batch size: 76, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:14:47,876 INFO [optim.py:368] (0/8) 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 17:15:05,427 INFO [train.py:904] (0/8) Epoch 7, batch 6400, loss[loss=0.2128, simple_loss=0.2927, pruned_loss=0.06644, over 16707.00 frames. ], tot_loss[loss=0.2408, simple_loss=0.3174, pruned_loss=0.08214, over 3073190.55 frames. ], batch size: 83, lr: 9.66e-03, grad_scale: 8.0 2023-04-28 17:15:22,347 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67313.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:15:44,859 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7867, 4.0742, 3.3719, 2.2653, 2.9335, 2.5206, 4.2631, 3.8771], device='cuda:0'), covar=tensor([0.2141, 0.0548, 0.1106, 0.1981, 0.1860, 0.1418, 0.0353, 0.0661], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0245, 0.0268, 0.0255, 0.0275, 0.0205, 0.0246, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:16:20,299 INFO [train.py:904] (0/8) Epoch 7, batch 6450, loss[loss=0.222, simple_loss=0.3109, pruned_loss=0.06659, over 16855.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3171, pruned_loss=0.08106, over 3082213.76 frames. ], batch size: 96, lr: 9.66e-03, grad_scale: 4.0 2023-04-28 17:16:24,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0858, 4.1853, 4.2397, 4.2275, 4.2445, 4.7535, 4.4868, 4.1958], device='cuda:0'), covar=tensor([0.1659, 0.1621, 0.1668, 0.1789, 0.2567, 0.1134, 0.1080, 0.2134], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0425, 0.0434, 0.0367, 0.0491, 0.0469, 0.0345, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 17:16:55,649 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.82 vs. limit=5.0 2023-04-28 17:17:01,277 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 17:17:16,632 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67387.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:17:22,058 INFO [optim.py:368] (0/8) 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 17:17:38,450 INFO [train.py:904] (0/8) Epoch 7, batch 6500, loss[loss=0.2516, simple_loss=0.3182, pruned_loss=0.09252, over 15378.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3143, pruned_loss=0.07945, over 3096911.37 frames. ], batch size: 191, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:18:29,448 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67435.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:18:36,138 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67439.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:18:57,756 INFO [train.py:904] (0/8) Epoch 7, batch 6550, loss[loss=0.2499, simple_loss=0.3304, pruned_loss=0.08471, over 16323.00 frames. ], tot_loss[loss=0.239, simple_loss=0.3174, pruned_loss=0.08028, over 3115799.82 frames. ], batch size: 35, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:19:56,004 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.411e+02 3.629e+02 4.533e+02 5.660e+02 1.346e+03, threshold=9.066e+02, percent-clipped=11.0 2023-04-28 17:20:13,101 INFO [train.py:904] (0/8) Epoch 7, batch 6600, loss[loss=0.2729, simple_loss=0.344, pruned_loss=0.1009, over 16714.00 frames. ], tot_loss[loss=0.2415, simple_loss=0.3201, pruned_loss=0.0815, over 3103789.33 frames. ], batch size: 134, lr: 9.65e-03, grad_scale: 4.0 2023-04-28 17:20:30,803 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67513.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:21:26,349 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1563, 3.2668, 3.2702, 5.1904, 4.4337, 4.7247, 2.2644, 3.6834], device='cuda:0'), covar=tensor([0.1096, 0.0556, 0.0812, 0.0080, 0.0316, 0.0270, 0.1098, 0.0578], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0148, 0.0169, 0.0104, 0.0198, 0.0197, 0.0167, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 17:21:29,935 INFO [train.py:904] (0/8) Epoch 7, batch 6650, loss[loss=0.2569, simple_loss=0.3296, pruned_loss=0.09213, over 15376.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3197, pruned_loss=0.08149, over 3122744.56 frames. ], batch size: 191, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:22:02,868 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67574.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:22:18,024 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4570, 2.0179, 2.2137, 4.2206, 1.9795, 2.6516, 2.1773, 2.1941], device='cuda:0'), covar=tensor([0.0751, 0.2828, 0.1643, 0.0292, 0.3405, 0.1696, 0.2475, 0.2614], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0352, 0.0294, 0.0319, 0.0391, 0.0380, 0.0316, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:22:27,321 INFO [optim.py:368] (0/8) 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 17:22:29,292 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7461, 1.5690, 2.1378, 2.6657, 2.5685, 3.0133, 1.7113, 2.8919], device='cuda:0'), covar=tensor([0.0095, 0.0289, 0.0170, 0.0136, 0.0145, 0.0079, 0.0303, 0.0054], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0155, 0.0137, 0.0138, 0.0148, 0.0103, 0.0157, 0.0094], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 17:22:43,253 INFO [train.py:904] (0/8) Epoch 7, batch 6700, loss[loss=0.276, simple_loss=0.3279, pruned_loss=0.112, over 11535.00 frames. ], tot_loss[loss=0.2419, simple_loss=0.3193, pruned_loss=0.08229, over 3093023.32 frames. ], batch size: 247, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:22:59,674 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:23:16,359 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5891, 2.0863, 1.5597, 1.9675, 2.5368, 2.2763, 2.6462, 2.7376], device='cuda:0'), covar=tensor([0.0070, 0.0244, 0.0345, 0.0313, 0.0142, 0.0224, 0.0110, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0174, 0.0174, 0.0172, 0.0168, 0.0175, 0.0165, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:23:56,596 INFO [train.py:904] (0/8) Epoch 7, batch 6750, loss[loss=0.2079, simple_loss=0.2899, pruned_loss=0.0629, over 16782.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3176, pruned_loss=0.08188, over 3085692.61 frames. ], batch size: 83, lr: 9.64e-03, grad_scale: 4.0 2023-04-28 17:24:04,400 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0035, 5.3784, 5.0788, 5.1313, 4.7580, 4.6438, 4.8027, 5.4548], device='cuda:0'), covar=tensor([0.0833, 0.0647, 0.0844, 0.0513, 0.0605, 0.0758, 0.0794, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0446, 0.0561, 0.0481, 0.0369, 0.0355, 0.0378, 0.0467, 0.0417], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:24:10,416 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67661.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:37,892 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67680.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:24:54,246 INFO [optim.py:368] (0/8) 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 17:24:55,934 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 7, batch 6800, loss[loss=0.2145, simple_loss=0.3058, pruned_loss=0.06159, over 16804.00 frames. ], tot_loss[loss=0.2412, simple_loss=0.3182, pruned_loss=0.08215, over 3080973.75 frames. ], batch size: 83, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:25:34,861 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9146, 3.3156, 3.3558, 1.9675, 3.0102, 3.2225, 3.1095, 1.7265], device='cuda:0'), covar=tensor([0.0371, 0.0026, 0.0029, 0.0307, 0.0060, 0.0078, 0.0046, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0056, 0.0060, 0.0118, 0.0065, 0.0077, 0.0067, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 17:26:06,219 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=67739.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:26:10,126 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67741.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:26:23,542 INFO [train.py:904] (0/8) Epoch 7, batch 6850, loss[loss=0.2534, simple_loss=0.3209, pruned_loss=0.0929, over 11736.00 frames. ], tot_loss[loss=0.241, simple_loss=0.3183, pruned_loss=0.08191, over 3098356.60 frames. ], batch size: 247, lr: 9.63e-03, grad_scale: 8.0 2023-04-28 17:26:25,739 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=67753.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:27:13,729 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=67787.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:27:20,978 INFO [optim.py:368] (0/8) 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 17:27:35,620 INFO [train.py:904] (0/8) Epoch 7, batch 6900, loss[loss=0.2845, simple_loss=0.3551, pruned_loss=0.1069, over 16230.00 frames. ], tot_loss[loss=0.2427, simple_loss=0.3214, pruned_loss=0.08194, over 3108842.30 frames. ], batch size: 165, lr: 9.63e-03, grad_scale: 2.0 2023-04-28 17:28:44,705 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6686, 3.8864, 4.0960, 1.6854, 4.2787, 4.3213, 3.1056, 3.0075], device='cuda:0'), covar=tensor([0.0654, 0.0129, 0.0133, 0.1207, 0.0052, 0.0067, 0.0313, 0.0416], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0095, 0.0083, 0.0140, 0.0071, 0.0086, 0.0120, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 17:28:46,556 INFO [train.py:904] (0/8) Epoch 7, batch 6950, loss[loss=0.2353, simple_loss=0.3113, pruned_loss=0.07964, over 16697.00 frames. ], tot_loss[loss=0.2451, simple_loss=0.3228, pruned_loss=0.08368, over 3098409.71 frames. ], batch size: 134, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:29:13,151 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=67869.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:29:46,384 INFO [optim.py:368] (0/8) 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 17:29:59,773 INFO [train.py:904] (0/8) Epoch 7, batch 7000, loss[loss=0.2222, simple_loss=0.3147, pruned_loss=0.06479, over 16674.00 frames. ], tot_loss[loss=0.243, simple_loss=0.3221, pruned_loss=0.082, over 3108500.59 frames. ], batch size: 134, lr: 9.62e-03, grad_scale: 2.0 2023-04-28 17:30:48,734 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8977, 3.8725, 4.3599, 4.3312, 4.3231, 4.0015, 4.0361, 3.9432], device='cuda:0'), covar=tensor([0.0284, 0.0473, 0.0325, 0.0383, 0.0423, 0.0343, 0.0790, 0.0469], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0269, 0.0271, 0.0267, 0.0320, 0.0291, 0.0386, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 17:31:13,270 INFO [train.py:904] (0/8) Epoch 7, batch 7050, loss[loss=0.2675, simple_loss=0.3366, pruned_loss=0.09919, over 16655.00 frames. ], tot_loss[loss=0.2435, simple_loss=0.3229, pruned_loss=0.08207, over 3092849.47 frames. ], batch size: 57, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:31:25,944 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 17:31:57,412 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=67981.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:32:14,208 INFO [optim.py:368] (0/8) 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 17:32:23,793 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-68000.pt 2023-04-28 17:32:29,899 INFO [train.py:904] (0/8) Epoch 7, batch 7100, loss[loss=0.2326, simple_loss=0.3155, pruned_loss=0.0748, over 16467.00 frames. ], tot_loss[loss=0.2434, simple_loss=0.3217, pruned_loss=0.08253, over 3066066.98 frames. ], batch size: 35, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:33:20,756 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68036.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:29,668 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68042.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:33:35,255 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6124, 4.4380, 4.6442, 4.8428, 4.9399, 4.5306, 4.9452, 4.9477], device='cuda:0'), covar=tensor([0.1213, 0.0902, 0.1262, 0.0486, 0.0445, 0.0614, 0.0456, 0.0427], device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0533, 0.0666, 0.0533, 0.0406, 0.0403, 0.0423, 0.0463], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:33:36,401 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3609, 3.9656, 3.7360, 1.9444, 2.9343, 2.6716, 3.6320, 4.0358], device='cuda:0'), covar=tensor([0.0269, 0.0510, 0.0512, 0.1708, 0.0797, 0.0778, 0.0641, 0.0743], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0130, 0.0155, 0.0141, 0.0135, 0.0125, 0.0137, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 17:33:39,224 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68048.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:33:44,541 INFO [train.py:904] (0/8) Epoch 7, batch 7150, loss[loss=0.2353, simple_loss=0.3179, pruned_loss=0.07634, over 16330.00 frames. ], tot_loss[loss=0.2422, simple_loss=0.3201, pruned_loss=0.08221, over 3064461.20 frames. ], batch size: 146, lr: 9.61e-03, grad_scale: 2.0 2023-04-28 17:34:36,486 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68087.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:34:38,181 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8508, 3.8564, 3.8410, 2.7681, 3.7921, 1.4895, 3.5243, 3.3139], device='cuda:0'), covar=tensor([0.0133, 0.0108, 0.0163, 0.0547, 0.0114, 0.2924, 0.0166, 0.0311], device='cuda:0'), in_proj_covar=tensor([0.0103, 0.0090, 0.0138, 0.0134, 0.0105, 0.0155, 0.0122, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:34:44,523 INFO [optim.py:368] (0/8) 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 17:34:51,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7396, 3.7308, 3.8347, 3.7347, 3.7504, 4.1948, 3.9042, 3.6530], device='cuda:0'), covar=tensor([0.2073, 0.1888, 0.1654, 0.2275, 0.2846, 0.1506, 0.1203, 0.2385], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0432, 0.0442, 0.0370, 0.0495, 0.0471, 0.0356, 0.0508], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 17:34:58,214 INFO [train.py:904] (0/8) Epoch 7, batch 7200, loss[loss=0.2103, simple_loss=0.3071, pruned_loss=0.05672, over 16786.00 frames. ], tot_loss[loss=0.2397, simple_loss=0.318, pruned_loss=0.08074, over 3051178.72 frames. ], batch size: 124, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:35:14,327 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0109, 1.7294, 2.4162, 2.8876, 2.5843, 3.2764, 1.9231, 3.1546], device='cuda:0'), covar=tensor([0.0110, 0.0313, 0.0201, 0.0154, 0.0169, 0.0081, 0.0312, 0.0063], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0156, 0.0136, 0.0138, 0.0147, 0.0102, 0.0154, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 17:35:18,475 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3530, 5.3476, 5.2881, 5.1355, 4.5505, 5.2723, 5.2568, 4.9825], device='cuda:0'), covar=tensor([0.0525, 0.0358, 0.0245, 0.0175, 0.1136, 0.0363, 0.0201, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0233, 0.0233, 0.0206, 0.0261, 0.0237, 0.0165, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:35:23,079 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-28 17:36:11,195 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68148.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:36:16,048 INFO [train.py:904] (0/8) Epoch 7, batch 7250, loss[loss=0.2263, simple_loss=0.2989, pruned_loss=0.07686, over 15389.00 frames. ], tot_loss[loss=0.2373, simple_loss=0.3156, pruned_loss=0.07948, over 3059904.12 frames. ], batch size: 190, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:36:39,820 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3595, 4.1324, 4.3481, 4.5323, 4.6958, 4.2636, 4.6495, 4.6721], device='cuda:0'), covar=tensor([0.1168, 0.1009, 0.1275, 0.0567, 0.0454, 0.0808, 0.0490, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0440, 0.0540, 0.0676, 0.0543, 0.0413, 0.0414, 0.0432, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:36:41,065 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68169.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:36:48,709 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-28 17:37:09,188 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9813, 4.8948, 4.7572, 3.6210, 4.8013, 1.6381, 4.5787, 4.6828], device='cuda:0'), covar=tensor([0.0090, 0.0067, 0.0137, 0.0472, 0.0077, 0.2436, 0.0110, 0.0165], device='cuda:0'), in_proj_covar=tensor([0.0104, 0.0091, 0.0140, 0.0135, 0.0106, 0.0156, 0.0123, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:37:16,281 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 7, batch 7300, loss[loss=0.2086, simple_loss=0.3028, pruned_loss=0.05723, over 16614.00 frames. ], tot_loss[loss=0.2352, simple_loss=0.314, pruned_loss=0.07818, over 3082942.59 frames. ], batch size: 62, lr: 9.60e-03, grad_scale: 4.0 2023-04-28 17:37:52,949 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68217.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:37:57,578 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7292, 1.7095, 1.4770, 1.4873, 1.7834, 1.4732, 1.7073, 1.8582], device='cuda:0'), covar=tensor([0.0064, 0.0136, 0.0208, 0.0184, 0.0105, 0.0152, 0.0096, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0171, 0.0172, 0.0170, 0.0166, 0.0174, 0.0163, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:38:44,626 INFO [train.py:904] (0/8) Epoch 7, batch 7350, loss[loss=0.222, simple_loss=0.3023, pruned_loss=0.07085, over 16409.00 frames. ], tot_loss[loss=0.235, simple_loss=0.314, pruned_loss=0.078, over 3086356.14 frames. ], batch size: 146, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:39:48,572 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.908e+02 3.085e+02 3.772e+02 4.630e+02 1.239e+03, threshold=7.544e+02, percent-clipped=2.0 2023-04-28 17:40:02,139 INFO [train.py:904] (0/8) Epoch 7, batch 7400, loss[loss=0.2237, simple_loss=0.3079, pruned_loss=0.0697, over 16706.00 frames. ], tot_loss[loss=0.2366, simple_loss=0.3157, pruned_loss=0.0787, over 3097455.06 frames. ], batch size: 134, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:40:53,646 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68336.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:40:54,929 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68337.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:12,135 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9391, 3.8136, 4.3752, 4.3277, 4.3307, 3.9963, 4.0554, 3.9083], device='cuda:0'), covar=tensor([0.0287, 0.0530, 0.0337, 0.0404, 0.0431, 0.0346, 0.0839, 0.0421], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0272, 0.0274, 0.0267, 0.0320, 0.0295, 0.0390, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 17:41:13,440 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:41:19,099 INFO [train.py:904] (0/8) Epoch 7, batch 7450, loss[loss=0.2085, simple_loss=0.2957, pruned_loss=0.06069, over 17102.00 frames. ], tot_loss[loss=0.2386, simple_loss=0.3173, pruned_loss=0.07997, over 3108174.45 frames. ], batch size: 49, lr: 9.59e-03, grad_scale: 4.0 2023-04-28 17:41:35,930 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1458, 4.1223, 4.2965, 4.1670, 4.2570, 4.7677, 4.3282, 4.0722], device='cuda:0'), covar=tensor([0.1323, 0.1887, 0.1727, 0.1852, 0.2442, 0.0960, 0.1400, 0.2435], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0426, 0.0442, 0.0370, 0.0491, 0.0469, 0.0353, 0.0503], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 17:42:02,829 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1115, 4.1186, 3.9580, 3.7772, 3.6064, 3.9686, 3.8047, 3.7306], device='cuda:0'), covar=tensor([0.0517, 0.0463, 0.0261, 0.0257, 0.0799, 0.0416, 0.0700, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0232, 0.0234, 0.0207, 0.0261, 0.0236, 0.0164, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:42:11,338 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.512e+02 3.665e+02 4.367e+02 5.646e+02 8.920e+02, threshold=8.734e+02, percent-clipped=5.0 2023-04-28 17:42:30,795 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68396.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:35,188 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:42:39,846 INFO [train.py:904] (0/8) Epoch 7, batch 7500, loss[loss=0.2504, simple_loss=0.3275, pruned_loss=0.08668, over 15387.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3181, pruned_loss=0.08059, over 3074102.59 frames. ], batch size: 191, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:43:34,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6324, 3.5234, 2.9428, 2.3667, 2.6761, 2.3770, 3.8656, 3.5819], device='cuda:0'), covar=tensor([0.2332, 0.0877, 0.1365, 0.1678, 0.1808, 0.1471, 0.0438, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0251, 0.0275, 0.0260, 0.0281, 0.0210, 0.0252, 0.0270], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:43:43,726 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68443.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 17:43:57,930 INFO [train.py:904] (0/8) Epoch 7, batch 7550, loss[loss=0.2288, simple_loss=0.3033, pruned_loss=0.07713, over 16740.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3176, pruned_loss=0.08125, over 3051555.98 frames. ], batch size: 62, lr: 9.58e-03, grad_scale: 4.0 2023-04-28 17:44:10,844 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:44:59,736 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 7, batch 7600, loss[loss=0.2158, simple_loss=0.2931, pruned_loss=0.06927, over 17118.00 frames. ], tot_loss[loss=0.2395, simple_loss=0.317, pruned_loss=0.08105, over 3073928.61 frames. ], batch size: 47, lr: 9.58e-03, grad_scale: 8.0 2023-04-28 17:46:12,990 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2147, 4.2927, 4.0619, 3.9606, 3.7065, 4.1237, 3.9851, 3.8696], device='cuda:0'), covar=tensor([0.0546, 0.0308, 0.0248, 0.0221, 0.0852, 0.0369, 0.0534, 0.0531], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0235, 0.0235, 0.0209, 0.0264, 0.0238, 0.0165, 0.0272], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:46:28,283 INFO [train.py:904] (0/8) Epoch 7, batch 7650, loss[loss=0.1919, simple_loss=0.2741, pruned_loss=0.0549, over 16393.00 frames. ], tot_loss[loss=0.2388, simple_loss=0.3161, pruned_loss=0.08074, over 3083476.14 frames. ], batch size: 35, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:47:08,468 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=68578.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:47:29,674 INFO [optim.py:368] (0/8) 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 17:47:42,701 INFO [train.py:904] (0/8) Epoch 7, batch 7700, loss[loss=0.3174, simple_loss=0.3635, pruned_loss=0.1356, over 11141.00 frames. ], tot_loss[loss=0.2407, simple_loss=0.3173, pruned_loss=0.08207, over 3082220.66 frames. ], batch size: 248, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:48:34,686 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68637.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:37,852 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:48:57,126 INFO [train.py:904] (0/8) Epoch 7, batch 7750, loss[loss=0.2354, simple_loss=0.3136, pruned_loss=0.07856, over 17043.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3171, pruned_loss=0.08176, over 3076152.40 frames. ], batch size: 53, lr: 9.57e-03, grad_scale: 8.0 2023-04-28 17:49:38,529 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 2023-04-28 17:49:44,497 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68685.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:49:44,630 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 17:50:10,427 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7618, 3.3899, 3.1026, 1.7898, 2.8094, 2.2718, 3.2163, 3.3854], device='cuda:0'), covar=tensor([0.0268, 0.0504, 0.0540, 0.1706, 0.0719, 0.0895, 0.0613, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0132, 0.0153, 0.0140, 0.0134, 0.0124, 0.0135, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 17:50:11,059 INFO [train.py:904] (0/8) Epoch 7, batch 7800, loss[loss=0.2588, simple_loss=0.3334, pruned_loss=0.09208, over 16411.00 frames. ], tot_loss[loss=0.2417, simple_loss=0.3182, pruned_loss=0.08263, over 3067983.74 frames. ], batch size: 146, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:50:42,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3102, 2.9437, 2.6181, 2.2319, 2.2452, 2.1605, 2.8326, 2.8797], device='cuda:0'), covar=tensor([0.1992, 0.0652, 0.1237, 0.1561, 0.1647, 0.1576, 0.0431, 0.0844], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0254, 0.0279, 0.0263, 0.0284, 0.0212, 0.0254, 0.0274], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:50:47,656 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0421, 2.4516, 2.3007, 2.9388, 2.3844, 3.2715, 1.7138, 2.7013], device='cuda:0'), covar=tensor([0.1122, 0.0441, 0.0864, 0.0112, 0.0154, 0.0331, 0.1222, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0148, 0.0170, 0.0105, 0.0199, 0.0197, 0.0168, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 17:51:12,803 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=68743.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 17:51:17,595 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=68746.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:51:25,173 INFO [train.py:904] (0/8) Epoch 7, batch 7850, loss[loss=0.2449, simple_loss=0.3263, pruned_loss=0.08177, over 17026.00 frames. ], tot_loss[loss=0.2423, simple_loss=0.3191, pruned_loss=0.08272, over 3058174.32 frames. ], batch size: 55, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:51:30,860 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=68755.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:51:58,838 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5056, 3.6724, 1.8020, 3.9519, 2.4932, 3.9432, 2.0400, 2.6868], device='cuda:0'), covar=tensor([0.0207, 0.0300, 0.1717, 0.0090, 0.0867, 0.0398, 0.1538, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0157, 0.0181, 0.0097, 0.0165, 0.0194, 0.0190, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 17:52:24,310 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=68791.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 17:52:26,861 INFO [optim.py:368] (0/8) 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 17:52:32,687 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0517, 1.2528, 1.7415, 2.0506, 2.1657, 2.3137, 1.5481, 2.1873], device='cuda:0'), covar=tensor([0.0116, 0.0261, 0.0147, 0.0169, 0.0131, 0.0087, 0.0249, 0.0060], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0154, 0.0135, 0.0136, 0.0146, 0.0101, 0.0154, 0.0092], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 17:52:40,555 INFO [train.py:904] (0/8) Epoch 7, batch 7900, loss[loss=0.2727, simple_loss=0.3478, pruned_loss=0.09881, over 15415.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.318, pruned_loss=0.08189, over 3067747.41 frames. ], batch size: 191, lr: 9.56e-03, grad_scale: 8.0 2023-04-28 17:53:58,005 INFO [train.py:904] (0/8) Epoch 7, batch 7950, loss[loss=0.2362, simple_loss=0.3106, pruned_loss=0.0809, over 15401.00 frames. ], tot_loss[loss=0.2421, simple_loss=0.3183, pruned_loss=0.08294, over 3048346.30 frames. ], batch size: 191, lr: 9.55e-03, grad_scale: 2.0 2023-04-28 17:54:02,132 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2463, 3.4481, 1.6834, 3.5899, 2.4262, 3.5775, 1.8568, 2.5997], device='cuda:0'), covar=tensor([0.0180, 0.0293, 0.1580, 0.0089, 0.0723, 0.0471, 0.1488, 0.0681], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0156, 0.0180, 0.0097, 0.0164, 0.0193, 0.0189, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 17:54:31,700 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9106, 5.5120, 5.6701, 5.5054, 5.5265, 6.0605, 5.5839, 5.3464], device='cuda:0'), covar=tensor([0.0824, 0.1589, 0.1875, 0.1768, 0.2369, 0.0875, 0.1170, 0.2276], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0433, 0.0446, 0.0374, 0.0498, 0.0474, 0.0356, 0.0510], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 17:55:03,364 INFO [optim.py:368] (0/8) 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 17:55:05,081 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6040, 2.6727, 1.6586, 2.7467, 2.1104, 2.7412, 1.9442, 2.3775], device='cuda:0'), covar=tensor([0.0224, 0.0338, 0.1141, 0.0127, 0.0545, 0.0520, 0.1048, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0157, 0.0182, 0.0097, 0.0165, 0.0196, 0.0190, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 17:55:12,759 INFO [train.py:904] (0/8) Epoch 7, batch 8000, loss[loss=0.2479, simple_loss=0.3276, pruned_loss=0.08408, over 16718.00 frames. ], tot_loss[loss=0.2425, simple_loss=0.3188, pruned_loss=0.08312, over 3061785.32 frames. ], batch size: 83, lr: 9.55e-03, grad_scale: 4.0 2023-04-28 17:56:00,102 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 7, batch 8050, loss[loss=0.2394, simple_loss=0.3341, pruned_loss=0.0724, over 16742.00 frames. ], tot_loss[loss=0.2413, simple_loss=0.3184, pruned_loss=0.08214, over 3086465.59 frames. ], batch size: 124, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:57:30,025 INFO [optim.py:368] (0/8) 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 17:57:41,498 INFO [train.py:904] (0/8) Epoch 7, batch 8100, loss[loss=0.2102, simple_loss=0.297, pruned_loss=0.06174, over 17141.00 frames. ], tot_loss[loss=0.24, simple_loss=0.3174, pruned_loss=0.08135, over 3084636.10 frames. ], batch size: 49, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:58:07,712 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-28 17:58:35,880 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-28 17:58:40,379 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69041.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:43,494 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69043.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:58:44,764 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.34 vs. limit=5.0 2023-04-28 17:58:46,221 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4541, 3.4291, 3.3750, 2.8199, 3.3632, 2.0666, 3.1125, 2.7209], device='cuda:0'), covar=tensor([0.0105, 0.0084, 0.0128, 0.0222, 0.0074, 0.1767, 0.0111, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0092, 0.0141, 0.0135, 0.0106, 0.0159, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 17:58:48,722 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9679, 1.6071, 2.3407, 2.9046, 2.5834, 3.3635, 2.0801, 3.2583], device='cuda:0'), covar=tensor([0.0095, 0.0289, 0.0188, 0.0133, 0.0154, 0.0071, 0.0268, 0.0047], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0152, 0.0133, 0.0133, 0.0145, 0.0100, 0.0151, 0.0091], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 17:58:55,994 INFO [train.py:904] (0/8) Epoch 7, batch 8150, loss[loss=0.2009, simple_loss=0.2762, pruned_loss=0.0628, over 16823.00 frames. ], tot_loss[loss=0.237, simple_loss=0.3143, pruned_loss=0.07989, over 3100119.02 frames. ], batch size: 116, lr: 9.54e-03, grad_scale: 4.0 2023-04-28 17:59:00,715 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 17:59:59,699 INFO [optim.py:368] (0/8) 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 18:00:10,172 INFO [train.py:904] (0/8) Epoch 7, batch 8200, loss[loss=0.2403, simple_loss=0.3221, pruned_loss=0.07925, over 16265.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3129, pruned_loss=0.08043, over 3058860.03 frames. ], batch size: 165, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:00:12,053 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69103.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:00:13,540 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69104.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:06,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5053, 2.0307, 1.6910, 1.7601, 2.3118, 2.0315, 2.3893, 2.4566], device='cuda:0'), covar=tensor([0.0061, 0.0232, 0.0262, 0.0268, 0.0152, 0.0229, 0.0108, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0099, 0.0172, 0.0170, 0.0170, 0.0166, 0.0173, 0.0163, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:01:30,737 INFO [train.py:904] (0/8) Epoch 7, batch 8250, loss[loss=0.1935, simple_loss=0.2777, pruned_loss=0.05467, over 12365.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3122, pruned_loss=0.07799, over 3054496.26 frames. ], batch size: 248, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:01:47,254 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69162.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:01:58,273 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69168.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:02:40,835 INFO [optim.py:368] (0/8) 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 18:02:52,452 INFO [train.py:904] (0/8) Epoch 7, batch 8300, loss[loss=0.2063, simple_loss=0.2985, pruned_loss=0.05702, over 16280.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3087, pruned_loss=0.07406, over 3062625.24 frames. ], batch size: 165, lr: 9.53e-03, grad_scale: 4.0 2023-04-28 18:03:25,524 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69223.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:28,575 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69225.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:35,009 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69229.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:03:42,684 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69234.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:04:00,664 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69245.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:04:11,891 INFO [train.py:904] (0/8) Epoch 7, batch 8350, loss[loss=0.2136, simple_loss=0.3006, pruned_loss=0.06336, over 16919.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3066, pruned_loss=0.07125, over 3069091.44 frames. ], batch size: 109, lr: 9.52e-03, grad_scale: 4.0 2023-04-28 18:04:46,459 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:04:59,336 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69282.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:06,574 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69286.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:05:20,699 INFO [optim.py:368] (0/8) 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 18:05:32,119 INFO [train.py:904] (0/8) Epoch 7, batch 8400, loss[loss=0.1809, simple_loss=0.2612, pruned_loss=0.05027, over 12289.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3037, pruned_loss=0.06906, over 3048004.60 frames. ], batch size: 247, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:05:39,380 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69306.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:05:47,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6034, 2.0981, 1.6136, 1.9018, 2.4372, 2.1838, 2.5613, 2.5857], device='cuda:0'), covar=tensor([0.0057, 0.0225, 0.0284, 0.0254, 0.0134, 0.0198, 0.0105, 0.0129], device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0171, 0.0168, 0.0167, 0.0164, 0.0171, 0.0157, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:06:02,144 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:23,816 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69334.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:06:34,629 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69341.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:06:52,100 INFO [train.py:904] (0/8) Epoch 7, batch 8450, loss[loss=0.1918, simple_loss=0.2865, pruned_loss=0.04858, over 16795.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3017, pruned_loss=0.06717, over 3042185.31 frames. ], batch size: 102, lr: 9.52e-03, grad_scale: 8.0 2023-04-28 18:07:18,130 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69368.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:38,205 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:50,680 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69389.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:58,604 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:07:59,258 INFO [optim.py:368] (0/8) 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 18:08:06,568 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69399.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:08:10,414 INFO [train.py:904] (0/8) Epoch 7, batch 8500, loss[loss=0.1805, simple_loss=0.272, pruned_loss=0.04451, over 16851.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2969, pruned_loss=0.06389, over 3039478.71 frames. ], batch size: 102, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:08:55,067 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:09:33,187 INFO [train.py:904] (0/8) Epoch 7, batch 8550, loss[loss=0.2138, simple_loss=0.3023, pruned_loss=0.06267, over 16430.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2937, pruned_loss=0.06208, over 3034519.84 frames. ], batch size: 146, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:09:41,235 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69455.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:10:31,829 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 18:10:34,621 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8649, 2.5268, 2.3203, 3.1225, 2.2292, 3.3619, 1.6186, 2.8139], device='cuda:0'), covar=tensor([0.1273, 0.0508, 0.0986, 0.0121, 0.0146, 0.0372, 0.1359, 0.0633], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0147, 0.0169, 0.0105, 0.0190, 0.0194, 0.0170, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 18:10:58,771 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.959e+02 3.006e+02 3.588e+02 4.597e+02 9.217e+02, threshold=7.177e+02, percent-clipped=6.0 2023-04-28 18:11:12,057 INFO [train.py:904] (0/8) Epoch 7, batch 8600, loss[loss=0.2075, simple_loss=0.2947, pruned_loss=0.06018, over 15376.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2933, pruned_loss=0.06067, over 3024960.46 frames. ], batch size: 191, lr: 9.51e-03, grad_scale: 8.0 2023-04-28 18:11:46,311 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69518.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:11:46,466 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2041, 3.1849, 3.1792, 1.7499, 3.3606, 3.4352, 2.7215, 2.7066], device='cuda:0'), covar=tensor([0.0692, 0.0137, 0.0135, 0.1064, 0.0047, 0.0081, 0.0327, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0091, 0.0077, 0.0134, 0.0064, 0.0081, 0.0113, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 18:11:55,643 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:12:47,748 INFO [train.py:904] (0/8) Epoch 7, batch 8650, loss[loss=0.184, simple_loss=0.2717, pruned_loss=0.04811, over 12317.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2912, pruned_loss=0.05846, over 3032478.83 frames. ], batch size: 247, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:13:01,690 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4791, 3.4500, 3.4179, 2.9816, 3.4467, 2.0065, 3.1863, 2.8799], device='cuda:0'), covar=tensor([0.0086, 0.0079, 0.0110, 0.0192, 0.0074, 0.1857, 0.0092, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0090, 0.0138, 0.0129, 0.0104, 0.0157, 0.0123, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:13:27,367 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4572, 4.4763, 4.9210, 4.8724, 4.8661, 4.5778, 4.5464, 4.2792], device='cuda:0'), covar=tensor([0.0229, 0.0403, 0.0302, 0.0387, 0.0485, 0.0283, 0.0764, 0.0475], device='cuda:0'), in_proj_covar=tensor([0.0261, 0.0267, 0.0268, 0.0258, 0.0311, 0.0286, 0.0377, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-28 18:13:53,662 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69581.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:14:19,153 INFO [optim.py:368] (0/8) 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 18:14:31,864 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69601.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:14:32,624 INFO [train.py:904] (0/8) Epoch 7, batch 8700, loss[loss=0.1979, simple_loss=0.2906, pruned_loss=0.0526, over 16688.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2876, pruned_loss=0.05651, over 3044688.97 frames. ], batch size: 134, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:15:02,643 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-28 18:15:22,291 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69629.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:15:37,387 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-28 18:15:45,342 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6587, 3.7473, 4.0508, 4.0293, 4.0363, 3.7844, 3.8296, 3.7725], device='cuda:0'), covar=tensor([0.0271, 0.0475, 0.0333, 0.0412, 0.0446, 0.0335, 0.0730, 0.0375], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0263, 0.0263, 0.0255, 0.0305, 0.0282, 0.0371, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-28 18:15:54,773 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-28 18:16:08,642 INFO [train.py:904] (0/8) Epoch 7, batch 8750, loss[loss=0.2039, simple_loss=0.2835, pruned_loss=0.06216, over 12169.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2871, pruned_loss=0.05578, over 3056669.43 frames. ], batch size: 248, lr: 9.50e-03, grad_scale: 8.0 2023-04-28 18:16:12,449 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4941, 3.4778, 2.7837, 2.1120, 2.4448, 2.1546, 3.7692, 3.4215], device='cuda:0'), covar=tensor([0.2317, 0.0738, 0.1343, 0.1832, 0.1818, 0.1638, 0.0348, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0240, 0.0263, 0.0250, 0.0254, 0.0203, 0.0238, 0.0254], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:16:46,484 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 18:17:04,332 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=69674.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:17:07,499 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69676.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:17:45,991 INFO [optim.py:368] (0/8) 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 18:17:55,736 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69699.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:17:56,079 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-28 18:18:01,172 INFO [train.py:904] (0/8) Epoch 7, batch 8800, loss[loss=0.1952, simple_loss=0.2817, pruned_loss=0.05438, over 15628.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2855, pruned_loss=0.05467, over 3064894.61 frames. ], batch size: 192, lr: 9.49e-03, grad_scale: 8.0 2023-04-28 18:18:47,928 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69724.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:11,470 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=69735.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 18:19:36,495 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69747.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:43,885 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=69750.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:19:46,940 INFO [train.py:904] (0/8) Epoch 7, batch 8850, loss[loss=0.1749, simple_loss=0.2655, pruned_loss=0.04212, over 12518.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2876, pruned_loss=0.05411, over 3043720.82 frames. ], batch size: 247, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:21:21,323 INFO [optim.py:368] (0/8) 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 18:21:34,929 INFO [train.py:904] (0/8) Epoch 7, batch 8900, loss[loss=0.1742, simple_loss=0.2735, pruned_loss=0.03739, over 16555.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2877, pruned_loss=0.05338, over 3041979.47 frames. ], batch size: 62, lr: 9.49e-03, grad_scale: 4.0 2023-04-28 18:21:43,087 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-28 18:22:06,795 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69818.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:22:21,393 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69824.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:23:38,929 INFO [train.py:904] (0/8) Epoch 7, batch 8950, loss[loss=0.1922, simple_loss=0.2799, pruned_loss=0.05222, over 16961.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2876, pruned_loss=0.05393, over 3055525.97 frames. ], batch size: 116, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:24:09,510 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69866.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:21,488 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69872.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:24:42,944 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69881.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:25:14,738 INFO [optim.py:368] (0/8) 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 18:25:27,333 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69901.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:25:27,979 INFO [train.py:904] (0/8) Epoch 7, batch 9000, loss[loss=0.1717, simple_loss=0.2627, pruned_loss=0.04033, over 16478.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.285, pruned_loss=0.05274, over 3063957.79 frames. ], batch size: 68, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:25:27,980 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 18:25:37,198 INFO [train.py:938] (0/8) Epoch 7, validation: loss=0.1647, simple_loss=0.2682, pruned_loss=0.03062, over 944034.00 frames. 2023-04-28 18:25:37,199 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 18:25:38,445 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1701, 4.1557, 4.0179, 3.5844, 4.0592, 1.6702, 3.8725, 3.8373], device='cuda:0'), covar=tensor([0.0073, 0.0068, 0.0110, 0.0223, 0.0075, 0.1962, 0.0096, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0090, 0.0138, 0.0127, 0.0104, 0.0158, 0.0122, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:26:33,161 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:26:33,304 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69929.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:27:12,000 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4089, 1.9614, 1.5974, 1.8032, 2.2297, 2.0405, 2.2693, 2.3795], device='cuda:0'), covar=tensor([0.0049, 0.0207, 0.0282, 0.0252, 0.0132, 0.0192, 0.0118, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0174, 0.0169, 0.0169, 0.0167, 0.0172, 0.0156, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:27:14,337 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69949.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:27:19,725 INFO [train.py:904] (0/8) Epoch 7, batch 9050, loss[loss=0.1951, simple_loss=0.28, pruned_loss=0.05507, over 12940.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2858, pruned_loss=0.0533, over 3052463.95 frames. ], batch size: 250, lr: 9.48e-03, grad_scale: 4.0 2023-04-28 18:27:31,447 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-28 18:28:08,334 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=69976.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:28:09,992 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=69977.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:28:21,346 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-04-28 18:28:50,068 INFO [optim.py:368] (0/8) 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 18:29:00,169 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-70000.pt 2023-04-28 18:29:06,536 INFO [train.py:904] (0/8) Epoch 7, batch 9100, loss[loss=0.1748, simple_loss=0.2761, pruned_loss=0.03674, over 16903.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2853, pruned_loss=0.05378, over 3051568.00 frames. ], batch size: 96, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:29:58,211 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70024.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:29:58,279 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70030.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 18:31:00,626 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70050.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:31:03,194 INFO [train.py:904] (0/8) Epoch 7, batch 9150, loss[loss=0.1912, simple_loss=0.2792, pruned_loss=0.05167, over 16794.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2862, pruned_loss=0.05363, over 3044248.92 frames. ], batch size: 124, lr: 9.47e-03, grad_scale: 4.0 2023-04-28 18:31:47,575 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70072.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:35,335 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.681e+02 3.453e+02 4.314e+02 7.152e+02, threshold=6.905e+02, percent-clipped=1.0 2023-04-28 18:32:39,007 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70098.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:32:45,144 INFO [train.py:904] (0/8) Epoch 7, batch 9200, loss[loss=0.2214, simple_loss=0.3098, pruned_loss=0.06651, over 16852.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2818, pruned_loss=0.05216, over 3065181.95 frames. ], batch size: 124, lr: 9.47e-03, grad_scale: 8.0 2023-04-28 18:33:01,874 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70110.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:33:27,302 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7553, 4.7415, 4.5270, 4.0817, 4.5420, 1.8320, 4.3780, 4.5435], device='cuda:0'), covar=tensor([0.0060, 0.0058, 0.0123, 0.0243, 0.0077, 0.1851, 0.0084, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0102, 0.0090, 0.0137, 0.0126, 0.0104, 0.0157, 0.0122, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:34:22,196 INFO [train.py:904] (0/8) Epoch 7, batch 9250, loss[loss=0.1835, simple_loss=0.2724, pruned_loss=0.04729, over 16741.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2813, pruned_loss=0.05235, over 3048834.32 frames. ], batch size: 124, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:35:01,752 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70171.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:36:01,193 INFO [optim.py:368] (0/8) 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 18:36:13,560 INFO [train.py:904] (0/8) Epoch 7, batch 9300, loss[loss=0.1728, simple_loss=0.2634, pruned_loss=0.04107, over 16702.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2801, pruned_loss=0.05188, over 3055983.00 frames. ], batch size: 62, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:36:46,119 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 18:37:59,028 INFO [train.py:904] (0/8) Epoch 7, batch 9350, loss[loss=0.1777, simple_loss=0.2721, pruned_loss=0.04162, over 16685.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2795, pruned_loss=0.05159, over 3064637.71 frames. ], batch size: 89, lr: 9.46e-03, grad_scale: 8.0 2023-04-28 18:38:17,887 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:39:31,318 INFO [optim.py:368] (0/8) 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 18:39:40,135 INFO [train.py:904] (0/8) Epoch 7, batch 9400, loss[loss=0.1647, simple_loss=0.2511, pruned_loss=0.0392, over 12784.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2791, pruned_loss=0.05114, over 3058884.56 frames. ], batch size: 249, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:39:58,514 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0290, 1.9746, 2.1822, 3.5446, 1.9006, 2.3837, 2.1200, 2.0138], device='cuda:0'), covar=tensor([0.0820, 0.2907, 0.1662, 0.0389, 0.3434, 0.1700, 0.2471, 0.2981], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0341, 0.0291, 0.0307, 0.0384, 0.0365, 0.0308, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:40:18,232 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=70321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:40:36,597 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70330.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 18:41:20,568 INFO [train.py:904] (0/8) Epoch 7, batch 9450, loss[loss=0.1894, simple_loss=0.2787, pruned_loss=0.05006, over 16761.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2811, pruned_loss=0.05178, over 3053145.95 frames. ], batch size: 124, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:42:14,167 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70378.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 18:42:50,915 INFO [optim.py:368] (0/8) 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 18:43:02,106 INFO [train.py:904] (0/8) Epoch 7, batch 9500, loss[loss=0.2, simple_loss=0.2844, pruned_loss=0.05781, over 16934.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2804, pruned_loss=0.05128, over 3062384.70 frames. ], batch size: 109, lr: 9.45e-03, grad_scale: 4.0 2023-04-28 18:43:26,146 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0752, 2.7099, 2.6085, 1.7512, 2.8912, 2.8749, 2.4942, 2.5010], device='cuda:0'), covar=tensor([0.0670, 0.0169, 0.0171, 0.1086, 0.0069, 0.0130, 0.0423, 0.0379], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0093, 0.0080, 0.0138, 0.0065, 0.0082, 0.0116, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 18:44:41,288 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-04-28 18:44:47,831 INFO [train.py:904] (0/8) Epoch 7, batch 9550, loss[loss=0.1882, simple_loss=0.2827, pruned_loss=0.04682, over 16856.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2804, pruned_loss=0.05112, over 3089547.40 frames. ], batch size: 96, lr: 9.44e-03, grad_scale: 4.0 2023-04-28 18:45:18,544 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:45:27,737 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9838, 1.7084, 1.5413, 1.4262, 1.9361, 1.6263, 1.7856, 1.9293], device='cuda:0'), covar=tensor([0.0061, 0.0198, 0.0263, 0.0253, 0.0132, 0.0181, 0.0114, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0098, 0.0177, 0.0172, 0.0172, 0.0169, 0.0174, 0.0158, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:46:08,411 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2875, 3.1083, 2.9103, 3.3534, 3.3816, 3.1755, 3.3298, 3.4285], device='cuda:0'), covar=tensor([0.1013, 0.1087, 0.2230, 0.1127, 0.1067, 0.3571, 0.1655, 0.1142], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0511, 0.0622, 0.0515, 0.0387, 0.0391, 0.0398, 0.0444], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:46:18,722 INFO [optim.py:368] (0/8) 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 18:46:26,851 INFO [train.py:904] (0/8) Epoch 7, batch 9600, loss[loss=0.2159, simple_loss=0.305, pruned_loss=0.06338, over 15338.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2823, pruned_loss=0.05223, over 3073723.81 frames. ], batch size: 191, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:47:38,833 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8332, 4.8742, 4.6459, 4.1876, 4.5636, 1.8380, 4.4788, 4.6126], device='cuda:0'), covar=tensor([0.0058, 0.0050, 0.0109, 0.0218, 0.0068, 0.1937, 0.0089, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0088, 0.0133, 0.0122, 0.0101, 0.0155, 0.0119, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:48:15,888 INFO [train.py:904] (0/8) Epoch 7, batch 9650, loss[loss=0.2144, simple_loss=0.3145, pruned_loss=0.05715, over 16211.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2843, pruned_loss=0.05245, over 3055405.78 frames. ], batch size: 165, lr: 9.44e-03, grad_scale: 8.0 2023-04-28 18:49:03,697 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-28 18:49:55,423 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.673e+02 3.296e+02 3.931e+02 1.187e+03, threshold=6.593e+02, percent-clipped=2.0 2023-04-28 18:50:06,089 INFO [train.py:904] (0/8) Epoch 7, batch 9700, loss[loss=0.1915, simple_loss=0.282, pruned_loss=0.0505, over 16914.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2835, pruned_loss=0.05223, over 3058833.97 frames. ], batch size: 109, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:50:33,215 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=70616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:51:47,811 INFO [train.py:904] (0/8) Epoch 7, batch 9750, loss[loss=0.1797, simple_loss=0.273, pruned_loss=0.04323, over 16767.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2811, pruned_loss=0.05184, over 3039099.43 frames. ], batch size: 124, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:52:23,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2994, 3.5831, 3.5684, 2.5308, 3.3758, 3.4157, 3.5527, 1.9484], device='cuda:0'), covar=tensor([0.0324, 0.0020, 0.0026, 0.0231, 0.0051, 0.0060, 0.0034, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0056, 0.0059, 0.0117, 0.0065, 0.0073, 0.0065, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 18:53:18,489 INFO [optim.py:368] (0/8) 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 18:53:27,266 INFO [train.py:904] (0/8) Epoch 7, batch 9800, loss[loss=0.2011, simple_loss=0.2998, pruned_loss=0.05118, over 16890.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2812, pruned_loss=0.05099, over 3053073.81 frames. ], batch size: 96, lr: 9.43e-03, grad_scale: 8.0 2023-04-28 18:54:17,039 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3961, 1.8379, 1.4511, 1.4500, 2.2076, 1.8693, 2.2099, 2.3415], device='cuda:0'), covar=tensor([0.0051, 0.0275, 0.0344, 0.0325, 0.0151, 0.0225, 0.0124, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0096, 0.0175, 0.0171, 0.0169, 0.0167, 0.0171, 0.0155, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 18:55:12,750 INFO [train.py:904] (0/8) Epoch 7, batch 9850, loss[loss=0.1818, simple_loss=0.2666, pruned_loss=0.04848, over 12396.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.282, pruned_loss=0.05063, over 3050731.17 frames. ], batch size: 248, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:55:43,260 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70766.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:56:28,339 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8115, 1.5147, 2.2015, 2.8332, 2.4456, 2.9025, 1.8144, 3.0186], device='cuda:0'), covar=tensor([0.0104, 0.0320, 0.0188, 0.0148, 0.0178, 0.0080, 0.0309, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0151, 0.0134, 0.0134, 0.0141, 0.0096, 0.0150, 0.0088], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 18:56:54,661 INFO [optim.py:368] (0/8) 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 18:57:04,352 INFO [train.py:904] (0/8) Epoch 7, batch 9900, loss[loss=0.2029, simple_loss=0.301, pruned_loss=0.05241, over 16762.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2829, pruned_loss=0.05081, over 3057092.61 frames. ], batch size: 83, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:57:24,172 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-28 18:57:33,010 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70814.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 18:58:26,291 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7407, 4.2501, 4.3991, 3.0640, 3.9757, 4.2578, 4.1036, 2.3277], device='cuda:0'), covar=tensor([0.0337, 0.0017, 0.0020, 0.0241, 0.0051, 0.0044, 0.0030, 0.0347], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0056, 0.0059, 0.0117, 0.0065, 0.0072, 0.0066, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0001, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 18:59:01,234 INFO [train.py:904] (0/8) Epoch 7, batch 9950, loss[loss=0.1767, simple_loss=0.2748, pruned_loss=0.0393, over 16792.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2847, pruned_loss=0.05101, over 3061840.83 frames. ], batch size: 83, lr: 9.42e-03, grad_scale: 8.0 2023-04-28 18:59:29,075 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 19:00:47,805 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.726e+02 3.249e+02 3.973e+02 1.098e+03, threshold=6.498e+02, percent-clipped=6.0 2023-04-28 19:01:00,574 INFO [train.py:904] (0/8) Epoch 7, batch 10000, loss[loss=0.1687, simple_loss=0.2579, pruned_loss=0.03976, over 12400.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2835, pruned_loss=0.05064, over 3063673.36 frames. ], batch size: 250, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:01:29,760 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=70916.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:01:41,585 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4679, 1.9526, 1.5266, 1.6954, 2.2781, 1.9809, 2.2993, 2.4257], device='cuda:0'), covar=tensor([0.0061, 0.0268, 0.0355, 0.0333, 0.0160, 0.0247, 0.0121, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0097, 0.0175, 0.0171, 0.0169, 0.0167, 0.0173, 0.0155, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:02:17,056 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 19:02:40,829 INFO [train.py:904] (0/8) Epoch 7, batch 10050, loss[loss=0.1857, simple_loss=0.2763, pruned_loss=0.04758, over 16744.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2834, pruned_loss=0.05035, over 3077631.74 frames. ], batch size: 76, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:02:51,303 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6555, 2.3783, 2.0914, 3.5831, 2.3730, 3.7327, 1.3338, 2.6230], device='cuda:0'), covar=tensor([0.1381, 0.0683, 0.1245, 0.0094, 0.0136, 0.0321, 0.1564, 0.0823], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0145, 0.0169, 0.0102, 0.0168, 0.0193, 0.0169, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 19:03:04,145 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=70964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:03:56,353 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=70992.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:04:04,313 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.660e+02 2.613e+02 3.026e+02 3.875e+02 7.173e+02, threshold=6.053e+02, percent-clipped=4.0 2023-04-28 19:04:12,154 INFO [train.py:904] (0/8) Epoch 7, batch 10100, loss[loss=0.1942, simple_loss=0.2801, pruned_loss=0.05414, over 16699.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2841, pruned_loss=0.05116, over 3075702.73 frames. ], batch size: 134, lr: 9.41e-03, grad_scale: 8.0 2023-04-28 19:05:06,994 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:05:30,841 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-7.pt 2023-04-28 19:05:54,946 INFO [train.py:904] (0/8) Epoch 8, batch 0, loss[loss=0.2721, simple_loss=0.3179, pruned_loss=0.1131, over 16953.00 frames. ], tot_loss[loss=0.2721, simple_loss=0.3179, pruned_loss=0.1131, over 16953.00 frames. ], batch size: 90, lr: 8.86e-03, grad_scale: 8.0 2023-04-28 19:05:54,947 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 19:06:02,580 INFO [train.py:938] (0/8) Epoch 8, validation: loss=0.1627, simple_loss=0.2663, pruned_loss=0.02958, over 944034.00 frames. 2023-04-28 19:06:02,581 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 19:06:03,999 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71053.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:06:13,863 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7313, 5.2557, 5.3443, 5.1870, 5.1989, 5.7428, 5.3210, 5.0916], device='cuda:0'), covar=tensor([0.0942, 0.1523, 0.1901, 0.1800, 0.2253, 0.0999, 0.1311, 0.2341], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0403, 0.0419, 0.0355, 0.0464, 0.0444, 0.0334, 0.0471], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:06:55,592 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 19:07:09,996 INFO [train.py:904] (0/8) Epoch 8, batch 50, loss[loss=0.2288, simple_loss=0.2944, pruned_loss=0.08164, over 16846.00 frames. ], tot_loss[loss=0.2312, simple_loss=0.3039, pruned_loss=0.07922, over 749075.57 frames. ], batch size: 102, lr: 8.86e-03, grad_scale: 1.0 2023-04-28 19:08:17,903 INFO [train.py:904] (0/8) Epoch 8, batch 100, loss[loss=0.1839, simple_loss=0.2702, pruned_loss=0.04884, over 17265.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2962, pruned_loss=0.07136, over 1314848.34 frames. ], batch size: 52, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:09:20,424 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.11 vs. limit=2.0 2023-04-28 19:09:23,827 INFO [optim.py:368] (0/8) 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 19:09:26,724 INFO [train.py:904] (0/8) Epoch 8, batch 150, loss[loss=0.264, simple_loss=0.3319, pruned_loss=0.09805, over 12052.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2918, pruned_loss=0.06846, over 1762248.08 frames. ], batch size: 246, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:10:28,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-28 19:10:33,459 INFO [train.py:904] (0/8) Epoch 8, batch 200, loss[loss=0.1833, simple_loss=0.2684, pruned_loss=0.04909, over 17232.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.291, pruned_loss=0.06762, over 2108651.99 frames. ], batch size: 45, lr: 8.85e-03, grad_scale: 1.0 2023-04-28 19:10:47,768 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0204, 4.0057, 4.4397, 3.3140, 4.1189, 4.3678, 4.1184, 2.8132], device='cuda:0'), covar=tensor([0.0288, 0.0033, 0.0022, 0.0200, 0.0038, 0.0031, 0.0032, 0.0273], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0063, 0.0062, 0.0119, 0.0066, 0.0074, 0.0067, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 19:11:40,026 INFO [optim.py:368] (0/8) 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 19:11:42,957 INFO [train.py:904] (0/8) Epoch 8, batch 250, loss[loss=0.2451, simple_loss=0.3098, pruned_loss=0.09019, over 15441.00 frames. ], tot_loss[loss=0.212, simple_loss=0.289, pruned_loss=0.06752, over 2366881.11 frames. ], batch size: 191, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:11:56,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6288, 1.4701, 1.9927, 2.3566, 2.4341, 2.4167, 1.6024, 2.5372], device='cuda:0'), covar=tensor([0.0091, 0.0289, 0.0173, 0.0150, 0.0140, 0.0113, 0.0267, 0.0058], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0155, 0.0138, 0.0138, 0.0144, 0.0098, 0.0152, 0.0090], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 19:12:47,162 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71348.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:12:49,393 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8632, 5.1805, 4.8575, 4.9240, 4.6814, 4.6075, 4.6691, 5.2387], device='cuda:0'), covar=tensor([0.0884, 0.0778, 0.1155, 0.0546, 0.0713, 0.0900, 0.0934, 0.0909], device='cuda:0'), in_proj_covar=tensor([0.0462, 0.0588, 0.0494, 0.0399, 0.0378, 0.0392, 0.0494, 0.0437], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:12:52,140 INFO [train.py:904] (0/8) Epoch 8, batch 300, loss[loss=0.1952, simple_loss=0.2784, pruned_loss=0.056, over 17111.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2858, pruned_loss=0.06597, over 2577755.65 frames. ], batch size: 55, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:13:38,975 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71386.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:13:58,500 INFO [optim.py:368] (0/8) 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 19:14:01,159 INFO [train.py:904] (0/8) Epoch 8, batch 350, loss[loss=0.1894, simple_loss=0.2827, pruned_loss=0.04807, over 17041.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2831, pruned_loss=0.06384, over 2735522.22 frames. ], batch size: 50, lr: 8.84e-03, grad_scale: 1.0 2023-04-28 19:15:10,468 INFO [train.py:904] (0/8) Epoch 8, batch 400, loss[loss=0.2249, simple_loss=0.2819, pruned_loss=0.08392, over 16670.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2804, pruned_loss=0.06304, over 2864548.93 frames. ], batch size: 89, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:15:19,840 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-04-28 19:16:17,823 INFO [optim.py:368] (0/8) 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 19:16:20,160 INFO [train.py:904] (0/8) Epoch 8, batch 450, loss[loss=0.1702, simple_loss=0.2472, pruned_loss=0.04659, over 16017.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2782, pruned_loss=0.06178, over 2967874.50 frames. ], batch size: 35, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:16:43,191 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7764, 3.8511, 4.2756, 3.2975, 3.9221, 4.1162, 3.9510, 2.3720], device='cuda:0'), covar=tensor([0.0337, 0.0047, 0.0031, 0.0195, 0.0049, 0.0059, 0.0043, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0064, 0.0063, 0.0120, 0.0067, 0.0077, 0.0068, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 19:17:04,080 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8918, 4.0262, 4.1031, 3.3277, 3.8652, 4.0826, 3.9699, 2.1722], device='cuda:0'), covar=tensor([0.0365, 0.0046, 0.0059, 0.0242, 0.0073, 0.0106, 0.0078, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0064, 0.0063, 0.0120, 0.0067, 0.0076, 0.0068, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 19:17:28,371 INFO [train.py:904] (0/8) Epoch 8, batch 500, loss[loss=0.2283, simple_loss=0.2869, pruned_loss=0.08483, over 16678.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2764, pruned_loss=0.06036, over 3049573.33 frames. ], batch size: 134, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:17:36,543 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-28 19:18:33,696 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.499e+02 3.026e+02 3.865e+02 9.130e+02, threshold=6.053e+02, percent-clipped=1.0 2023-04-28 19:18:37,356 INFO [train.py:904] (0/8) Epoch 8, batch 550, loss[loss=0.185, simple_loss=0.2633, pruned_loss=0.05337, over 17217.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2756, pruned_loss=0.0592, over 3114809.90 frames. ], batch size: 44, lr: 8.83e-03, grad_scale: 2.0 2023-04-28 19:18:56,766 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:33,239 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-28 19:19:41,725 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71648.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:19:45,784 INFO [train.py:904] (0/8) Epoch 8, batch 600, loss[loss=0.186, simple_loss=0.2613, pruned_loss=0.05537, over 15866.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.275, pruned_loss=0.0592, over 3156869.88 frames. ], batch size: 35, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:20:19,061 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:23,350 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9992, 1.8335, 2.3652, 2.9611, 2.7290, 3.2318, 2.1691, 3.1902], device='cuda:0'), covar=tensor([0.0137, 0.0297, 0.0207, 0.0165, 0.0179, 0.0133, 0.0266, 0.0088], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0157, 0.0141, 0.0141, 0.0148, 0.0103, 0.0155, 0.0093], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 19:20:33,413 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=71686.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:39,644 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-04-28 19:20:45,907 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71696.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:20:50,943 INFO [optim.py:368] (0/8) 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 19:20:53,321 INFO [train.py:904] (0/8) Epoch 8, batch 650, loss[loss=0.1747, simple_loss=0.2445, pruned_loss=0.05244, over 16289.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2734, pruned_loss=0.05855, over 3188813.49 frames. ], batch size: 36, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:21:37,503 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=71734.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:22:01,906 INFO [train.py:904] (0/8) Epoch 8, batch 700, loss[loss=0.1959, simple_loss=0.2649, pruned_loss=0.06343, over 16810.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2739, pruned_loss=0.05852, over 3218293.06 frames. ], batch size: 116, lr: 8.82e-03, grad_scale: 2.0 2023-04-28 19:22:48,467 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1931, 5.0926, 5.0122, 4.6776, 4.5458, 5.0194, 5.0262, 4.6166], device='cuda:0'), covar=tensor([0.0490, 0.0382, 0.0227, 0.0215, 0.0975, 0.0348, 0.0269, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0263, 0.0261, 0.0233, 0.0292, 0.0265, 0.0178, 0.0303], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 19:23:05,954 INFO [optim.py:368] (0/8) 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 19:23:08,649 INFO [train.py:904] (0/8) Epoch 8, batch 750, loss[loss=0.1945, simple_loss=0.2825, pruned_loss=0.05318, over 16635.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2744, pruned_loss=0.05933, over 3226734.54 frames. ], batch size: 57, lr: 8.81e-03, grad_scale: 2.0 2023-04-28 19:23:57,354 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 19:24:17,947 INFO [train.py:904] (0/8) Epoch 8, batch 800, loss[loss=0.2052, simple_loss=0.2907, pruned_loss=0.05983, over 16628.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2745, pruned_loss=0.0591, over 3239470.22 frames. ], batch size: 62, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:24:32,334 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5080, 3.4112, 3.8811, 2.7638, 3.6335, 3.8941, 3.7331, 2.4693], device='cuda:0'), covar=tensor([0.0342, 0.0180, 0.0030, 0.0242, 0.0054, 0.0060, 0.0051, 0.0283], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0064, 0.0063, 0.0118, 0.0067, 0.0076, 0.0069, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 19:25:23,720 INFO [optim.py:368] (0/8) 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 19:25:25,978 INFO [train.py:904] (0/8) Epoch 8, batch 850, loss[loss=0.2131, simple_loss=0.2797, pruned_loss=0.07323, over 16876.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2735, pruned_loss=0.05905, over 3253795.45 frames. ], batch size: 109, lr: 8.81e-03, grad_scale: 4.0 2023-04-28 19:25:49,786 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=71919.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:26:18,805 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6362, 2.6825, 2.4747, 4.0715, 3.3813, 4.0921, 1.3633, 2.8278], device='cuda:0'), covar=tensor([0.1320, 0.0628, 0.1081, 0.0125, 0.0219, 0.0362, 0.1413, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0152, 0.0174, 0.0113, 0.0192, 0.0205, 0.0171, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 19:26:32,702 INFO [train.py:904] (0/8) Epoch 8, batch 900, loss[loss=0.1959, simple_loss=0.2778, pruned_loss=0.05698, over 16668.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2722, pruned_loss=0.05789, over 3268124.44 frames. ], batch size: 57, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:26:57,152 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9799, 4.1850, 4.4302, 3.1434, 3.9614, 4.4440, 4.0444, 2.9138], device='cuda:0'), covar=tensor([0.0309, 0.0025, 0.0029, 0.0220, 0.0049, 0.0045, 0.0041, 0.0247], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0063, 0.0062, 0.0117, 0.0067, 0.0075, 0.0068, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 19:27:00,523 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=71972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:13,330 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=71980.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:27:38,438 INFO [optim.py:368] (0/8) 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 19:27:38,824 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-72000.pt 2023-04-28 19:27:44,219 INFO [train.py:904] (0/8) Epoch 8, batch 950, loss[loss=0.1949, simple_loss=0.263, pruned_loss=0.06336, over 16881.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2713, pruned_loss=0.05696, over 3284100.11 frames. ], batch size: 96, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:28:52,164 INFO [train.py:904] (0/8) Epoch 8, batch 1000, loss[loss=0.181, simple_loss=0.2762, pruned_loss=0.04294, over 17260.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2699, pruned_loss=0.0561, over 3296109.18 frames. ], batch size: 52, lr: 8.80e-03, grad_scale: 4.0 2023-04-28 19:29:14,661 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 19:29:58,314 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.648e+02 3.149e+02 3.724e+02 6.885e+02, threshold=6.299e+02, percent-clipped=3.0 2023-04-28 19:30:01,627 INFO [train.py:904] (0/8) Epoch 8, batch 1050, loss[loss=0.1782, simple_loss=0.2749, pruned_loss=0.04077, over 17133.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2702, pruned_loss=0.05634, over 3308635.32 frames. ], batch size: 48, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:30:40,193 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:31:10,582 INFO [train.py:904] (0/8) Epoch 8, batch 1100, loss[loss=0.2077, simple_loss=0.2673, pruned_loss=0.07402, over 16746.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2699, pruned_loss=0.05647, over 3302743.20 frames. ], batch size: 124, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:03,077 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72190.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:32:15,694 INFO [optim.py:368] (0/8) 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 19:32:18,231 INFO [train.py:904] (0/8) Epoch 8, batch 1150, loss[loss=0.2212, simple_loss=0.2831, pruned_loss=0.07967, over 16269.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2694, pruned_loss=0.05617, over 3306024.50 frames. ], batch size: 165, lr: 8.79e-03, grad_scale: 4.0 2023-04-28 19:32:56,151 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7650, 2.2444, 2.3716, 4.3814, 2.1954, 2.8558, 2.3480, 2.4556], device='cuda:0'), covar=tensor([0.0737, 0.2731, 0.1649, 0.0340, 0.3083, 0.1777, 0.2496, 0.2690], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0361, 0.0305, 0.0325, 0.0398, 0.0402, 0.0326, 0.0432], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:33:26,646 INFO [train.py:904] (0/8) Epoch 8, batch 1200, loss[loss=0.198, simple_loss=0.2877, pruned_loss=0.05417, over 17060.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2689, pruned_loss=0.05612, over 3303240.85 frames. ], batch size: 55, lr: 8.79e-03, grad_scale: 8.0 2023-04-28 19:33:53,237 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:33:56,994 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72275.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:34:30,643 INFO [optim.py:368] (0/8) 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 19:34:32,979 INFO [train.py:904] (0/8) Epoch 8, batch 1250, loss[loss=0.1893, simple_loss=0.2648, pruned_loss=0.05691, over 16796.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2684, pruned_loss=0.05658, over 3309350.29 frames. ], batch size: 102, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:34:58,452 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:06,118 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:35:41,562 INFO [train.py:904] (0/8) Epoch 8, batch 1300, loss[loss=0.1721, simple_loss=0.2469, pruned_loss=0.04864, over 15486.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2692, pruned_loss=0.05703, over 3306744.50 frames. ], batch size: 191, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:35:44,137 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-28 19:36:30,388 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72387.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:36:49,368 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.738e+02 2.612e+02 3.072e+02 3.829e+02 7.494e+02, threshold=6.144e+02, percent-clipped=3.0 2023-04-28 19:36:52,120 INFO [train.py:904] (0/8) Epoch 8, batch 1350, loss[loss=0.1997, simple_loss=0.2699, pruned_loss=0.06477, over 16334.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2696, pruned_loss=0.05697, over 3313275.75 frames. ], batch size: 165, lr: 8.78e-03, grad_scale: 8.0 2023-04-28 19:37:22,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7371, 2.7281, 2.3089, 4.0417, 3.5012, 4.0472, 1.4630, 2.8926], device='cuda:0'), covar=tensor([0.1396, 0.0611, 0.1159, 0.0126, 0.0240, 0.0366, 0.1431, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0154, 0.0174, 0.0116, 0.0195, 0.0206, 0.0172, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 19:38:01,927 INFO [train.py:904] (0/8) Epoch 8, batch 1400, loss[loss=0.1941, simple_loss=0.2625, pruned_loss=0.06287, over 16809.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2691, pruned_loss=0.05652, over 3308717.62 frames. ], batch size: 76, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:38:45,986 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 19:38:47,782 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72485.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:39:07,467 INFO [optim.py:368] (0/8) 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 19:39:10,639 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-28 19:39:11,115 INFO [train.py:904] (0/8) Epoch 8, batch 1450, loss[loss=0.2014, simple_loss=0.2644, pruned_loss=0.06919, over 16389.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2685, pruned_loss=0.05676, over 3298742.46 frames. ], batch size: 146, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:40:20,582 INFO [train.py:904] (0/8) Epoch 8, batch 1500, loss[loss=0.1919, simple_loss=0.2729, pruned_loss=0.0554, over 16658.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2674, pruned_loss=0.05623, over 3301173.72 frames. ], batch size: 62, lr: 8.77e-03, grad_scale: 8.0 2023-04-28 19:40:52,055 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72575.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:41:25,451 INFO [optim.py:368] (0/8) 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 19:41:28,321 INFO [train.py:904] (0/8) Epoch 8, batch 1550, loss[loss=0.2124, simple_loss=0.2707, pruned_loss=0.07709, over 16572.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2684, pruned_loss=0.05733, over 3304598.43 frames. ], batch size: 75, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:41:58,084 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72623.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:42:39,334 INFO [train.py:904] (0/8) Epoch 8, batch 1600, loss[loss=0.2131, simple_loss=0.3046, pruned_loss=0.06084, over 16719.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2712, pruned_loss=0.05836, over 3302537.62 frames. ], batch size: 57, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:43:20,422 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=72682.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:43:21,801 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1377, 3.9456, 4.1591, 4.3839, 4.4476, 4.0197, 4.1662, 4.4056], device='cuda:0'), covar=tensor([0.1123, 0.0910, 0.1287, 0.0539, 0.0567, 0.1338, 0.1383, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0620, 0.0774, 0.0625, 0.0471, 0.0478, 0.0484, 0.0532], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:43:25,630 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 19:43:44,410 INFO [optim.py:368] (0/8) 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 19:43:47,317 INFO [train.py:904] (0/8) Epoch 8, batch 1650, loss[loss=0.2173, simple_loss=0.3148, pruned_loss=0.05986, over 17066.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2739, pruned_loss=0.05874, over 3306180.50 frames. ], batch size: 53, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:44:06,050 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9522, 4.3363, 4.6598, 2.1649, 4.9414, 4.8816, 3.4668, 3.6384], device='cuda:0'), covar=tensor([0.0679, 0.0113, 0.0138, 0.0932, 0.0036, 0.0068, 0.0297, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0094, 0.0083, 0.0137, 0.0070, 0.0091, 0.0119, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 19:44:58,251 INFO [train.py:904] (0/8) Epoch 8, batch 1700, loss[loss=0.202, simple_loss=0.2868, pruned_loss=0.05855, over 17063.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2752, pruned_loss=0.05891, over 3310456.42 frames. ], batch size: 53, lr: 8.76e-03, grad_scale: 8.0 2023-04-28 19:45:06,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7608, 6.1256, 5.8185, 6.0455, 5.4986, 5.2528, 5.6603, 6.1977], device='cuda:0'), covar=tensor([0.0855, 0.0716, 0.1328, 0.0531, 0.0672, 0.0614, 0.0714, 0.0772], device='cuda:0'), in_proj_covar=tensor([0.0482, 0.0616, 0.0509, 0.0411, 0.0386, 0.0400, 0.0512, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:45:44,195 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72785.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:46:05,113 INFO [optim.py:368] (0/8) 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 19:46:07,556 INFO [train.py:904] (0/8) Epoch 8, batch 1750, loss[loss=0.1991, simple_loss=0.2876, pruned_loss=0.05525, over 17081.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.276, pruned_loss=0.05896, over 3310900.19 frames. ], batch size: 53, lr: 8.75e-03, grad_scale: 8.0 2023-04-28 19:46:50,155 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=72833.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:46:57,513 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7720, 4.7433, 5.3176, 5.2875, 5.2354, 4.8922, 4.8186, 4.5950], device='cuda:0'), covar=tensor([0.0267, 0.0398, 0.0283, 0.0363, 0.0394, 0.0288, 0.0791, 0.0432], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0311, 0.0310, 0.0295, 0.0351, 0.0329, 0.0432, 0.0266], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 19:47:02,030 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2676, 1.9761, 2.5819, 3.2004, 2.9217, 3.5199, 2.5517, 3.4834], device='cuda:0'), covar=tensor([0.0109, 0.0268, 0.0190, 0.0136, 0.0139, 0.0093, 0.0233, 0.0076], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0158, 0.0145, 0.0144, 0.0151, 0.0105, 0.0157, 0.0098], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 19:47:16,120 INFO [train.py:904] (0/8) Epoch 8, batch 1800, loss[loss=0.2379, simple_loss=0.3047, pruned_loss=0.08551, over 16493.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2766, pruned_loss=0.05865, over 3317916.05 frames. ], batch size: 146, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:48:15,319 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72894.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:48:21,247 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2089, 3.3213, 1.9142, 3.4315, 2.4332, 3.4463, 1.9305, 2.6401], device='cuda:0'), covar=tensor([0.0180, 0.0350, 0.1227, 0.0155, 0.0693, 0.0481, 0.1255, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0163, 0.0180, 0.0107, 0.0164, 0.0201, 0.0191, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 19:48:24,994 INFO [optim.py:368] (0/8) 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 19:48:26,878 INFO [train.py:904] (0/8) Epoch 8, batch 1850, loss[loss=0.1829, simple_loss=0.2772, pruned_loss=0.04433, over 16685.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2775, pruned_loss=0.05859, over 3314884.05 frames. ], batch size: 57, lr: 8.75e-03, grad_scale: 4.0 2023-04-28 19:49:35,224 INFO [train.py:904] (0/8) Epoch 8, batch 1900, loss[loss=0.2048, simple_loss=0.2956, pruned_loss=0.05698, over 16773.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2771, pruned_loss=0.05816, over 3313397.81 frames. ], batch size: 62, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:49:39,132 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=72955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:49:52,052 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=72964.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 19:50:05,757 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7690, 3.9441, 2.0595, 4.1707, 2.7107, 4.1168, 2.1803, 2.9715], device='cuda:0'), covar=tensor([0.0172, 0.0292, 0.1366, 0.0154, 0.0719, 0.0453, 0.1289, 0.0606], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0163, 0.0180, 0.0107, 0.0164, 0.0201, 0.0190, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 19:50:16,708 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=72982.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:50:41,099 INFO [optim.py:368] (0/8) 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 19:50:43,039 INFO [train.py:904] (0/8) Epoch 8, batch 1950, loss[loss=0.188, simple_loss=0.2852, pruned_loss=0.0454, over 16667.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2768, pruned_loss=0.05742, over 3315672.58 frames. ], batch size: 57, lr: 8.74e-03, grad_scale: 4.0 2023-04-28 19:50:49,462 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6783, 3.8892, 4.1032, 1.9147, 4.4186, 4.3805, 3.1129, 3.2597], device='cuda:0'), covar=tensor([0.0704, 0.0151, 0.0183, 0.1070, 0.0052, 0.0095, 0.0385, 0.0351], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0095, 0.0085, 0.0139, 0.0071, 0.0094, 0.0122, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 19:51:11,991 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0954, 3.4104, 3.1318, 1.9551, 2.8534, 2.3045, 3.5643, 3.4343], device='cuda:0'), covar=tensor([0.0211, 0.0599, 0.0584, 0.1471, 0.0700, 0.0856, 0.0428, 0.0684], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0138, 0.0154, 0.0140, 0.0135, 0.0124, 0.0134, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 19:51:14,252 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73025.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 19:51:21,211 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73030.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:51:50,198 INFO [train.py:904] (0/8) Epoch 8, batch 2000, loss[loss=0.1973, simple_loss=0.2951, pruned_loss=0.04975, over 16803.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2772, pruned_loss=0.05733, over 3324854.16 frames. ], batch size: 57, lr: 8.74e-03, grad_scale: 8.0 2023-04-28 19:52:54,130 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6692, 2.6523, 2.3311, 2.4352, 2.9599, 2.6900, 3.5937, 3.2439], device='cuda:0'), covar=tensor([0.0051, 0.0235, 0.0277, 0.0266, 0.0165, 0.0243, 0.0131, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0183, 0.0180, 0.0180, 0.0182, 0.0186, 0.0186, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:52:58,786 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 8, batch 2050, loss[loss=0.202, simple_loss=0.2863, pruned_loss=0.05885, over 17045.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2767, pruned_loss=0.05762, over 3321279.67 frames. ], batch size: 55, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:54:09,331 INFO [train.py:904] (0/8) Epoch 8, batch 2100, loss[loss=0.2144, simple_loss=0.2892, pruned_loss=0.06982, over 16828.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2785, pruned_loss=0.05823, over 3315976.77 frames. ], batch size: 90, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:54:15,736 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 19:54:47,101 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7263, 3.3445, 2.6018, 5.0471, 4.3851, 4.8007, 1.4648, 3.3140], device='cuda:0'), covar=tensor([0.1430, 0.0542, 0.1162, 0.0131, 0.0249, 0.0309, 0.1519, 0.0714], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0153, 0.0172, 0.0118, 0.0197, 0.0206, 0.0171, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 19:55:14,762 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.921e+02 2.680e+02 3.272e+02 3.798e+02 7.634e+02, threshold=6.545e+02, percent-clipped=2.0 2023-04-28 19:55:16,488 INFO [train.py:904] (0/8) Epoch 8, batch 2150, loss[loss=0.231, simple_loss=0.2982, pruned_loss=0.0819, over 16701.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2792, pruned_loss=0.05871, over 3322393.68 frames. ], batch size: 134, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:56:21,720 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73250.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:56:23,763 INFO [train.py:904] (0/8) Epoch 8, batch 2200, loss[loss=0.1645, simple_loss=0.2505, pruned_loss=0.03927, over 17186.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2796, pruned_loss=0.05925, over 3317266.60 frames. ], batch size: 44, lr: 8.73e-03, grad_scale: 8.0 2023-04-28 19:57:20,368 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73293.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:57:29,786 INFO [optim.py:368] (0/8) 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 19:57:31,982 INFO [train.py:904] (0/8) Epoch 8, batch 2250, loss[loss=0.1443, simple_loss=0.2278, pruned_loss=0.03038, over 16800.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2793, pruned_loss=0.05937, over 3312864.24 frames. ], batch size: 39, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:57:46,102 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6478, 4.4176, 4.6315, 4.8595, 4.9773, 4.4579, 4.7842, 4.9048], device='cuda:0'), covar=tensor([0.1289, 0.1094, 0.1404, 0.0644, 0.0622, 0.0895, 0.1149, 0.0826], device='cuda:0'), in_proj_covar=tensor([0.0495, 0.0619, 0.0770, 0.0619, 0.0469, 0.0473, 0.0485, 0.0530], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:57:54,747 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-28 19:57:56,916 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73320.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 19:58:24,293 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3765, 3.3412, 3.4081, 2.8842, 3.3294, 2.1492, 3.1234, 2.6756], device='cuda:0'), covar=tensor([0.0113, 0.0089, 0.0141, 0.0211, 0.0074, 0.1683, 0.0133, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0103, 0.0154, 0.0149, 0.0119, 0.0166, 0.0141, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 19:58:40,769 INFO [train.py:904] (0/8) Epoch 8, batch 2300, loss[loss=0.1908, simple_loss=0.2706, pruned_loss=0.05548, over 16884.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2794, pruned_loss=0.05985, over 3319516.49 frames. ], batch size: 90, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 19:58:44,170 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73354.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 19:59:10,962 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4383, 3.5801, 3.9859, 2.6632, 3.5194, 3.8620, 3.7160, 2.3430], device='cuda:0'), covar=tensor([0.0337, 0.0177, 0.0023, 0.0242, 0.0075, 0.0079, 0.0055, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0066, 0.0063, 0.0118, 0.0068, 0.0079, 0.0071, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 19:59:24,038 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-28 19:59:48,861 INFO [optim.py:368] (0/8) 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 19:59:49,989 INFO [train.py:904] (0/8) Epoch 8, batch 2350, loss[loss=0.1688, simple_loss=0.2681, pruned_loss=0.0348, over 17142.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2804, pruned_loss=0.06097, over 3316697.93 frames. ], batch size: 48, lr: 8.72e-03, grad_scale: 8.0 2023-04-28 20:00:25,400 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 20:00:58,129 INFO [train.py:904] (0/8) Epoch 8, batch 2400, loss[loss=0.2188, simple_loss=0.2921, pruned_loss=0.07276, over 16842.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2816, pruned_loss=0.0616, over 3314099.53 frames. ], batch size: 83, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:01:20,069 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 20:02:04,600 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.637e+02 3.145e+02 3.743e+02 9.455e+02, threshold=6.290e+02, percent-clipped=3.0 2023-04-28 20:02:05,790 INFO [train.py:904] (0/8) Epoch 8, batch 2450, loss[loss=0.2035, simple_loss=0.2921, pruned_loss=0.05743, over 17046.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2816, pruned_loss=0.06046, over 3314865.83 frames. ], batch size: 53, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:02:52,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0008, 4.1972, 4.5075, 2.1452, 4.8571, 4.7788, 3.3812, 3.4536], device='cuda:0'), covar=tensor([0.0665, 0.0144, 0.0169, 0.1111, 0.0040, 0.0079, 0.0371, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0096, 0.0085, 0.0139, 0.0069, 0.0093, 0.0121, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 20:03:12,598 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73550.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:03:14,173 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 20:03:14,452 INFO [train.py:904] (0/8) Epoch 8, batch 2500, loss[loss=0.1996, simple_loss=0.2943, pruned_loss=0.05239, over 17144.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2803, pruned_loss=0.0595, over 3320574.57 frames. ], batch size: 49, lr: 8.71e-03, grad_scale: 8.0 2023-04-28 20:04:07,460 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73590.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:11,993 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7988, 4.9046, 5.2874, 5.2820, 5.3451, 4.9904, 4.8293, 4.6966], device='cuda:0'), covar=tensor([0.0406, 0.0497, 0.0505, 0.0599, 0.0560, 0.0449, 0.1259, 0.0432], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0306, 0.0306, 0.0292, 0.0347, 0.0324, 0.0426, 0.0262], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 20:04:18,255 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73598.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:04:22,388 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.774e+02 3.514e+02 4.177e+02 6.912e+02, threshold=7.028e+02, percent-clipped=4.0 2023-04-28 20:04:23,444 INFO [train.py:904] (0/8) Epoch 8, batch 2550, loss[loss=0.2048, simple_loss=0.2881, pruned_loss=0.06072, over 16667.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2805, pruned_loss=0.05943, over 3323575.69 frames. ], batch size: 89, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:04:48,882 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73620.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:04:58,944 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 20:05:29,145 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73649.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,290 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73651.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:05:32,986 INFO [train.py:904] (0/8) Epoch 8, batch 2600, loss[loss=0.1819, simple_loss=0.2604, pruned_loss=0.05168, over 16835.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2803, pruned_loss=0.0592, over 3331660.77 frames. ], batch size: 102, lr: 8.70e-03, grad_scale: 8.0 2023-04-28 20:05:55,331 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73668.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 20:05:56,977 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-28 20:06:43,000 INFO [optim.py:368] (0/8) 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 20:06:43,016 INFO [train.py:904] (0/8) Epoch 8, batch 2650, loss[loss=0.1829, simple_loss=0.2726, pruned_loss=0.04664, over 16012.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2816, pruned_loss=0.05902, over 3322866.50 frames. ], batch size: 35, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:23,357 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=73732.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:07:35,273 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6600, 3.7042, 4.0503, 3.9973, 4.0641, 3.7995, 3.8598, 3.7620], device='cuda:0'), covar=tensor([0.0335, 0.0536, 0.0409, 0.0466, 0.0406, 0.0350, 0.0671, 0.0487], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0305, 0.0303, 0.0290, 0.0344, 0.0320, 0.0423, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 20:07:41,856 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-28 20:07:50,957 INFO [train.py:904] (0/8) Epoch 8, batch 2700, loss[loss=0.1923, simple_loss=0.2786, pruned_loss=0.05302, over 16483.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2812, pruned_loss=0.05858, over 3321396.38 frames. ], batch size: 75, lr: 8.70e-03, grad_scale: 4.0 2023-04-28 20:07:55,736 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7850, 1.6084, 2.2061, 2.5881, 2.6813, 2.5955, 1.7991, 2.7933], device='cuda:0'), covar=tensor([0.0075, 0.0275, 0.0186, 0.0140, 0.0124, 0.0130, 0.0258, 0.0067], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0160, 0.0144, 0.0149, 0.0155, 0.0109, 0.0160, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 20:08:01,839 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6313, 2.7313, 1.7874, 2.7469, 2.1682, 2.8301, 1.9413, 2.4067], device='cuda:0'), covar=tensor([0.0222, 0.0349, 0.1236, 0.0215, 0.0703, 0.0458, 0.1135, 0.0558], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0163, 0.0180, 0.0108, 0.0164, 0.0202, 0.0190, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 20:08:38,971 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9982, 5.4168, 5.6565, 5.4245, 5.4440, 6.0415, 5.5435, 5.1975], device='cuda:0'), covar=tensor([0.0892, 0.1672, 0.1461, 0.1417, 0.2335, 0.0907, 0.1141, 0.2118], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0464, 0.0480, 0.0406, 0.0536, 0.0513, 0.0387, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:08:46,257 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=73793.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:08:57,115 INFO [optim.py:368] (0/8) 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 20:08:57,136 INFO [train.py:904] (0/8) Epoch 8, batch 2750, loss[loss=0.214, simple_loss=0.3048, pruned_loss=0.06158, over 17106.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2819, pruned_loss=0.05783, over 3325968.35 frames. ], batch size: 53, lr: 8.69e-03, grad_scale: 4.0 2023-04-28 20:09:58,642 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-28 20:10:05,128 INFO [train.py:904] (0/8) Epoch 8, batch 2800, loss[loss=0.1917, simple_loss=0.2664, pruned_loss=0.05853, over 16766.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2812, pruned_loss=0.0582, over 3334778.24 frames. ], batch size: 102, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:10:19,713 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-04-28 20:11:12,685 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1701, 4.4643, 2.2541, 4.7119, 3.1190, 4.6896, 2.7444, 3.4317], device='cuda:0'), covar=tensor([0.0176, 0.0242, 0.1460, 0.0187, 0.0688, 0.0362, 0.1213, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0166, 0.0182, 0.0110, 0.0166, 0.0205, 0.0192, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 20:11:14,610 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.850e+02 2.765e+02 3.378e+02 4.166e+02 1.030e+03, threshold=6.755e+02, percent-clipped=2.0 2023-04-28 20:11:14,632 INFO [train.py:904] (0/8) Epoch 8, batch 2850, loss[loss=0.1916, simple_loss=0.2689, pruned_loss=0.05717, over 16234.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2794, pruned_loss=0.05795, over 3337405.92 frames. ], batch size: 165, lr: 8.69e-03, grad_scale: 8.0 2023-04-28 20:11:38,804 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 20:12:01,555 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-28 20:12:16,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=73946.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:20,778 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=73949.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:12:22,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1946, 5.1905, 4.9820, 4.3921, 5.0753, 1.9833, 4.8028, 5.0370], device='cuda:0'), covar=tensor([0.0061, 0.0053, 0.0120, 0.0334, 0.0070, 0.1959, 0.0100, 0.0128], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0105, 0.0157, 0.0152, 0.0122, 0.0167, 0.0143, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:12:23,984 INFO [train.py:904] (0/8) Epoch 8, batch 2900, loss[loss=0.3027, simple_loss=0.3368, pruned_loss=0.1343, over 11807.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2776, pruned_loss=0.05765, over 3336930.96 frames. ], batch size: 247, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:13:28,886 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=73997.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:13:32,419 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-74000.pt 2023-04-28 20:13:38,231 INFO [optim.py:368] (0/8) 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 20:13:38,247 INFO [train.py:904] (0/8) Epoch 8, batch 2950, loss[loss=0.2201, simple_loss=0.2857, pruned_loss=0.07722, over 16729.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2773, pruned_loss=0.0589, over 3325311.88 frames. ], batch size: 134, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:14:02,695 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-04-28 20:14:46,255 INFO [train.py:904] (0/8) Epoch 8, batch 3000, loss[loss=0.1854, simple_loss=0.277, pruned_loss=0.04689, over 17190.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2769, pruned_loss=0.05851, over 3326727.24 frames. ], batch size: 46, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:14:46,255 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 20:14:55,853 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 20:15:13,404 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74064.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:15:41,954 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-28 20:15:45,915 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74088.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:16:06,767 INFO [optim.py:368] (0/8) 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 20:16:06,788 INFO [train.py:904] (0/8) Epoch 8, batch 3050, loss[loss=0.1947, simple_loss=0.287, pruned_loss=0.05119, over 17116.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2768, pruned_loss=0.05851, over 3330400.92 frames. ], batch size: 48, lr: 8.68e-03, grad_scale: 8.0 2023-04-28 20:16:38,676 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74125.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:17:13,378 INFO [train.py:904] (0/8) Epoch 8, batch 3100, loss[loss=0.1655, simple_loss=0.2475, pruned_loss=0.04178, over 16802.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2771, pruned_loss=0.05853, over 3336665.06 frames. ], batch size: 39, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:18:21,291 INFO [optim.py:368] (0/8) 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 20:18:21,307 INFO [train.py:904] (0/8) Epoch 8, batch 3150, loss[loss=0.1887, simple_loss=0.2637, pruned_loss=0.05679, over 16192.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2764, pruned_loss=0.05813, over 3339339.23 frames. ], batch size: 164, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:18:35,676 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2906, 3.2415, 3.4928, 2.4652, 3.1439, 3.5212, 3.2794, 2.0345], device='cuda:0'), covar=tensor([0.0307, 0.0100, 0.0031, 0.0225, 0.0068, 0.0054, 0.0047, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0117, 0.0065, 0.0063, 0.0116, 0.0069, 0.0079, 0.0071, 0.0111], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:19:14,381 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74240.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:23,299 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74246.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:19:32,380 INFO [train.py:904] (0/8) Epoch 8, batch 3200, loss[loss=0.2438, simple_loss=0.31, pruned_loss=0.08879, over 12311.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2746, pruned_loss=0.05738, over 3339800.90 frames. ], batch size: 246, lr: 8.67e-03, grad_scale: 8.0 2023-04-28 20:19:52,494 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7777, 1.5514, 2.1536, 2.5381, 2.6409, 2.5403, 1.6617, 2.6816], device='cuda:0'), covar=tensor([0.0078, 0.0274, 0.0196, 0.0177, 0.0140, 0.0139, 0.0278, 0.0080], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0162, 0.0147, 0.0150, 0.0156, 0.0111, 0.0160, 0.0101], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 20:20:30,672 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:40,644 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:20:41,307 INFO [train.py:904] (0/8) Epoch 8, batch 3250, loss[loss=0.1662, simple_loss=0.2643, pruned_loss=0.03402, over 17091.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.274, pruned_loss=0.05723, over 3335978.48 frames. ], batch size: 47, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:20:42,360 INFO [optim.py:368] (0/8) 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 20:21:12,993 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3260, 3.9490, 3.9616, 1.9900, 3.0348, 2.5113, 3.6407, 3.7664], device='cuda:0'), covar=tensor([0.0320, 0.0698, 0.0444, 0.1696, 0.0777, 0.0895, 0.0700, 0.1106], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0141, 0.0154, 0.0140, 0.0135, 0.0123, 0.0136, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 20:21:52,436 INFO [train.py:904] (0/8) Epoch 8, batch 3300, loss[loss=0.2321, simple_loss=0.2981, pruned_loss=0.08308, over 16945.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2765, pruned_loss=0.05871, over 3324268.07 frames. ], batch size: 116, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:22:36,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3855, 4.1316, 3.9406, 1.9632, 3.0270, 2.6465, 3.7235, 4.0754], device='cuda:0'), covar=tensor([0.0302, 0.0656, 0.0527, 0.1772, 0.0809, 0.0924, 0.0723, 0.0858], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0142, 0.0154, 0.0139, 0.0135, 0.0123, 0.0136, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 20:22:41,330 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74388.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:01,525 INFO [train.py:904] (0/8) Epoch 8, batch 3350, loss[loss=0.1997, simple_loss=0.2886, pruned_loss=0.05542, over 16716.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2773, pruned_loss=0.05849, over 3322955.93 frames. ], batch size: 57, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:23:02,761 INFO [optim.py:368] (0/8) 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 20:23:27,945 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74420.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:23:49,805 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74436.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:24:11,180 INFO [train.py:904] (0/8) Epoch 8, batch 3400, loss[loss=0.2177, simple_loss=0.2852, pruned_loss=0.07505, over 16854.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.278, pruned_loss=0.05857, over 3320352.39 frames. ], batch size: 116, lr: 8.66e-03, grad_scale: 4.0 2023-04-28 20:24:16,584 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4317, 5.3869, 5.2786, 4.9337, 4.8391, 5.3119, 5.3120, 4.9535], device='cuda:0'), covar=tensor([0.0461, 0.0295, 0.0231, 0.0211, 0.0985, 0.0326, 0.0193, 0.0522], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0287, 0.0283, 0.0253, 0.0316, 0.0286, 0.0192, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:25:01,059 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9885, 3.8220, 3.9544, 4.1851, 4.2495, 3.8277, 4.0131, 4.2209], device='cuda:0'), covar=tensor([0.1039, 0.0875, 0.1309, 0.0507, 0.0530, 0.1477, 0.1275, 0.0566], device='cuda:0'), in_proj_covar=tensor([0.0509, 0.0627, 0.0785, 0.0638, 0.0483, 0.0485, 0.0497, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:25:21,723 INFO [train.py:904] (0/8) Epoch 8, batch 3450, loss[loss=0.1664, simple_loss=0.249, pruned_loss=0.04188, over 17215.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2764, pruned_loss=0.05814, over 3322871.68 frames. ], batch size: 44, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:25:22,842 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.611e+02 3.016e+02 3.656e+02 6.729e+02, threshold=6.032e+02, percent-clipped=3.0 2023-04-28 20:26:30,771 INFO [train.py:904] (0/8) Epoch 8, batch 3500, loss[loss=0.154, simple_loss=0.2389, pruned_loss=0.03459, over 16854.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2749, pruned_loss=0.05694, over 3331338.40 frames. ], batch size: 42, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:26:35,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8621, 2.0166, 2.2695, 3.1552, 2.0694, 2.2695, 2.2446, 2.0439], device='cuda:0'), covar=tensor([0.0759, 0.2500, 0.1418, 0.0446, 0.2926, 0.1651, 0.2131, 0.2625], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0370, 0.0310, 0.0331, 0.0399, 0.0416, 0.0331, 0.0440], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:26:58,729 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-28 20:27:31,844 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74595.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:33,599 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74596.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:27:42,854 INFO [train.py:904] (0/8) Epoch 8, batch 3550, loss[loss=0.2418, simple_loss=0.3006, pruned_loss=0.09154, over 16888.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2742, pruned_loss=0.05666, over 3326107.81 frames. ], batch size: 116, lr: 8.65e-03, grad_scale: 4.0 2023-04-28 20:27:43,962 INFO [optim.py:368] (0/8) 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 20:28:51,895 INFO [train.py:904] (0/8) Epoch 8, batch 3600, loss[loss=0.198, simple_loss=0.2689, pruned_loss=0.06352, over 16463.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2742, pruned_loss=0.057, over 3324239.99 frames. ], batch size: 146, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:28:56,979 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 8, batch 3650, loss[loss=0.1959, simple_loss=0.2625, pruned_loss=0.06466, over 16485.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2729, pruned_loss=0.05736, over 3319762.04 frames. ], batch size: 75, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:30:02,111 INFO [optim.py:368] (0/8) 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 20:30:27,787 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:31:07,134 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1840, 3.5071, 3.2050, 2.0333, 2.7810, 2.3924, 3.5904, 3.4479], device='cuda:0'), covar=tensor([0.0209, 0.0610, 0.0510, 0.1467, 0.0717, 0.0826, 0.0497, 0.0749], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0142, 0.0154, 0.0139, 0.0134, 0.0123, 0.0135, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 20:31:13,893 INFO [train.py:904] (0/8) Epoch 8, batch 3700, loss[loss=0.1976, simple_loss=0.2749, pruned_loss=0.06013, over 15366.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2716, pruned_loss=0.05871, over 3314524.27 frames. ], batch size: 190, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:31:38,747 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74768.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:31:44,037 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6229, 2.6133, 1.7943, 2.6003, 2.1511, 2.7087, 2.0512, 2.3417], device='cuda:0'), covar=tensor([0.0224, 0.0327, 0.1326, 0.0159, 0.0657, 0.0381, 0.1028, 0.0504], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0163, 0.0181, 0.0111, 0.0165, 0.0205, 0.0190, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 20:32:29,723 INFO [train.py:904] (0/8) Epoch 8, batch 3750, loss[loss=0.177, simple_loss=0.263, pruned_loss=0.04551, over 17230.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2724, pruned_loss=0.06003, over 3304526.09 frames. ], batch size: 45, lr: 8.64e-03, grad_scale: 8.0 2023-04-28 20:32:30,693 INFO [optim.py:368] (0/8) 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 20:33:41,202 INFO [train.py:904] (0/8) Epoch 8, batch 3800, loss[loss=0.209, simple_loss=0.2952, pruned_loss=0.06134, over 16379.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2735, pruned_loss=0.06197, over 3301265.92 frames. ], batch size: 35, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:21,837 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0338, 1.9539, 1.5729, 1.7231, 2.2808, 1.9928, 2.2201, 2.3835], device='cuda:0'), covar=tensor([0.0118, 0.0227, 0.0307, 0.0299, 0.0135, 0.0230, 0.0134, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0185, 0.0181, 0.0182, 0.0182, 0.0186, 0.0188, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:34:44,764 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=74896.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:34:51,955 INFO [train.py:904] (0/8) Epoch 8, batch 3850, loss[loss=0.1959, simple_loss=0.2661, pruned_loss=0.06285, over 16905.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2737, pruned_loss=0.06259, over 3289221.02 frames. ], batch size: 96, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:34:53,140 INFO [optim.py:368] (0/8) 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 20:34:55,854 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6130, 4.6385, 4.7328, 4.7912, 4.6931, 5.1951, 4.8246, 4.5428], device='cuda:0'), covar=tensor([0.1238, 0.1518, 0.1365, 0.1598, 0.2528, 0.0955, 0.1090, 0.2094], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0460, 0.0476, 0.0403, 0.0528, 0.0503, 0.0383, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:35:03,437 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74909.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:35:52,062 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=74944.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:01,461 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=74951.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:02,293 INFO [train.py:904] (0/8) Epoch 8, batch 3900, loss[loss=0.191, simple_loss=0.262, pruned_loss=0.06, over 16851.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2727, pruned_loss=0.06245, over 3296895.16 frames. ], batch size: 102, lr: 8.63e-03, grad_scale: 8.0 2023-04-28 20:36:20,361 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=74964.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:36:27,533 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=74970.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:37:12,360 INFO [train.py:904] (0/8) Epoch 8, batch 3950, loss[loss=0.201, simple_loss=0.2594, pruned_loss=0.07124, over 16728.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2716, pruned_loss=0.06283, over 3304965.80 frames. ], batch size: 134, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:37:14,096 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 2.648e+02 3.131e+02 3.735e+02 8.073e+02, threshold=6.262e+02, percent-clipped=3.0 2023-04-28 20:37:46,142 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75025.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:38:23,638 INFO [train.py:904] (0/8) Epoch 8, batch 4000, loss[loss=0.1676, simple_loss=0.2487, pruned_loss=0.04322, over 16892.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2722, pruned_loss=0.06343, over 3307929.04 frames. ], batch size: 42, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:38:38,860 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.17 vs. limit=5.0 2023-04-28 20:38:50,688 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1568, 5.1775, 4.9387, 4.6820, 4.6050, 5.0172, 4.9271, 4.7300], device='cuda:0'), covar=tensor([0.0526, 0.0245, 0.0207, 0.0253, 0.0884, 0.0288, 0.0318, 0.0586], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0274, 0.0272, 0.0243, 0.0303, 0.0278, 0.0186, 0.0311], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:39:36,990 INFO [train.py:904] (0/8) Epoch 8, batch 4050, loss[loss=0.2008, simple_loss=0.2807, pruned_loss=0.06042, over 16748.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2726, pruned_loss=0.06235, over 3296680.66 frames. ], batch size: 124, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:39:38,165 INFO [optim.py:368] (0/8) 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 20:39:42,414 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5497, 5.5072, 5.3617, 5.0945, 4.9499, 5.3940, 5.3208, 5.1197], device='cuda:0'), covar=tensor([0.0488, 0.0251, 0.0186, 0.0194, 0.0901, 0.0295, 0.0224, 0.0487], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0271, 0.0269, 0.0241, 0.0301, 0.0276, 0.0184, 0.0308], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:40:49,058 INFO [train.py:904] (0/8) Epoch 8, batch 4100, loss[loss=0.2187, simple_loss=0.2998, pruned_loss=0.06879, over 16246.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2738, pruned_loss=0.06156, over 3288886.78 frames. ], batch size: 165, lr: 8.62e-03, grad_scale: 8.0 2023-04-28 20:41:00,949 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9472, 4.0534, 3.7908, 3.6358, 3.4865, 3.9582, 3.5848, 3.6856], device='cuda:0'), covar=tensor([0.0556, 0.0410, 0.0253, 0.0229, 0.0677, 0.0387, 0.0968, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0269, 0.0267, 0.0239, 0.0298, 0.0274, 0.0182, 0.0305], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:41:26,778 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5730, 2.1877, 2.2709, 4.3515, 2.0747, 2.7241, 2.3112, 2.4561], device='cuda:0'), covar=tensor([0.0735, 0.2820, 0.1712, 0.0286, 0.3175, 0.1723, 0.2362, 0.2475], device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0375, 0.0312, 0.0329, 0.0399, 0.0424, 0.0337, 0.0444], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:41:44,533 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.40 vs. limit=5.0 2023-04-28 20:42:02,429 INFO [train.py:904] (0/8) Epoch 8, batch 4150, loss[loss=0.2676, simple_loss=0.3274, pruned_loss=0.1039, over 11162.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2819, pruned_loss=0.06455, over 3247024.39 frames. ], batch size: 248, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:42:04,252 INFO [optim.py:368] (0/8) 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 20:42:42,701 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75228.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:06,710 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7258, 4.5095, 4.7315, 4.9072, 5.0066, 4.4287, 5.0090, 5.0008], device='cuda:0'), covar=tensor([0.1047, 0.0871, 0.1140, 0.0475, 0.0379, 0.0755, 0.0393, 0.0386], device='cuda:0'), in_proj_covar=tensor([0.0473, 0.0579, 0.0720, 0.0587, 0.0444, 0.0447, 0.0452, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:43:18,736 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75251.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:20,236 INFO [train.py:904] (0/8) Epoch 8, batch 4200, loss[loss=0.2323, simple_loss=0.3213, pruned_loss=0.07171, over 15391.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.289, pruned_loss=0.06679, over 3204653.94 frames. ], batch size: 190, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:43:40,419 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75265.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:47,269 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=75270.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:43:52,702 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 20:44:16,449 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:30,196 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:44:34,933 INFO [train.py:904] (0/8) Epoch 8, batch 4250, loss[loss=0.2069, simple_loss=0.2964, pruned_loss=0.05866, over 17246.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2926, pruned_loss=0.06657, over 3198564.49 frames. ], batch size: 52, lr: 8.61e-03, grad_scale: 8.0 2023-04-28 20:44:36,197 INFO [optim.py:368] (0/8) 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 20:45:01,393 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=75331.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 20:45:31,062 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2486, 4.3521, 4.1487, 3.9684, 3.7367, 4.2354, 3.9299, 3.9088], device='cuda:0'), covar=tensor([0.0552, 0.0265, 0.0227, 0.0219, 0.0872, 0.0265, 0.0605, 0.0521], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0258, 0.0258, 0.0229, 0.0286, 0.0260, 0.0176, 0.0291], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:45:32,451 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8235, 3.5163, 3.2726, 1.8227, 2.7695, 2.1442, 3.2979, 3.4165], device='cuda:0'), covar=tensor([0.0243, 0.0506, 0.0531, 0.1724, 0.0747, 0.0952, 0.0604, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0135, 0.0152, 0.0137, 0.0132, 0.0122, 0.0133, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 20:45:33,498 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6490, 2.3143, 2.1180, 4.5082, 2.1309, 2.8343, 2.4105, 2.5589], device='cuda:0'), covar=tensor([0.0676, 0.2652, 0.1721, 0.0260, 0.3045, 0.1675, 0.2337, 0.2474], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0366, 0.0305, 0.0322, 0.0392, 0.0410, 0.0328, 0.0434], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:45:48,491 INFO [train.py:904] (0/8) Epoch 8, batch 4300, loss[loss=0.24, simple_loss=0.3239, pruned_loss=0.0781, over 16541.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2943, pruned_loss=0.06582, over 3196412.47 frames. ], batch size: 62, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:47:02,669 INFO [train.py:904] (0/8) Epoch 8, batch 4350, loss[loss=0.2385, simple_loss=0.3172, pruned_loss=0.07996, over 16451.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2982, pruned_loss=0.06717, over 3191412.71 frames. ], batch size: 146, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:47:03,853 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.683e+02 2.615e+02 3.065e+02 3.856e+02 8.729e+02, threshold=6.129e+02, percent-clipped=2.0 2023-04-28 20:47:19,140 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3771, 1.9855, 1.5561, 1.8794, 2.3251, 2.1165, 2.5710, 2.6218], device='cuda:0'), covar=tensor([0.0086, 0.0255, 0.0325, 0.0313, 0.0137, 0.0243, 0.0116, 0.0137], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0182, 0.0178, 0.0179, 0.0180, 0.0183, 0.0179, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:48:17,344 INFO [train.py:904] (0/8) Epoch 8, batch 4400, loss[loss=0.2042, simple_loss=0.2906, pruned_loss=0.05889, over 16748.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3002, pruned_loss=0.06839, over 3163332.28 frames. ], batch size: 83, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:48:38,161 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0028, 3.5353, 3.4807, 2.2459, 3.2630, 3.5142, 3.3921, 1.6783], device='cuda:0'), covar=tensor([0.0415, 0.0024, 0.0028, 0.0303, 0.0049, 0.0053, 0.0035, 0.0364], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0064, 0.0063, 0.0119, 0.0069, 0.0077, 0.0070, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:49:26,874 INFO [train.py:904] (0/8) Epoch 8, batch 4450, loss[loss=0.2345, simple_loss=0.3175, pruned_loss=0.07577, over 15388.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3025, pruned_loss=0.06857, over 3181690.63 frames. ], batch size: 190, lr: 8.60e-03, grad_scale: 8.0 2023-04-28 20:49:28,914 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.480e+02 2.422e+02 2.913e+02 3.508e+02 6.103e+02, threshold=5.826e+02, percent-clipped=0.0 2023-04-28 20:50:38,188 INFO [train.py:904] (0/8) Epoch 8, batch 4500, loss[loss=0.2455, simple_loss=0.3199, pruned_loss=0.08552, over 16424.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3028, pruned_loss=0.06882, over 3183989.18 frames. ], batch size: 146, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:50:57,752 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75565.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:24,927 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75584.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:51:27,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6158, 4.5310, 4.3814, 3.7762, 4.5101, 1.6788, 4.2136, 4.1097], device='cuda:0'), covar=tensor([0.0047, 0.0037, 0.0086, 0.0255, 0.0042, 0.2044, 0.0080, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0100, 0.0148, 0.0145, 0.0116, 0.0160, 0.0135, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:51:41,127 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-28 20:51:51,104 INFO [train.py:904] (0/8) Epoch 8, batch 4550, loss[loss=0.2081, simple_loss=0.2934, pruned_loss=0.06143, over 16765.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3031, pruned_loss=0.06932, over 3186772.88 frames. ], batch size: 124, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:51:52,275 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.538e+02 2.189e+02 2.672e+02 3.111e+02 5.807e+02, threshold=5.345e+02, percent-clipped=0.0 2023-04-28 20:52:06,458 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75613.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:17,126 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75620.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:52:24,434 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=75626.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 20:52:52,908 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3938, 3.5382, 3.2589, 3.0926, 2.9871, 3.3599, 3.1808, 3.1116], device='cuda:0'), covar=tensor([0.0522, 0.0283, 0.0255, 0.0226, 0.0644, 0.0303, 0.1489, 0.0548], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0251, 0.0252, 0.0225, 0.0280, 0.0253, 0.0173, 0.0285], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:53:02,645 INFO [train.py:904] (0/8) Epoch 8, batch 4600, loss[loss=0.2295, simple_loss=0.3095, pruned_loss=0.07479, over 15366.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3036, pruned_loss=0.06939, over 3190573.82 frames. ], batch size: 190, lr: 8.59e-03, grad_scale: 8.0 2023-04-28 20:53:25,294 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75668.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:53:31,646 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 20:54:08,974 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-28 20:54:12,069 INFO [train.py:904] (0/8) Epoch 8, batch 4650, loss[loss=0.2115, simple_loss=0.2939, pruned_loss=0.06459, over 16942.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3021, pruned_loss=0.0692, over 3200116.17 frames. ], batch size: 109, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:54:13,263 INFO [optim.py:368] (0/8) 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 20:55:23,494 INFO [train.py:904] (0/8) Epoch 8, batch 4700, loss[loss=0.2013, simple_loss=0.281, pruned_loss=0.06076, over 17029.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.299, pruned_loss=0.06772, over 3200347.50 frames. ], batch size: 53, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:55:33,642 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2860, 4.0943, 4.2239, 2.7096, 3.6701, 4.0521, 3.7781, 2.1563], device='cuda:0'), covar=tensor([0.0364, 0.0020, 0.0015, 0.0250, 0.0045, 0.0063, 0.0040, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0063, 0.0062, 0.0119, 0.0068, 0.0078, 0.0070, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 20:56:31,939 INFO [train.py:904] (0/8) Epoch 8, batch 4750, loss[loss=0.2037, simple_loss=0.285, pruned_loss=0.06117, over 15332.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2947, pruned_loss=0.06551, over 3206802.96 frames. ], batch size: 190, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:56:33,073 INFO [optim.py:368] (0/8) 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 20:56:35,880 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.09 vs. limit=5.0 2023-04-28 20:57:20,500 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1088, 3.2670, 3.3711, 1.5639, 3.6291, 3.6702, 2.7327, 2.6254], device='cuda:0'), covar=tensor([0.0921, 0.0187, 0.0177, 0.1315, 0.0063, 0.0080, 0.0413, 0.0486], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0098, 0.0084, 0.0139, 0.0068, 0.0090, 0.0118, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 20:57:22,548 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 20:57:44,152 INFO [train.py:904] (0/8) Epoch 8, batch 4800, loss[loss=0.244, simple_loss=0.3135, pruned_loss=0.08728, over 11971.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2916, pruned_loss=0.06336, over 3206627.93 frames. ], batch size: 246, lr: 8.58e-03, grad_scale: 8.0 2023-04-28 20:58:32,003 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75884.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 20:58:57,793 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1805, 3.2225, 3.3040, 1.5792, 3.5475, 3.5527, 2.7082, 2.5469], device='cuda:0'), covar=tensor([0.0717, 0.0173, 0.0152, 0.1172, 0.0058, 0.0088, 0.0344, 0.0453], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0098, 0.0084, 0.0139, 0.0068, 0.0091, 0.0118, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 20:58:58,551 INFO [train.py:904] (0/8) Epoch 8, batch 4850, loss[loss=0.2248, simple_loss=0.324, pruned_loss=0.06284, over 16178.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2924, pruned_loss=0.06312, over 3182635.76 frames. ], batch size: 165, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 20:59:01,505 INFO [optim.py:368] (0/8) 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 20:59:23,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0527, 2.2987, 1.7922, 2.0245, 2.6592, 2.4194, 2.8921, 2.9143], device='cuda:0'), covar=tensor([0.0055, 0.0265, 0.0347, 0.0329, 0.0163, 0.0242, 0.0122, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0180, 0.0178, 0.0179, 0.0178, 0.0183, 0.0178, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 20:59:36,805 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=75926.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 20:59:45,980 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75932.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:00:17,706 INFO [train.py:904] (0/8) Epoch 8, batch 4900, loss[loss=0.1945, simple_loss=0.2822, pruned_loss=0.05344, over 16672.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2911, pruned_loss=0.06162, over 3177383.01 frames. ], batch size: 134, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:00:20,918 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 21:00:26,963 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7075, 3.4006, 3.0822, 1.7534, 2.6645, 2.2485, 3.3448, 3.3929], device='cuda:0'), covar=tensor([0.0217, 0.0470, 0.0517, 0.1593, 0.0707, 0.0811, 0.0572, 0.0606], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0134, 0.0153, 0.0139, 0.0131, 0.0122, 0.0133, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:00:38,884 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-28 21:00:49,333 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=75974.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:01:28,725 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-76000.pt 2023-04-28 21:01:34,533 INFO [train.py:904] (0/8) Epoch 8, batch 4950, loss[loss=0.2017, simple_loss=0.2918, pruned_loss=0.05574, over 16730.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2905, pruned_loss=0.06089, over 3201135.34 frames. ], batch size: 124, lr: 8.57e-03, grad_scale: 8.0 2023-04-28 21:01:36,820 INFO [optim.py:368] (0/8) 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 21:02:03,120 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0601, 5.5395, 5.8318, 5.6385, 5.6515, 6.1828, 5.8051, 5.5343], device='cuda:0'), covar=tensor([0.0627, 0.1594, 0.1315, 0.1320, 0.2022, 0.0713, 0.0939, 0.1712], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0428, 0.0445, 0.0378, 0.0503, 0.0473, 0.0360, 0.0513], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 21:02:45,331 INFO [train.py:904] (0/8) Epoch 8, batch 5000, loss[loss=0.2051, simple_loss=0.3003, pruned_loss=0.05492, over 16706.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2928, pruned_loss=0.06146, over 3202281.82 frames. ], batch size: 124, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:55,764 INFO [train.py:904] (0/8) Epoch 8, batch 5050, loss[loss=0.188, simple_loss=0.2697, pruned_loss=0.05312, over 17160.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2933, pruned_loss=0.06127, over 3209435.11 frames. ], batch size: 47, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:03:57,928 INFO [optim.py:368] (0/8) 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] (0/8) attn_weights_entropy = tensor([3.7972, 3.6524, 3.8032, 3.9304, 4.0338, 3.6114, 3.9988, 4.0253], device='cuda:0'), covar=tensor([0.1005, 0.0854, 0.1147, 0.0529, 0.0439, 0.1670, 0.0491, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0468, 0.0571, 0.0713, 0.0583, 0.0439, 0.0443, 0.0447, 0.0497], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:05:07,315 INFO [train.py:904] (0/8) Epoch 8, batch 5100, loss[loss=0.2007, simple_loss=0.287, pruned_loss=0.05721, over 15392.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.291, pruned_loss=0.06024, over 3209545.52 frames. ], batch size: 190, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:20,875 INFO [train.py:904] (0/8) Epoch 8, batch 5150, loss[loss=0.197, simple_loss=0.2902, pruned_loss=0.05194, over 16485.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2909, pruned_loss=0.05928, over 3202812.92 frames. ], batch size: 75, lr: 8.56e-03, grad_scale: 8.0 2023-04-28 21:06:24,101 INFO [optim.py:368] (0/8) 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 21:07:26,976 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2990, 1.9991, 2.2112, 4.0461, 1.8915, 2.5123, 2.1111, 2.2326], device='cuda:0'), covar=tensor([0.0901, 0.2995, 0.1789, 0.0344, 0.3501, 0.1957, 0.2712, 0.2635], device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0364, 0.0307, 0.0321, 0.0395, 0.0407, 0.0327, 0.0433], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:07:33,726 INFO [train.py:904] (0/8) Epoch 8, batch 5200, loss[loss=0.2036, simple_loss=0.2838, pruned_loss=0.06169, over 16910.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2888, pruned_loss=0.05864, over 3208552.05 frames. ], batch size: 109, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:07:59,748 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76269.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:08:15,466 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 21:08:35,853 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76294.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:08:48,742 INFO [train.py:904] (0/8) Epoch 8, batch 5250, loss[loss=0.203, simple_loss=0.3018, pruned_loss=0.0521, over 16870.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2866, pruned_loss=0.05837, over 3203328.01 frames. ], batch size: 96, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:08:51,148 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.594e+02 2.354e+02 2.757e+02 3.436e+02 6.643e+02, threshold=5.515e+02, percent-clipped=1.0 2023-04-28 21:09:19,339 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76323.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:09:30,938 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76330.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:10:01,836 INFO [train.py:904] (0/8) Epoch 8, batch 5300, loss[loss=0.214, simple_loss=0.2873, pruned_loss=0.07036, over 11896.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2829, pruned_loss=0.05725, over 3191570.84 frames. ], batch size: 246, lr: 8.55e-03, grad_scale: 8.0 2023-04-28 21:10:06,393 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76355.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:10:48,042 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76384.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 21:10:50,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6223, 4.7490, 4.5350, 4.2875, 3.8680, 4.5986, 4.4525, 4.2614], device='cuda:0'), covar=tensor([0.0603, 0.0414, 0.0288, 0.0241, 0.1254, 0.0399, 0.0340, 0.0605], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0261, 0.0259, 0.0231, 0.0290, 0.0265, 0.0175, 0.0299], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:11:13,479 INFO [train.py:904] (0/8) Epoch 8, batch 5350, loss[loss=0.235, simple_loss=0.3013, pruned_loss=0.08435, over 12013.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.282, pruned_loss=0.05669, over 3193412.52 frames. ], batch size: 248, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:11:15,925 INFO [optim.py:368] (0/8) 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 21:11:56,110 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4361, 3.5090, 2.7541, 2.0660, 2.5580, 2.2113, 3.6955, 3.4016], device='cuda:0'), covar=tensor([0.2338, 0.0673, 0.1369, 0.1956, 0.1973, 0.1567, 0.0432, 0.0856], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0249, 0.0276, 0.0263, 0.0278, 0.0210, 0.0259, 0.0276], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:12:26,520 INFO [train.py:904] (0/8) Epoch 8, batch 5400, loss[loss=0.2379, simple_loss=0.3145, pruned_loss=0.08066, over 12013.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2845, pruned_loss=0.05751, over 3199428.77 frames. ], batch size: 246, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:33,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9655, 3.5086, 3.2270, 1.8017, 2.7766, 2.2891, 3.3294, 3.3713], device='cuda:0'), covar=tensor([0.0215, 0.0456, 0.0519, 0.1595, 0.0695, 0.0805, 0.0530, 0.0759], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0132, 0.0154, 0.0139, 0.0132, 0.0121, 0.0134, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:13:43,650 INFO [train.py:904] (0/8) Epoch 8, batch 5450, loss[loss=0.2269, simple_loss=0.3111, pruned_loss=0.07134, over 16470.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2881, pruned_loss=0.0594, over 3189177.50 frames. ], batch size: 146, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:13:46,711 INFO [optim.py:368] (0/8) 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 21:13:49,191 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9060, 5.3048, 4.6801, 5.1665, 4.6729, 4.5159, 5.0083, 5.3478], device='cuda:0'), covar=tensor([0.1844, 0.1464, 0.2551, 0.1128, 0.1373, 0.1476, 0.1527, 0.1604], device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0584, 0.0493, 0.0402, 0.0373, 0.0382, 0.0488, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:13:53,358 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.8227, 6.1418, 5.8056, 5.9757, 5.4670, 5.2439, 5.6970, 6.2877], device='cuda:0'), covar=tensor([0.0756, 0.0631, 0.0969, 0.0544, 0.0618, 0.0554, 0.0664, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0463, 0.0584, 0.0492, 0.0402, 0.0373, 0.0382, 0.0488, 0.0428], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:14:08,667 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9448, 2.4791, 2.2796, 2.9607, 2.3586, 3.3079, 1.7262, 2.7192], device='cuda:0'), covar=tensor([0.1105, 0.0501, 0.0958, 0.0131, 0.0160, 0.0356, 0.1245, 0.0612], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0153, 0.0175, 0.0118, 0.0201, 0.0204, 0.0172, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 21:14:50,992 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-28 21:15:01,752 INFO [train.py:904] (0/8) Epoch 8, batch 5500, loss[loss=0.2593, simple_loss=0.3397, pruned_loss=0.08942, over 16806.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2973, pruned_loss=0.0654, over 3168744.94 frames. ], batch size: 83, lr: 8.54e-03, grad_scale: 8.0 2023-04-28 21:16:16,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4581, 3.5000, 3.1981, 2.9756, 3.0419, 3.3081, 3.2741, 3.1061], device='cuda:0'), covar=tensor([0.0536, 0.0418, 0.0221, 0.0197, 0.0514, 0.0334, 0.0857, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0259, 0.0255, 0.0226, 0.0285, 0.0263, 0.0173, 0.0295], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:16:22,254 INFO [train.py:904] (0/8) Epoch 8, batch 5550, loss[loss=0.2344, simple_loss=0.3074, pruned_loss=0.08073, over 16616.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3055, pruned_loss=0.07153, over 3146589.44 frames. ], batch size: 57, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:16:26,052 INFO [optim.py:368] (0/8) 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 21:17:00,122 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76625.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 21:17:00,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0067, 4.9541, 5.5039, 5.4867, 5.5092, 5.0703, 5.0570, 4.5864], device='cuda:0'), covar=tensor([0.0225, 0.0439, 0.0304, 0.0368, 0.0357, 0.0297, 0.0785, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0288, 0.0292, 0.0279, 0.0326, 0.0309, 0.0412, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 21:17:40,838 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76650.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:17:43,818 INFO [train.py:904] (0/8) Epoch 8, batch 5600, loss[loss=0.2758, simple_loss=0.3446, pruned_loss=0.1035, over 15128.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3109, pruned_loss=0.07602, over 3130052.38 frames. ], batch size: 190, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:18:12,609 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 21:18:28,373 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=76679.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:19:06,823 INFO [train.py:904] (0/8) Epoch 8, batch 5650, loss[loss=0.2871, simple_loss=0.3415, pruned_loss=0.1164, over 11872.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3173, pruned_loss=0.08168, over 3089022.88 frames. ], batch size: 247, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:19:10,210 INFO [optim.py:368] (0/8) 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 21:20:27,963 INFO [train.py:904] (0/8) Epoch 8, batch 5700, loss[loss=0.2417, simple_loss=0.3239, pruned_loss=0.0798, over 16645.00 frames. ], tot_loss[loss=0.2416, simple_loss=0.3179, pruned_loss=0.08261, over 3080588.36 frames. ], batch size: 62, lr: 8.53e-03, grad_scale: 8.0 2023-04-28 21:21:49,371 INFO [train.py:904] (0/8) Epoch 8, batch 5750, loss[loss=0.2318, simple_loss=0.3175, pruned_loss=0.07302, over 16743.00 frames. ], tot_loss[loss=0.2458, simple_loss=0.3215, pruned_loss=0.08501, over 3039295.67 frames. ], batch size: 89, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:21:54,077 INFO [optim.py:368] (0/8) 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 21:22:51,979 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=76839.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:23:12,553 INFO [train.py:904] (0/8) Epoch 8, batch 5800, loss[loss=0.2297, simple_loss=0.3076, pruned_loss=0.07588, over 16244.00 frames. ], tot_loss[loss=0.2449, simple_loss=0.3217, pruned_loss=0.08406, over 3030295.89 frames. ], batch size: 35, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:23:24,773 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9540, 2.0155, 2.3285, 3.2255, 2.1162, 2.2918, 2.2033, 2.0764], device='cuda:0'), covar=tensor([0.0831, 0.2654, 0.1463, 0.0484, 0.3171, 0.1833, 0.2392, 0.2785], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0362, 0.0304, 0.0317, 0.0393, 0.0401, 0.0323, 0.0427], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:24:04,070 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-28 21:24:30,498 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=76900.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:24:33,004 INFO [train.py:904] (0/8) Epoch 8, batch 5850, loss[loss=0.215, simple_loss=0.2989, pruned_loss=0.06558, over 16751.00 frames. ], tot_loss[loss=0.2409, simple_loss=0.3188, pruned_loss=0.08154, over 3051564.92 frames. ], batch size: 124, lr: 8.52e-03, grad_scale: 4.0 2023-04-28 21:24:37,989 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 3.434e+02 4.196e+02 5.124e+02 1.086e+03, threshold=8.393e+02, percent-clipped=1.0 2023-04-28 21:25:09,380 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76925.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 21:25:44,601 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7073, 3.3646, 3.1993, 1.8527, 2.7753, 2.2347, 3.2892, 3.2799], device='cuda:0'), covar=tensor([0.0202, 0.0465, 0.0515, 0.1645, 0.0675, 0.0867, 0.0548, 0.0755], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0132, 0.0154, 0.0139, 0.0133, 0.0122, 0.0134, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:25:52,720 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76950.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:25:56,483 INFO [train.py:904] (0/8) Epoch 8, batch 5900, loss[loss=0.2192, simple_loss=0.3042, pruned_loss=0.06715, over 16924.00 frames. ], tot_loss[loss=0.2396, simple_loss=0.3178, pruned_loss=0.08075, over 3058793.54 frames. ], batch size: 109, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:26:33,747 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=76973.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:26:42,694 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=76979.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:27:06,866 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-28 21:27:10,471 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 8, batch 5950, loss[loss=0.2387, simple_loss=0.3254, pruned_loss=0.07597, over 17010.00 frames. ], tot_loss[loss=0.2383, simple_loss=0.3183, pruned_loss=0.07919, over 3078899.35 frames. ], batch size: 41, lr: 8.51e-03, grad_scale: 4.0 2023-04-28 21:27:21,565 INFO [optim.py:368] (0/8) 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 21:27:22,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5717, 5.4877, 5.2514, 4.6619, 5.4203, 1.9014, 5.1169, 5.3140], device='cuda:0'), covar=tensor([0.0051, 0.0041, 0.0102, 0.0315, 0.0051, 0.1976, 0.0081, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0097, 0.0146, 0.0141, 0.0114, 0.0161, 0.0131, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:27:55,399 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77027.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:28:33,578 INFO [train.py:904] (0/8) Epoch 8, batch 6000, loss[loss=0.2047, simple_loss=0.2906, pruned_loss=0.0594, over 16702.00 frames. ], tot_loss[loss=0.2378, simple_loss=0.3174, pruned_loss=0.07909, over 3081327.79 frames. ], batch size: 124, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:28:33,579 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 21:28:44,113 INFO [train.py:938] (0/8) Epoch 8, validation: loss=0.1707, simple_loss=0.284, pruned_loss=0.02871, over 944034.00 frames. 2023-04-28 21:28:44,113 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 21:29:48,607 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77095.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:30:00,318 INFO [train.py:904] (0/8) Epoch 8, batch 6050, loss[loss=0.2035, simple_loss=0.3008, pruned_loss=0.05306, over 16566.00 frames. ], tot_loss[loss=0.2369, simple_loss=0.3165, pruned_loss=0.0786, over 3103382.99 frames. ], batch size: 68, lr: 8.51e-03, grad_scale: 8.0 2023-04-28 21:30:04,233 INFO [optim.py:368] (0/8) 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 21:30:46,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8458, 5.3024, 5.4977, 5.3391, 5.3134, 5.8705, 5.4185, 5.2086], device='cuda:0'), covar=tensor([0.0903, 0.1694, 0.1553, 0.1729, 0.2517, 0.0896, 0.1237, 0.2268], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0431, 0.0455, 0.0383, 0.0509, 0.0479, 0.0369, 0.0518], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 21:30:53,650 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 21:31:19,269 INFO [train.py:904] (0/8) Epoch 8, batch 6100, loss[loss=0.205, simple_loss=0.2897, pruned_loss=0.06015, over 17272.00 frames. ], tot_loss[loss=0.2355, simple_loss=0.3158, pruned_loss=0.07758, over 3109496.31 frames. ], batch size: 52, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:31:26,129 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77156.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:31:47,762 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-04-28 21:32:26,665 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77195.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:32:37,679 INFO [train.py:904] (0/8) Epoch 8, batch 6150, loss[loss=0.2276, simple_loss=0.3084, pruned_loss=0.07336, over 16737.00 frames. ], tot_loss[loss=0.2338, simple_loss=0.3135, pruned_loss=0.07704, over 3107689.44 frames. ], batch size: 124, lr: 8.50e-03, grad_scale: 8.0 2023-04-28 21:32:42,675 INFO [optim.py:368] (0/8) 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 21:32:52,983 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77211.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:32:56,382 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5510, 5.5453, 5.2269, 4.6353, 5.3250, 2.0338, 5.1415, 5.2833], device='cuda:0'), covar=tensor([0.0050, 0.0035, 0.0104, 0.0322, 0.0054, 0.1972, 0.0084, 0.0109], device='cuda:0'), in_proj_covar=tensor([0.0111, 0.0098, 0.0146, 0.0142, 0.0115, 0.0162, 0.0131, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:32:57,691 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2794, 4.3889, 4.5682, 4.4544, 4.4041, 4.9696, 4.4980, 4.2044], device='cuda:0'), covar=tensor([0.1329, 0.1678, 0.1400, 0.1525, 0.2366, 0.0870, 0.1304, 0.2301], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0435, 0.0457, 0.0385, 0.0514, 0.0481, 0.0371, 0.0522], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 21:32:59,670 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-28 21:33:09,824 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5297, 3.5132, 2.7572, 2.1363, 2.5065, 2.0981, 3.6685, 3.4195], device='cuda:0'), covar=tensor([0.2537, 0.0669, 0.1456, 0.2025, 0.2123, 0.1798, 0.0450, 0.0913], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0252, 0.0275, 0.0263, 0.0280, 0.0211, 0.0258, 0.0279], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:33:59,284 INFO [train.py:904] (0/8) Epoch 8, batch 6200, loss[loss=0.2589, simple_loss=0.3155, pruned_loss=0.1012, over 11732.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3111, pruned_loss=0.07599, over 3110527.58 frames. ], batch size: 246, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:34:31,216 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:32,466 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77273.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:39,069 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77277.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:34:51,420 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 21:35:16,105 INFO [train.py:904] (0/8) Epoch 8, batch 6250, loss[loss=0.2185, simple_loss=0.311, pruned_loss=0.06295, over 16500.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3103, pruned_loss=0.07566, over 3094966.14 frames. ], batch size: 75, lr: 8.50e-03, grad_scale: 4.0 2023-04-28 21:35:22,798 INFO [optim.py:368] (0/8) 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 21:35:51,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7085, 4.1854, 4.0469, 2.0893, 3.2602, 2.7712, 4.0924, 4.2311], device='cuda:0'), covar=tensor([0.0208, 0.0487, 0.0462, 0.1634, 0.0687, 0.0818, 0.0460, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0135, 0.0156, 0.0141, 0.0136, 0.0125, 0.0136, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:36:07,337 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77334.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:11,759 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8873, 1.9233, 2.2821, 3.1739, 2.0558, 2.2342, 2.1370, 1.9796], device='cuda:0'), covar=tensor([0.0847, 0.2654, 0.1513, 0.0453, 0.3261, 0.1946, 0.2534, 0.2711], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0364, 0.0303, 0.0318, 0.0397, 0.0405, 0.0325, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:36:12,949 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77338.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:36:35,485 INFO [train.py:904] (0/8) Epoch 8, batch 6300, loss[loss=0.2234, simple_loss=0.3052, pruned_loss=0.07086, over 16933.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.31, pruned_loss=0.07437, over 3118865.35 frames. ], batch size: 109, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:37:54,086 INFO [train.py:904] (0/8) Epoch 8, batch 6350, loss[loss=0.2068, simple_loss=0.2954, pruned_loss=0.05908, over 16858.00 frames. ], tot_loss[loss=0.2321, simple_loss=0.3117, pruned_loss=0.07624, over 3103657.10 frames. ], batch size: 90, lr: 8.49e-03, grad_scale: 4.0 2023-04-28 21:38:00,462 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.093e+02 3.385e+02 4.021e+02 5.215e+02 8.857e+02, threshold=8.042e+02, percent-clipped=2.0 2023-04-28 21:38:01,057 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7757, 4.7330, 4.5622, 3.9659, 4.6125, 1.6770, 4.4039, 4.4347], device='cuda:0'), covar=tensor([0.0060, 0.0056, 0.0099, 0.0281, 0.0063, 0.2015, 0.0092, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0097, 0.0144, 0.0140, 0.0114, 0.0160, 0.0130, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:39:09,935 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77451.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 21:39:11,224 INFO [train.py:904] (0/8) Epoch 8, batch 6400, loss[loss=0.2261, simple_loss=0.3053, pruned_loss=0.07349, over 16700.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3121, pruned_loss=0.07759, over 3109378.15 frames. ], batch size: 124, lr: 8.49e-03, grad_scale: 8.0 2023-04-28 21:39:33,256 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-28 21:39:52,276 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5836, 2.5711, 2.2082, 3.5942, 2.6792, 3.8175, 1.3010, 2.7081], device='cuda:0'), covar=tensor([0.1344, 0.0613, 0.1188, 0.0112, 0.0202, 0.0387, 0.1603, 0.0763], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0152, 0.0176, 0.0119, 0.0203, 0.0206, 0.0175, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:40:17,157 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77495.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:40:26,243 INFO [train.py:904] (0/8) Epoch 8, batch 6450, loss[loss=0.2101, simple_loss=0.289, pruned_loss=0.0656, over 16820.00 frames. ], tot_loss[loss=0.2337, simple_loss=0.3124, pruned_loss=0.07746, over 3089243.60 frames. ], batch size: 116, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:40:33,082 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.067e+02 3.422e+02 4.688e+02 6.041e+02 9.477e+02, threshold=9.377e+02, percent-clipped=7.0 2023-04-28 21:41:31,680 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77543.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:41:45,633 INFO [train.py:904] (0/8) Epoch 8, batch 6500, loss[loss=0.1957, simple_loss=0.2856, pruned_loss=0.05297, over 16770.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3095, pruned_loss=0.07572, over 3111345.69 frames. ], batch size: 89, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:42:09,125 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77567.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:42:27,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5608, 2.6529, 1.6713, 2.7940, 2.1468, 2.7822, 2.0151, 2.3426], device='cuda:0'), covar=tensor([0.0219, 0.0346, 0.1296, 0.0142, 0.0629, 0.0503, 0.1133, 0.0543], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0158, 0.0182, 0.0102, 0.0163, 0.0197, 0.0191, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:42:32,212 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3115, 2.9185, 2.5820, 2.2183, 2.2303, 2.1147, 2.9027, 2.9133], device='cuda:0'), covar=tensor([0.1892, 0.0700, 0.1152, 0.1590, 0.1600, 0.1561, 0.0424, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0252, 0.0273, 0.0264, 0.0280, 0.0211, 0.0258, 0.0279], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:43:05,187 INFO [train.py:904] (0/8) Epoch 8, batch 6550, loss[loss=0.2388, simple_loss=0.3286, pruned_loss=0.07454, over 15411.00 frames. ], tot_loss[loss=0.2329, simple_loss=0.3125, pruned_loss=0.07669, over 3098835.13 frames. ], batch size: 192, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:43:11,241 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.922e+02 3.133e+02 3.759e+02 4.463e+02 1.371e+03, threshold=7.519e+02, percent-clipped=1.0 2023-04-28 21:43:27,649 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77616.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:47,313 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77629.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:43:53,415 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 8, batch 6600, loss[loss=0.2263, simple_loss=0.3038, pruned_loss=0.07438, over 17018.00 frames. ], tot_loss[loss=0.2342, simple_loss=0.3139, pruned_loss=0.07721, over 3083607.40 frames. ], batch size: 55, lr: 8.48e-03, grad_scale: 8.0 2023-04-28 21:45:00,797 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:00,899 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77677.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:45:38,855 INFO [train.py:904] (0/8) Epoch 8, batch 6650, loss[loss=0.2356, simple_loss=0.3043, pruned_loss=0.08342, over 16800.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3142, pruned_loss=0.07804, over 3098098.41 frames. ], batch size: 124, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:45:39,853 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 21:45:43,376 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-28 21:45:45,539 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.195e+02 3.528e+02 4.149e+02 5.029e+02 1.289e+03, threshold=8.299e+02, percent-clipped=5.0 2023-04-28 21:46:35,366 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=77738.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:46:53,216 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77751.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 21:46:54,556 INFO [train.py:904] (0/8) Epoch 8, batch 6700, loss[loss=0.2128, simple_loss=0.2929, pruned_loss=0.0663, over 16983.00 frames. ], tot_loss[loss=0.2345, simple_loss=0.3129, pruned_loss=0.07808, over 3101304.65 frames. ], batch size: 55, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:06,141 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77799.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:48:10,955 INFO [train.py:904] (0/8) Epoch 8, batch 6750, loss[loss=0.2323, simple_loss=0.3041, pruned_loss=0.0802, over 16693.00 frames. ], tot_loss[loss=0.2348, simple_loss=0.3125, pruned_loss=0.07858, over 3090660.49 frames. ], batch size: 124, lr: 8.47e-03, grad_scale: 4.0 2023-04-28 21:48:18,415 INFO [optim.py:368] (0/8) 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 21:48:24,725 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6367, 3.8824, 2.9411, 2.3178, 2.9220, 2.2763, 4.1164, 3.6871], device='cuda:0'), covar=tensor([0.2424, 0.0714, 0.1353, 0.1819, 0.1852, 0.1546, 0.0404, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0254, 0.0274, 0.0264, 0.0280, 0.0211, 0.0259, 0.0278], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:48:43,107 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4949, 3.5906, 1.6067, 3.8895, 2.4311, 3.8433, 1.8461, 2.7016], device='cuda:0'), covar=tensor([0.0177, 0.0284, 0.1726, 0.0081, 0.0789, 0.0395, 0.1545, 0.0595], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0159, 0.0181, 0.0102, 0.0164, 0.0197, 0.0192, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:49:14,519 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5655, 4.7606, 4.9155, 4.8239, 4.7517, 5.3645, 4.8746, 4.6258], device='cuda:0'), covar=tensor([0.1055, 0.1560, 0.1539, 0.1558, 0.2510, 0.0816, 0.1253, 0.2195], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0440, 0.0469, 0.0397, 0.0522, 0.0489, 0.0379, 0.0530], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 21:49:25,321 INFO [train.py:904] (0/8) Epoch 8, batch 6800, loss[loss=0.2925, simple_loss=0.3457, pruned_loss=0.1196, over 11189.00 frames. ], tot_loss[loss=0.2343, simple_loss=0.3124, pruned_loss=0.07815, over 3096527.60 frames. ], batch size: 246, lr: 8.47e-03, grad_scale: 8.0 2023-04-28 21:49:49,087 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77867.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:50:44,170 INFO [train.py:904] (0/8) Epoch 8, batch 6850, loss[loss=0.2122, simple_loss=0.3217, pruned_loss=0.05133, over 16820.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3127, pruned_loss=0.07717, over 3118047.93 frames. ], batch size: 102, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:50:53,215 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77915.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:24,211 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77929.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:29,979 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=77933.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:51:59,519 INFO [train.py:904] (0/8) Epoch 8, batch 6900, loss[loss=0.2288, simple_loss=0.3158, pruned_loss=0.07091, over 16421.00 frames. ], tot_loss[loss=0.2336, simple_loss=0.3147, pruned_loss=0.07629, over 3125905.51 frames. ], batch size: 75, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:52:31,121 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=77972.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:39,051 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=77981.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:52:57,682 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8042, 3.0029, 2.8718, 4.9266, 3.9661, 4.4967, 1.6037, 3.1959], device='cuda:0'), covar=tensor([0.1334, 0.0620, 0.1070, 0.0121, 0.0371, 0.0343, 0.1539, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0151, 0.0173, 0.0118, 0.0199, 0.0203, 0.0173, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 21:53:15,578 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-78000.pt 2023-04-28 21:53:20,815 INFO [train.py:904] (0/8) Epoch 8, batch 6950, loss[loss=0.2249, simple_loss=0.3067, pruned_loss=0.07152, over 16667.00 frames. ], tot_loss[loss=0.2368, simple_loss=0.3167, pruned_loss=0.07843, over 3124094.38 frames. ], batch size: 62, lr: 8.46e-03, grad_scale: 4.0 2023-04-28 21:53:29,770 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.004e+02 3.295e+02 4.352e+02 5.795e+02 9.816e+02, threshold=8.703e+02, percent-clipped=9.0 2023-04-28 21:53:38,926 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0151, 3.0407, 1.7242, 3.2922, 2.2555, 3.2642, 1.8920, 2.4593], device='cuda:0'), covar=tensor([0.0234, 0.0367, 0.1454, 0.0113, 0.0832, 0.0435, 0.1426, 0.0697], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0159, 0.0181, 0.0101, 0.0164, 0.0197, 0.0192, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 21:53:40,444 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 21:54:06,096 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-28 21:54:10,667 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 8, batch 7000, loss[loss=0.2275, simple_loss=0.2939, pruned_loss=0.08052, over 11817.00 frames. ], tot_loss[loss=0.2357, simple_loss=0.3164, pruned_loss=0.07749, over 3124000.08 frames. ], batch size: 247, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:55:29,098 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3025, 4.2858, 4.1846, 3.5604, 4.2091, 1.5793, 3.9566, 3.8672], device='cuda:0'), covar=tensor([0.0070, 0.0055, 0.0112, 0.0286, 0.0062, 0.2081, 0.0098, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0097, 0.0144, 0.0140, 0.0114, 0.0161, 0.0129, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:55:53,788 INFO [train.py:904] (0/8) Epoch 8, batch 7050, loss[loss=0.2144, simple_loss=0.3025, pruned_loss=0.06318, over 16689.00 frames. ], tot_loss[loss=0.2356, simple_loss=0.317, pruned_loss=0.07714, over 3122401.48 frames. ], batch size: 57, lr: 8.45e-03, grad_scale: 4.0 2023-04-28 21:56:03,902 INFO [optim.py:368] (0/8) 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 21:56:42,412 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1232, 4.2281, 3.3062, 2.5957, 3.2023, 2.6193, 4.5948, 4.0566], device='cuda:0'), covar=tensor([0.2093, 0.0704, 0.1356, 0.1713, 0.2114, 0.1481, 0.0375, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0256, 0.0276, 0.0264, 0.0281, 0.0212, 0.0261, 0.0278], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:57:11,219 INFO [train.py:904] (0/8) Epoch 8, batch 7100, loss[loss=0.2754, simple_loss=0.3291, pruned_loss=0.1109, over 11431.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3164, pruned_loss=0.07791, over 3104490.93 frames. ], batch size: 246, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:26,607 INFO [train.py:904] (0/8) Epoch 8, batch 7150, loss[loss=0.2078, simple_loss=0.2934, pruned_loss=0.06112, over 16813.00 frames. ], tot_loss[loss=0.234, simple_loss=0.314, pruned_loss=0.07697, over 3113275.58 frames. ], batch size: 83, lr: 8.45e-03, grad_scale: 2.0 2023-04-28 21:58:36,177 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.673e+02 3.732e+02 4.325e+02 5.377e+02 1.184e+03, threshold=8.651e+02, percent-clipped=7.0 2023-04-28 21:59:09,623 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78232.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 21:59:27,417 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7785, 1.7199, 1.5136, 1.4615, 1.8412, 1.5913, 1.7471, 1.9021], device='cuda:0'), covar=tensor([0.0065, 0.0166, 0.0236, 0.0233, 0.0107, 0.0175, 0.0097, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0110, 0.0180, 0.0178, 0.0181, 0.0178, 0.0180, 0.0175, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 21:59:39,856 INFO [train.py:904] (0/8) Epoch 8, batch 7200, loss[loss=0.2008, simple_loss=0.2889, pruned_loss=0.0564, over 15259.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3122, pruned_loss=0.07567, over 3109678.93 frames. ], batch size: 190, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:00:10,003 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78272.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:00:18,684 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6920, 4.6713, 4.5250, 3.9441, 4.5710, 1.7074, 4.3919, 4.3967], device='cuda:0'), covar=tensor([0.0059, 0.0055, 0.0103, 0.0280, 0.0061, 0.2066, 0.0082, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0095, 0.0141, 0.0138, 0.0113, 0.0158, 0.0127, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:00:20,379 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9868, 4.0191, 4.4180, 4.3943, 4.3714, 4.0703, 4.1071, 3.9420], device='cuda:0'), covar=tensor([0.0251, 0.0428, 0.0303, 0.0342, 0.0343, 0.0317, 0.0743, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0294, 0.0295, 0.0285, 0.0335, 0.0312, 0.0414, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 22:00:44,969 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78293.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:00:59,560 INFO [train.py:904] (0/8) Epoch 8, batch 7250, loss[loss=0.201, simple_loss=0.2805, pruned_loss=0.06075, over 16621.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3099, pruned_loss=0.07409, over 3115697.00 frames. ], batch size: 57, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:01:10,050 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.993e+02 2.959e+02 3.722e+02 4.557e+02 8.668e+02, threshold=7.444e+02, percent-clipped=1.0 2023-04-28 22:01:26,988 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78320.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:01:47,568 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:01:53,452 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9425, 3.5861, 3.2352, 1.7267, 2.7583, 2.4844, 3.3525, 3.6837], device='cuda:0'), covar=tensor([0.0301, 0.0545, 0.0628, 0.1950, 0.0865, 0.0901, 0.0718, 0.0757], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0132, 0.0155, 0.0140, 0.0133, 0.0124, 0.0135, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 22:02:16,551 INFO [train.py:904] (0/8) Epoch 8, batch 7300, loss[loss=0.2348, simple_loss=0.3156, pruned_loss=0.07697, over 16505.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3089, pruned_loss=0.0737, over 3107064.34 frames. ], batch size: 68, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:03:02,230 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78381.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:03:34,475 INFO [train.py:904] (0/8) Epoch 8, batch 7350, loss[loss=0.1886, simple_loss=0.2764, pruned_loss=0.05033, over 17098.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3093, pruned_loss=0.07458, over 3086183.67 frames. ], batch size: 47, lr: 8.44e-03, grad_scale: 4.0 2023-04-28 22:03:41,037 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.66 vs. limit=5.0 2023-04-28 22:03:45,282 INFO [optim.py:368] (0/8) 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 22:03:58,851 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9650, 2.5636, 2.7483, 4.8072, 2.2223, 2.9263, 2.5767, 2.8391], device='cuda:0'), covar=tensor([0.0706, 0.2560, 0.1618, 0.0274, 0.3366, 0.1665, 0.2315, 0.2638], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0368, 0.0306, 0.0319, 0.0401, 0.0404, 0.0326, 0.0433], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:04:55,420 INFO [train.py:904] (0/8) Epoch 8, batch 7400, loss[loss=0.2491, simple_loss=0.3305, pruned_loss=0.08384, over 16741.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3099, pruned_loss=0.07518, over 3077754.68 frames. ], batch size: 134, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:05:07,828 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78460.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:05:30,166 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 22:05:32,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4902, 4.5413, 5.0225, 4.9870, 4.9168, 4.5536, 4.5843, 4.2838], device='cuda:0'), covar=tensor([0.0261, 0.0405, 0.0256, 0.0329, 0.0433, 0.0298, 0.0794, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0288, 0.0294, 0.0280, 0.0331, 0.0309, 0.0407, 0.0254], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 22:06:13,357 INFO [train.py:904] (0/8) Epoch 8, batch 7450, loss[loss=0.2184, simple_loss=0.3133, pruned_loss=0.0618, over 16867.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3116, pruned_loss=0.07657, over 3081812.40 frames. ], batch size: 116, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:06:26,557 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.920e+02 3.338e+02 4.187e+02 5.103e+02 1.080e+03, threshold=8.375e+02, percent-clipped=2.0 2023-04-28 22:06:43,919 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7554, 3.1081, 2.6737, 4.6267, 3.5615, 4.3563, 1.4041, 3.2339], device='cuda:0'), covar=tensor([0.1346, 0.0567, 0.1017, 0.0124, 0.0193, 0.0343, 0.1577, 0.0660], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0154, 0.0174, 0.0119, 0.0202, 0.0205, 0.0175, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 22:06:46,805 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78521.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:07:34,693 INFO [train.py:904] (0/8) Epoch 8, batch 7500, loss[loss=0.2803, simple_loss=0.339, pruned_loss=0.1108, over 11663.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3129, pruned_loss=0.07704, over 3052962.52 frames. ], batch size: 248, lr: 8.43e-03, grad_scale: 4.0 2023-04-28 22:07:42,130 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78556.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:08:33,289 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78588.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:08:55,088 INFO [train.py:904] (0/8) Epoch 8, batch 7550, loss[loss=0.2225, simple_loss=0.3061, pruned_loss=0.06941, over 16394.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3108, pruned_loss=0.07598, over 3086661.45 frames. ], batch size: 146, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:09:05,666 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.406e+02 3.301e+02 4.162e+02 5.431e+02 1.258e+03, threshold=8.325e+02, percent-clipped=7.0 2023-04-28 22:09:19,458 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=78617.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:09:21,631 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-28 22:10:11,725 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0932, 3.1748, 3.3527, 1.5405, 3.5372, 3.5674, 2.6563, 2.5020], device='cuda:0'), covar=tensor([0.0741, 0.0177, 0.0139, 0.1179, 0.0057, 0.0093, 0.0375, 0.0480], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0094, 0.0080, 0.0134, 0.0065, 0.0086, 0.0115, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 22:10:12,508 INFO [train.py:904] (0/8) Epoch 8, batch 7600, loss[loss=0.2535, simple_loss=0.3365, pruned_loss=0.08526, over 16289.00 frames. ], tot_loss[loss=0.2311, simple_loss=0.31, pruned_loss=0.07609, over 3083224.22 frames. ], batch size: 165, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:10:45,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9438, 4.9061, 5.4713, 5.3964, 5.3732, 5.0192, 4.9837, 4.6268], device='cuda:0'), covar=tensor([0.0244, 0.0388, 0.0240, 0.0312, 0.0403, 0.0294, 0.0820, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0296, 0.0299, 0.0289, 0.0341, 0.0317, 0.0418, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-28 22:11:30,206 INFO [train.py:904] (0/8) Epoch 8, batch 7650, loss[loss=0.2172, simple_loss=0.3018, pruned_loss=0.06631, over 17177.00 frames. ], tot_loss[loss=0.2318, simple_loss=0.3106, pruned_loss=0.07654, over 3082185.25 frames. ], batch size: 46, lr: 8.42e-03, grad_scale: 8.0 2023-04-28 22:11:40,450 INFO [optim.py:368] (0/8) 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 22:12:21,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9784, 2.3683, 1.8361, 1.9723, 2.7779, 2.3535, 2.9997, 3.0204], device='cuda:0'), covar=tensor([0.0069, 0.0236, 0.0343, 0.0316, 0.0142, 0.0236, 0.0128, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0180, 0.0179, 0.0181, 0.0178, 0.0180, 0.0175, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:12:45,901 INFO [train.py:904] (0/8) Epoch 8, batch 7700, loss[loss=0.2295, simple_loss=0.3106, pruned_loss=0.07421, over 16704.00 frames. ], tot_loss[loss=0.2324, simple_loss=0.3109, pruned_loss=0.07695, over 3092949.87 frames. ], batch size: 62, lr: 8.42e-03, grad_scale: 4.0 2023-04-28 22:13:33,970 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9107, 3.8765, 3.8480, 2.7756, 3.8468, 1.5965, 3.6075, 3.4453], device='cuda:0'), covar=tensor([0.0184, 0.0127, 0.0189, 0.0616, 0.0127, 0.2987, 0.0182, 0.0358], device='cuda:0'), in_proj_covar=tensor([0.0109, 0.0096, 0.0142, 0.0139, 0.0113, 0.0161, 0.0127, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:14:03,993 INFO [train.py:904] (0/8) Epoch 8, batch 7750, loss[loss=0.2303, simple_loss=0.3152, pruned_loss=0.07267, over 16268.00 frames. ], tot_loss[loss=0.2323, simple_loss=0.311, pruned_loss=0.07675, over 3104976.64 frames. ], batch size: 165, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:14:17,810 INFO [optim.py:368] (0/8) 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 22:14:26,091 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78816.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:15:01,318 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-28 22:15:04,084 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3463, 3.1056, 2.9675, 1.8881, 2.6807, 2.1886, 2.9447, 3.2345], device='cuda:0'), covar=tensor([0.0304, 0.0569, 0.0586, 0.1703, 0.0759, 0.0998, 0.0646, 0.0621], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0135, 0.0157, 0.0142, 0.0134, 0.0125, 0.0137, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 22:15:19,816 INFO [train.py:904] (0/8) Epoch 8, batch 7800, loss[loss=0.2089, simple_loss=0.2984, pruned_loss=0.0597, over 16682.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3117, pruned_loss=0.07723, over 3109367.13 frames. ], batch size: 89, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:16,586 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=78888.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:16:38,229 INFO [train.py:904] (0/8) Epoch 8, batch 7850, loss[loss=0.2361, simple_loss=0.3167, pruned_loss=0.07776, over 16521.00 frames. ], tot_loss[loss=0.2349, simple_loss=0.3135, pruned_loss=0.07813, over 3092252.56 frames. ], batch size: 146, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:16:50,962 INFO [optim.py:368] (0/8) 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 22:16:53,431 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=78912.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:20,949 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0148, 2.3738, 2.2671, 2.9369, 2.2982, 3.2307, 1.6644, 2.6987], device='cuda:0'), covar=tensor([0.1137, 0.0541, 0.0981, 0.0145, 0.0172, 0.0409, 0.1338, 0.0718], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0153, 0.0173, 0.0120, 0.0202, 0.0203, 0.0174, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 22:17:29,118 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=78936.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:17:29,288 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9651, 3.0974, 1.6162, 3.2614, 2.2278, 3.2949, 1.9111, 2.5914], device='cuda:0'), covar=tensor([0.0241, 0.0365, 0.1567, 0.0125, 0.0821, 0.0500, 0.1454, 0.0629], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0158, 0.0182, 0.0103, 0.0166, 0.0197, 0.0193, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 22:17:41,679 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2681, 4.0887, 4.3141, 4.4765, 4.5683, 4.1606, 4.5080, 4.5363], device='cuda:0'), covar=tensor([0.1243, 0.0960, 0.1234, 0.0554, 0.0530, 0.0961, 0.0674, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0570, 0.0701, 0.0581, 0.0445, 0.0429, 0.0465, 0.0507], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:17:54,236 INFO [train.py:904] (0/8) Epoch 8, batch 7900, loss[loss=0.2496, simple_loss=0.3287, pruned_loss=0.0853, over 15343.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3126, pruned_loss=0.07784, over 3081637.93 frames. ], batch size: 190, lr: 8.41e-03, grad_scale: 2.0 2023-04-28 22:18:10,289 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-28 22:18:15,517 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=78965.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:19:13,458 INFO [train.py:904] (0/8) Epoch 8, batch 7950, loss[loss=0.2105, simple_loss=0.2926, pruned_loss=0.06421, over 16857.00 frames. ], tot_loss[loss=0.2334, simple_loss=0.3123, pruned_loss=0.07726, over 3107257.51 frames. ], batch size: 96, lr: 8.40e-03, grad_scale: 2.0 2023-04-28 22:19:25,529 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9025, 3.1520, 3.1557, 2.0979, 2.9057, 3.1780, 3.0231, 1.8186], device='cuda:0'), covar=tensor([0.0357, 0.0036, 0.0043, 0.0286, 0.0067, 0.0070, 0.0049, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0060, 0.0063, 0.0119, 0.0067, 0.0079, 0.0069, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 22:19:26,655 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8552, 4.1564, 3.9324, 4.0335, 3.6825, 3.7701, 3.7876, 4.1175], device='cuda:0'), covar=tensor([0.1030, 0.0829, 0.1070, 0.0579, 0.0769, 0.1435, 0.0830, 0.0925], device='cuda:0'), in_proj_covar=tensor([0.0465, 0.0577, 0.0496, 0.0398, 0.0363, 0.0389, 0.0487, 0.0431], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:19:28,067 INFO [optim.py:368] (0/8) 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 22:19:49,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9727, 3.1275, 3.1395, 2.0962, 2.8848, 3.1593, 2.9989, 1.8257], device='cuda:0'), covar=tensor([0.0382, 0.0038, 0.0042, 0.0304, 0.0063, 0.0074, 0.0049, 0.0370], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0061, 0.0064, 0.0120, 0.0068, 0.0080, 0.0069, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 22:19:51,953 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79026.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:20:32,285 INFO [train.py:904] (0/8) Epoch 8, batch 8000, loss[loss=0.2332, simple_loss=0.3132, pruned_loss=0.07662, over 15418.00 frames. ], tot_loss[loss=0.2354, simple_loss=0.3137, pruned_loss=0.07861, over 3080062.38 frames. ], batch size: 191, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:21:40,434 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4545, 1.3925, 1.9394, 2.2942, 2.3590, 2.6441, 1.5522, 2.5576], device='cuda:0'), covar=tensor([0.0124, 0.0362, 0.0198, 0.0209, 0.0181, 0.0110, 0.0339, 0.0069], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0160, 0.0143, 0.0142, 0.0150, 0.0108, 0.0158, 0.0099], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-28 22:21:48,624 INFO [train.py:904] (0/8) Epoch 8, batch 8050, loss[loss=0.2279, simple_loss=0.324, pruned_loss=0.06587, over 16517.00 frames. ], tot_loss[loss=0.2358, simple_loss=0.3139, pruned_loss=0.07878, over 3067272.98 frames. ], batch size: 68, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:22:02,035 INFO [optim.py:368] (0/8) 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 22:22:11,151 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:23:05,323 INFO [train.py:904] (0/8) Epoch 8, batch 8100, loss[loss=0.1973, simple_loss=0.2891, pruned_loss=0.05276, over 16694.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.313, pruned_loss=0.07784, over 3070471.96 frames. ], batch size: 89, lr: 8.40e-03, grad_scale: 4.0 2023-04-28 22:23:23,754 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79164.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:23:46,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8735, 1.9742, 2.2589, 3.1165, 2.1347, 2.2904, 2.1760, 2.0721], device='cuda:0'), covar=tensor([0.0894, 0.2602, 0.1635, 0.0541, 0.3164, 0.1755, 0.2446, 0.2798], device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0371, 0.0308, 0.0319, 0.0407, 0.0409, 0.0328, 0.0436], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:24:22,974 INFO [train.py:904] (0/8) Epoch 8, batch 8150, loss[loss=0.2479, simple_loss=0.3168, pruned_loss=0.08949, over 11576.00 frames. ], tot_loss[loss=0.2319, simple_loss=0.3104, pruned_loss=0.07672, over 3078239.66 frames. ], batch size: 246, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:24:36,888 INFO [optim.py:368] (0/8) 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 22:24:38,560 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79212.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:25:01,506 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0349, 3.4716, 3.5059, 2.4266, 3.2715, 3.4725, 3.3899, 1.9514], device='cuda:0'), covar=tensor([0.0396, 0.0029, 0.0036, 0.0269, 0.0065, 0.0078, 0.0043, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0061, 0.0064, 0.0120, 0.0069, 0.0080, 0.0069, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 22:25:23,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3953, 3.4896, 1.9568, 3.7458, 2.4812, 3.7280, 1.9870, 2.7072], device='cuda:0'), covar=tensor([0.0176, 0.0308, 0.1565, 0.0093, 0.0778, 0.0518, 0.1443, 0.0657], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0157, 0.0182, 0.0104, 0.0166, 0.0199, 0.0192, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 22:25:42,520 INFO [train.py:904] (0/8) Epoch 8, batch 8200, loss[loss=0.2489, simple_loss=0.3106, pruned_loss=0.09366, over 11266.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3069, pruned_loss=0.0755, over 3086358.17 frames. ], batch size: 247, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:25:55,511 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79260.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:26:28,026 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8460, 2.3152, 2.3189, 2.9742, 1.9120, 3.2991, 1.5717, 2.7051], device='cuda:0'), covar=tensor([0.1316, 0.0572, 0.0936, 0.0157, 0.0131, 0.0396, 0.1485, 0.0694], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0151, 0.0170, 0.0118, 0.0199, 0.0199, 0.0171, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 22:26:36,237 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79285.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:26:45,644 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-28 22:27:04,327 INFO [train.py:904] (0/8) Epoch 8, batch 8250, loss[loss=0.2092, simple_loss=0.2909, pruned_loss=0.06375, over 12118.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3062, pruned_loss=0.07317, over 3075874.15 frames. ], batch size: 250, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:27:19,447 INFO [optim.py:368] (0/8) 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 22:27:37,071 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79321.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:27:57,011 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79333.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:17,618 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79346.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:28:26,465 INFO [train.py:904] (0/8) Epoch 8, batch 8300, loss[loss=0.1989, simple_loss=0.2881, pruned_loss=0.05486, over 15358.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3029, pruned_loss=0.06977, over 3058640.77 frames. ], batch size: 190, lr: 8.39e-03, grad_scale: 4.0 2023-04-28 22:28:39,564 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-28 22:28:46,849 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79364.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:28:57,530 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8959, 2.1557, 1.8188, 1.9215, 2.5383, 2.2658, 2.7970, 2.7903], device='cuda:0'), covar=tensor([0.0060, 0.0278, 0.0339, 0.0319, 0.0156, 0.0233, 0.0127, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0179, 0.0178, 0.0177, 0.0175, 0.0179, 0.0173, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:29:04,947 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79375.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:29:36,182 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79394.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:29:48,256 INFO [train.py:904] (0/8) Epoch 8, batch 8350, loss[loss=0.2409, simple_loss=0.3041, pruned_loss=0.08878, over 11623.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3023, pruned_loss=0.06795, over 3046772.14 frames. ], batch size: 246, lr: 8.38e-03, grad_scale: 4.0 2023-04-28 22:30:02,874 INFO [optim.py:368] (0/8) 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 22:30:26,203 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79425.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:30:43,715 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79436.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:31:09,033 INFO [train.py:904] (0/8) Epoch 8, batch 8400, loss[loss=0.1783, simple_loss=0.2772, pruned_loss=0.03974, over 16686.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2988, pruned_loss=0.06456, over 3064230.50 frames. ], batch size: 89, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:27,067 INFO [train.py:904] (0/8) Epoch 8, batch 8450, loss[loss=0.2182, simple_loss=0.294, pruned_loss=0.07116, over 12355.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2966, pruned_loss=0.06295, over 3047333.25 frames. ], batch size: 247, lr: 8.38e-03, grad_scale: 8.0 2023-04-28 22:32:42,131 INFO [optim.py:368] (0/8) 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 22:33:02,004 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-28 22:33:47,278 INFO [train.py:904] (0/8) Epoch 8, batch 8500, loss[loss=0.1615, simple_loss=0.242, pruned_loss=0.04053, over 12110.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2918, pruned_loss=0.05993, over 3035172.93 frames. ], batch size: 246, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:34:40,424 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79585.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:35:09,525 INFO [train.py:904] (0/8) Epoch 8, batch 8550, loss[loss=0.1838, simple_loss=0.2613, pruned_loss=0.05321, over 12100.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2891, pruned_loss=0.05821, over 3031050.12 frames. ], batch size: 248, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:35:26,484 INFO [optim.py:368] (0/8) 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 22:35:47,397 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79621.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:28,328 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79641.0, num_to_drop=1, layers_to_drop={3} 2023-04-28 22:36:30,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1596, 2.8631, 2.8441, 1.9998, 2.5991, 2.1391, 2.6989, 2.9749], device='cuda:0'), covar=tensor([0.0330, 0.0714, 0.0448, 0.1529, 0.0698, 0.0940, 0.0615, 0.0699], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0127, 0.0148, 0.0136, 0.0128, 0.0121, 0.0131, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 22:36:36,798 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79646.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:36:48,478 INFO [train.py:904] (0/8) Epoch 8, batch 8600, loss[loss=0.1977, simple_loss=0.2697, pruned_loss=0.06281, over 12137.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2892, pruned_loss=0.0573, over 3013011.59 frames. ], batch size: 246, lr: 8.37e-03, grad_scale: 8.0 2023-04-28 22:36:59,395 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:37:24,929 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79669.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:38:02,636 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79689.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:38:11,332 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-28 22:38:26,210 INFO [train.py:904] (0/8) Epoch 8, batch 8650, loss[loss=0.1767, simple_loss=0.2796, pruned_loss=0.03683, over 16604.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2878, pruned_loss=0.05606, over 3029505.29 frames. ], batch size: 89, lr: 8.37e-03, grad_scale: 4.0 2023-04-28 22:38:42,140 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0570, 3.8542, 3.8743, 4.2024, 4.3685, 3.9704, 4.2935, 4.3323], device='cuda:0'), covar=tensor([0.1326, 0.1103, 0.2036, 0.1007, 0.0807, 0.1286, 0.0997, 0.0892], device='cuda:0'), in_proj_covar=tensor([0.0451, 0.0551, 0.0673, 0.0565, 0.0428, 0.0421, 0.0449, 0.0492], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:38:50,267 INFO [optim.py:368] (0/8) 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 22:39:04,763 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79718.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:09,075 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79720.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:39:22,971 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6104, 3.9220, 2.0818, 4.2135, 2.7384, 4.0956, 2.2857, 3.0377], device='cuda:0'), covar=tensor([0.0181, 0.0226, 0.1434, 0.0079, 0.0757, 0.0374, 0.1380, 0.0580], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0151, 0.0176, 0.0099, 0.0160, 0.0188, 0.0188, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-28 22:39:32,718 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79731.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 22:39:48,762 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3163, 4.4870, 4.5814, 4.4860, 4.4538, 4.9705, 4.5673, 4.3352], device='cuda:0'), covar=tensor([0.1171, 0.1588, 0.1491, 0.1592, 0.2535, 0.0958, 0.1248, 0.2070], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0412, 0.0442, 0.0368, 0.0486, 0.0465, 0.0362, 0.0490], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:40:00,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5414, 3.4560, 2.8122, 2.2333, 2.3023, 2.2483, 3.7222, 3.2476], device='cuda:0'), covar=tensor([0.2479, 0.0701, 0.1477, 0.2078, 0.2217, 0.1723, 0.0413, 0.0959], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0244, 0.0268, 0.0255, 0.0259, 0.0206, 0.0247, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:40:12,071 INFO [train.py:904] (0/8) Epoch 8, batch 8700, loss[loss=0.1703, simple_loss=0.264, pruned_loss=0.03824, over 16747.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2849, pruned_loss=0.05446, over 3048371.52 frames. ], batch size: 83, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:40:32,999 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=79762.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:40:55,651 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2268, 2.2819, 1.8505, 2.0130, 2.6146, 2.3548, 3.0692, 3.0332], device='cuda:0'), covar=tensor([0.0066, 0.0291, 0.0360, 0.0311, 0.0186, 0.0262, 0.0129, 0.0134], device='cuda:0'), in_proj_covar=tensor([0.0107, 0.0179, 0.0175, 0.0176, 0.0174, 0.0177, 0.0167, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:41:28,442 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0396, 2.6779, 2.6848, 1.9238, 2.8720, 2.8990, 2.5088, 2.4151], device='cuda:0'), covar=tensor([0.0651, 0.0175, 0.0170, 0.0898, 0.0065, 0.0118, 0.0376, 0.0416], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0092, 0.0079, 0.0135, 0.0063, 0.0084, 0.0114, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 22:41:50,478 INFO [train.py:904] (0/8) Epoch 8, batch 8750, loss[loss=0.1734, simple_loss=0.2578, pruned_loss=0.04452, over 12007.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.284, pruned_loss=0.05358, over 3046808.42 frames. ], batch size: 250, lr: 8.36e-03, grad_scale: 4.0 2023-04-28 22:42:15,365 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.669e+02 3.166e+02 4.113e+02 7.290e+02, threshold=6.332e+02, percent-clipped=2.0 2023-04-28 22:42:44,004 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=79823.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:43:44,539 INFO [train.py:904] (0/8) Epoch 8, batch 8800, loss[loss=0.1849, simple_loss=0.2765, pruned_loss=0.04659, over 15471.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2824, pruned_loss=0.05218, over 3048129.66 frames. ], batch size: 191, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:44:26,930 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6394, 2.7021, 1.7293, 2.8855, 2.0636, 2.8189, 2.0147, 2.3869], device='cuda:0'), covar=tensor([0.0255, 0.0360, 0.1488, 0.0163, 0.0910, 0.0575, 0.1396, 0.0745], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0151, 0.0179, 0.0100, 0.0162, 0.0189, 0.0189, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-28 22:45:31,368 INFO [train.py:904] (0/8) Epoch 8, batch 8850, loss[loss=0.1884, simple_loss=0.294, pruned_loss=0.04145, over 15289.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2844, pruned_loss=0.05173, over 3024400.88 frames. ], batch size: 190, lr: 8.36e-03, grad_scale: 8.0 2023-04-28 22:45:51,490 INFO [optim.py:368] (0/8) 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 22:46:57,784 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=79941.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:46:57,909 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79941.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:47:20,943 INFO [train.py:904] (0/8) Epoch 8, batch 8900, loss[loss=0.1998, simple_loss=0.2883, pruned_loss=0.05566, over 16826.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2839, pruned_loss=0.05062, over 3029521.56 frames. ], batch size: 124, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:48:54,506 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=79989.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:48:54,540 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=79989.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:49:22,543 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-80000.pt 2023-04-28 22:49:29,485 INFO [train.py:904] (0/8) Epoch 8, batch 8950, loss[loss=0.2021, simple_loss=0.2825, pruned_loss=0.06082, over 12497.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.284, pruned_loss=0.05081, over 3053478.25 frames. ], batch size: 247, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:49:50,486 INFO [optim.py:368] (0/8) 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 22:49:54,213 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80013.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:00,379 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0436, 2.7947, 2.7743, 2.0428, 2.5011, 2.0617, 2.6134, 2.9001], device='cuda:0'), covar=tensor([0.0274, 0.0645, 0.0494, 0.1489, 0.0721, 0.1037, 0.0659, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0127, 0.0151, 0.0137, 0.0129, 0.0122, 0.0132, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-28 22:50:08,341 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80020.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:50:32,409 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80031.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 22:50:46,187 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80037.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:51:17,202 INFO [train.py:904] (0/8) Epoch 8, batch 9000, loss[loss=0.1969, simple_loss=0.2719, pruned_loss=0.06092, over 12404.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2805, pruned_loss=0.04909, over 3076193.60 frames. ], batch size: 247, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:51:17,203 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 22:51:25,055 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2790, 3.4481, 1.9766, 3.6270, 2.5653, 3.5957, 2.2324, 2.9018], device='cuda:0'), covar=tensor([0.0208, 0.0301, 0.1467, 0.0126, 0.0715, 0.0462, 0.1280, 0.0587], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0149, 0.0177, 0.0098, 0.0159, 0.0186, 0.0187, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-28 22:51:27,535 INFO [train.py:938] (0/8) Epoch 8, validation: loss=0.1608, simple_loss=0.265, pruned_loss=0.02828, over 944034.00 frames. 2023-04-28 22:51:27,536 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 22:52:04,340 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80068.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:52:26,104 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80079.0, num_to_drop=1, layers_to_drop={0} 2023-04-28 22:52:44,322 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5683, 3.6440, 2.7761, 2.1079, 2.4017, 2.1176, 3.7753, 3.4120], device='cuda:0'), covar=tensor([0.2335, 0.0595, 0.1381, 0.2146, 0.1944, 0.1637, 0.0391, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0243, 0.0267, 0.0254, 0.0250, 0.0204, 0.0248, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:53:14,319 INFO [train.py:904] (0/8) Epoch 8, batch 9050, loss[loss=0.189, simple_loss=0.2801, pruned_loss=0.04899, over 16509.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2823, pruned_loss=0.05024, over 3071606.67 frames. ], batch size: 75, lr: 8.35e-03, grad_scale: 8.0 2023-04-28 22:53:35,360 INFO [optim.py:368] (0/8) 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 22:53:48,906 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80118.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:54:36,456 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2447, 1.7626, 1.3566, 1.4579, 2.1141, 1.7900, 2.1557, 2.2942], device='cuda:0'), covar=tensor([0.0077, 0.0384, 0.0430, 0.0415, 0.0188, 0.0310, 0.0127, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0104, 0.0181, 0.0175, 0.0179, 0.0174, 0.0180, 0.0167, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:54:59,320 INFO [train.py:904] (0/8) Epoch 8, batch 9100, loss[loss=0.1785, simple_loss=0.2797, pruned_loss=0.03872, over 16877.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2812, pruned_loss=0.05044, over 3077563.30 frames. ], batch size: 102, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:55:48,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5413, 3.5004, 3.4967, 3.0697, 3.4540, 1.8626, 3.2897, 2.8816], device='cuda:0'), covar=tensor([0.0094, 0.0091, 0.0122, 0.0223, 0.0083, 0.2075, 0.0114, 0.0178], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0094, 0.0139, 0.0131, 0.0110, 0.0161, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:56:48,077 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7423, 1.9415, 1.4871, 1.7190, 2.4585, 2.1139, 2.6524, 2.7264], device='cuda:0'), covar=tensor([0.0060, 0.0331, 0.0415, 0.0394, 0.0177, 0.0286, 0.0116, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0182, 0.0177, 0.0179, 0.0174, 0.0181, 0.0168, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 22:56:59,482 INFO [train.py:904] (0/8) Epoch 8, batch 9150, loss[loss=0.1942, simple_loss=0.2938, pruned_loss=0.04728, over 17125.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2821, pruned_loss=0.0501, over 3095879.18 frames. ], batch size: 49, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:57:20,189 INFO [optim.py:368] (0/8) 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 22:58:26,867 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80241.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:58:44,955 INFO [train.py:904] (0/8) Epoch 8, batch 9200, loss[loss=0.1869, simple_loss=0.2775, pruned_loss=0.04815, over 16740.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2782, pruned_loss=0.04933, over 3091545.17 frames. ], batch size: 124, lr: 8.34e-03, grad_scale: 8.0 2023-04-28 22:59:42,521 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80283.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 22:59:55,915 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80289.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:00:22,124 INFO [train.py:904] (0/8) Epoch 8, batch 9250, loss[loss=0.1857, simple_loss=0.2752, pruned_loss=0.04809, over 16798.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2783, pruned_loss=0.04966, over 3086165.99 frames. ], batch size: 124, lr: 8.34e-03, grad_scale: 4.0 2023-04-28 23:00:42,873 INFO [optim.py:368] (0/8) 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 23:00:44,290 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80313.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:01:34,654 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-28 23:01:36,423 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-28 23:01:45,506 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-28 23:01:55,875 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80344.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:02:12,083 INFO [train.py:904] (0/8) Epoch 8, batch 9300, loss[loss=0.1531, simple_loss=0.2495, pruned_loss=0.02834, over 16715.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2765, pruned_loss=0.049, over 3076980.64 frames. ], batch size: 83, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:02:31,765 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80361.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:03:23,022 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4839, 3.7207, 3.9141, 1.7809, 4.1764, 4.2187, 3.1355, 3.0641], device='cuda:0'), covar=tensor([0.0740, 0.0152, 0.0157, 0.1200, 0.0040, 0.0074, 0.0325, 0.0402], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0092, 0.0078, 0.0134, 0.0062, 0.0084, 0.0112, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 23:03:38,113 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3718, 4.6589, 4.4455, 4.4446, 4.1215, 4.1399, 4.2028, 4.6589], device='cuda:0'), covar=tensor([0.0841, 0.0809, 0.0992, 0.0535, 0.0714, 0.1186, 0.0798, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0445, 0.0565, 0.0472, 0.0385, 0.0350, 0.0376, 0.0470, 0.0420], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:03:42,316 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80393.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:03:57,485 INFO [train.py:904] (0/8) Epoch 8, batch 9350, loss[loss=0.1935, simple_loss=0.2806, pruned_loss=0.05316, over 17000.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2763, pruned_loss=0.04856, over 3102000.00 frames. ], batch size: 109, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:04:07,926 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7122, 2.1150, 2.2284, 4.3458, 2.0525, 2.5828, 2.1926, 2.2132], device='cuda:0'), covar=tensor([0.0680, 0.3043, 0.1877, 0.0267, 0.3524, 0.1863, 0.2756, 0.3045], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0352, 0.0301, 0.0306, 0.0387, 0.0387, 0.0316, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:04:22,284 INFO [optim.py:368] (0/8) 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 23:04:33,562 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80418.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:05:40,907 INFO [train.py:904] (0/8) Epoch 8, batch 9400, loss[loss=0.1621, simple_loss=0.2459, pruned_loss=0.0392, over 12919.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2756, pruned_loss=0.04828, over 3079947.88 frames. ], batch size: 248, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:05:46,234 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80454.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:05:54,615 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3664, 2.4543, 2.0171, 2.2181, 2.8987, 2.5938, 3.1995, 3.0823], device='cuda:0'), covar=tensor([0.0048, 0.0270, 0.0328, 0.0298, 0.0168, 0.0207, 0.0116, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0180, 0.0176, 0.0177, 0.0174, 0.0179, 0.0168, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:06:04,678 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80463.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:09,459 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80466.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:06:51,713 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3493, 4.6336, 4.4431, 4.4174, 4.0720, 4.0412, 4.1569, 4.6245], device='cuda:0'), covar=tensor([0.0839, 0.0752, 0.0758, 0.0534, 0.0653, 0.1233, 0.0822, 0.0816], device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0553, 0.0462, 0.0378, 0.0344, 0.0370, 0.0460, 0.0413], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:07:07,272 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80494.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:07:19,876 INFO [train.py:904] (0/8) Epoch 8, batch 9450, loss[loss=0.1767, simple_loss=0.2665, pruned_loss=0.04348, over 16754.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2779, pruned_loss=0.04871, over 3089905.94 frames. ], batch size: 62, lr: 8.33e-03, grad_scale: 4.0 2023-04-28 23:07:38,813 INFO [optim.py:368] (0/8) 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 23:07:54,008 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-28 23:07:56,208 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-28 23:08:04,395 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80524.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:08:58,872 INFO [train.py:904] (0/8) Epoch 8, batch 9500, loss[loss=0.182, simple_loss=0.271, pruned_loss=0.04652, over 16378.00 frames. ], tot_loss[loss=0.187, simple_loss=0.277, pruned_loss=0.04845, over 3068427.10 frames. ], batch size: 35, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:09:08,276 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80555.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:10:46,570 INFO [train.py:904] (0/8) Epoch 8, batch 9550, loss[loss=0.196, simple_loss=0.2943, pruned_loss=0.04881, over 16884.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2767, pruned_loss=0.04854, over 3084735.68 frames. ], batch size: 96, lr: 8.32e-03, grad_scale: 4.0 2023-04-28 23:11:10,123 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.584e+02 3.110e+02 3.705e+02 8.746e+02, threshold=6.220e+02, percent-clipped=3.0 2023-04-28 23:12:03,974 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80639.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:12:10,089 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-28 23:12:27,096 INFO [train.py:904] (0/8) Epoch 8, batch 9600, loss[loss=0.1819, simple_loss=0.2666, pruned_loss=0.04856, over 12397.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2783, pruned_loss=0.04946, over 3087201.95 frames. ], batch size: 248, lr: 8.32e-03, grad_scale: 8.0 2023-04-28 23:14:15,059 INFO [train.py:904] (0/8) Epoch 8, batch 9650, loss[loss=0.2205, simple_loss=0.3171, pruned_loss=0.06193, over 16845.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2799, pruned_loss=0.04954, over 3086980.62 frames. ], batch size: 124, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:14:33,445 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1343, 3.2420, 3.6079, 3.5666, 3.5790, 3.3292, 3.3762, 3.4193], device='cuda:0'), covar=tensor([0.0367, 0.0636, 0.0395, 0.0441, 0.0435, 0.0448, 0.0773, 0.0436], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0272, 0.0275, 0.0271, 0.0310, 0.0293, 0.0378, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-28 23:14:41,955 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-28 23:14:42,863 INFO [optim.py:368] (0/8) 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 23:15:29,966 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80736.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:15:59,214 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80749.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:16:03,279 INFO [train.py:904] (0/8) Epoch 8, batch 9700, loss[loss=0.2092, simple_loss=0.2896, pruned_loss=0.06443, over 11978.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2787, pruned_loss=0.04973, over 3059865.11 frames. ], batch size: 248, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:16:46,618 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=80773.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:38,618 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80797.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:17:46,353 INFO [train.py:904] (0/8) Epoch 8, batch 9750, loss[loss=0.1929, simple_loss=0.2863, pruned_loss=0.04973, over 16274.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2782, pruned_loss=0.05008, over 3060128.44 frames. ], batch size: 166, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:18:08,281 INFO [optim.py:368] (0/8) 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 23:18:20,467 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80819.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:18:55,718 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=80834.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:24,257 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=80850.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:19:26,325 INFO [train.py:904] (0/8) Epoch 8, batch 9800, loss[loss=0.2078, simple_loss=0.3019, pruned_loss=0.05688, over 16902.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2778, pruned_loss=0.04875, over 3062629.36 frames. ], batch size: 116, lr: 8.31e-03, grad_scale: 8.0 2023-04-28 23:19:47,764 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-28 23:20:06,815 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-28 23:21:11,983 INFO [train.py:904] (0/8) Epoch 8, batch 9850, loss[loss=0.1865, simple_loss=0.2827, pruned_loss=0.04519, over 16382.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2786, pruned_loss=0.04824, over 3055511.77 frames. ], batch size: 146, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:21:33,322 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.396e+02 3.091e+02 3.765e+02 1.204e+03, threshold=6.182e+02, percent-clipped=3.0 2023-04-28 23:22:36,825 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=80939.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:23:04,053 INFO [train.py:904] (0/8) Epoch 8, batch 9900, loss[loss=0.1975, simple_loss=0.289, pruned_loss=0.05297, over 15293.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2792, pruned_loss=0.04805, over 3057729.12 frames. ], batch size: 191, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:23:11,345 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5352, 3.5346, 3.4678, 3.1005, 3.4730, 1.9642, 3.2828, 3.0863], device='cuda:0'), covar=tensor([0.0104, 0.0102, 0.0138, 0.0184, 0.0091, 0.1899, 0.0108, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0106, 0.0095, 0.0137, 0.0127, 0.0110, 0.0163, 0.0123, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-28 23:23:59,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3564, 4.6724, 4.4365, 4.4763, 4.1532, 4.1206, 4.2921, 4.6697], device='cuda:0'), covar=tensor([0.0926, 0.0858, 0.0822, 0.0560, 0.0727, 0.1264, 0.0793, 0.0833], device='cuda:0'), in_proj_covar=tensor([0.0454, 0.0577, 0.0473, 0.0394, 0.0356, 0.0384, 0.0475, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:24:29,136 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=80987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:24:57,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4953, 4.8333, 4.5979, 4.6100, 4.2986, 4.2434, 4.2824, 4.8270], device='cuda:0'), covar=tensor([0.0995, 0.0862, 0.0878, 0.0504, 0.0758, 0.1143, 0.0937, 0.0954], device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0575, 0.0472, 0.0392, 0.0354, 0.0382, 0.0474, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:25:03,336 INFO [train.py:904] (0/8) Epoch 8, batch 9950, loss[loss=0.1945, simple_loss=0.2883, pruned_loss=0.05035, over 16743.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2812, pruned_loss=0.0483, over 3063999.91 frames. ], batch size: 134, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:25:29,446 INFO [optim.py:368] (0/8) 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 23:27:01,960 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81049.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:27:07,190 INFO [train.py:904] (0/8) Epoch 8, batch 10000, loss[loss=0.1816, simple_loss=0.2658, pruned_loss=0.04872, over 12650.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2794, pruned_loss=0.04754, over 3088420.53 frames. ], batch size: 250, lr: 8.30e-03, grad_scale: 8.0 2023-04-28 23:28:04,991 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 23:28:30,677 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81092.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:28:40,343 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 8, batch 10050, loss[loss=0.1853, simple_loss=0.2801, pruned_loss=0.04527, over 16721.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2799, pruned_loss=0.04785, over 3079542.85 frames. ], batch size: 83, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:29:08,275 INFO [optim.py:368] (0/8) 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 23:29:15,170 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:20,585 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81119.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:35,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8042, 3.8137, 4.0421, 2.1242, 4.3207, 4.3800, 3.3240, 3.2674], device='cuda:0'), covar=tensor([0.0644, 0.0157, 0.0150, 0.1120, 0.0042, 0.0063, 0.0294, 0.0332], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0092, 0.0076, 0.0135, 0.0063, 0.0083, 0.0112, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 23:29:39,837 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81129.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:29:44,196 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-28 23:30:10,854 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6564, 2.7714, 2.3910, 4.0792, 2.8268, 4.0236, 1.3340, 3.0589], device='cuda:0'), covar=tensor([0.1276, 0.0580, 0.1086, 0.0117, 0.0130, 0.0312, 0.1435, 0.0585], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0148, 0.0169, 0.0111, 0.0172, 0.0194, 0.0168, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 23:30:18,758 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81150.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:30:21,531 INFO [train.py:904] (0/8) Epoch 8, batch 10100, loss[loss=0.1527, simple_loss=0.2468, pruned_loss=0.02926, over 15385.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2807, pruned_loss=0.04842, over 3083156.93 frames. ], batch size: 190, lr: 8.29e-03, grad_scale: 8.0 2023-04-28 23:30:42,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6289, 2.0592, 1.6417, 1.7756, 2.4694, 2.1132, 2.5229, 2.5994], device='cuda:0'), covar=tensor([0.0069, 0.0296, 0.0368, 0.0349, 0.0170, 0.0274, 0.0117, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0105, 0.0183, 0.0178, 0.0179, 0.0176, 0.0181, 0.0168, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:30:51,344 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81167.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:14,318 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81177.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:16,892 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2176, 3.5124, 3.6200, 2.5979, 3.3211, 3.5471, 3.3582, 2.0187], device='cuda:0'), covar=tensor([0.0354, 0.0031, 0.0029, 0.0219, 0.0059, 0.0060, 0.0049, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0060, 0.0062, 0.0117, 0.0067, 0.0075, 0.0067, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 23:31:37,454 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81198.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:31:42,610 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-8.pt 2023-04-28 23:32:08,885 INFO [train.py:904] (0/8) Epoch 9, batch 0, loss[loss=0.3303, simple_loss=0.3576, pruned_loss=0.1515, over 16877.00 frames. ], tot_loss[loss=0.3303, simple_loss=0.3576, pruned_loss=0.1515, over 16877.00 frames. ], batch size: 109, lr: 7.85e-03, grad_scale: 8.0 2023-04-28 23:32:08,886 INFO [train.py:929] (0/8) Computing validation loss 2023-04-28 23:32:16,264 INFO [train.py:938] (0/8) Epoch 9, validation: loss=0.1602, simple_loss=0.2637, pruned_loss=0.02837, over 944034.00 frames. 2023-04-28 23:32:16,264 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-28 23:32:33,672 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5180, 4.2854, 4.4909, 4.6664, 4.8448, 4.4013, 4.7633, 4.7420], device='cuda:0'), covar=tensor([0.1468, 0.1191, 0.1774, 0.0870, 0.0682, 0.0792, 0.0721, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0443, 0.0540, 0.0663, 0.0549, 0.0414, 0.0407, 0.0434, 0.0477], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:32:36,786 INFO [optim.py:368] (0/8) 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 23:33:25,080 INFO [train.py:904] (0/8) Epoch 9, batch 50, loss[loss=0.2263, simple_loss=0.3068, pruned_loss=0.07291, over 15330.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.2944, pruned_loss=0.07229, over 757545.92 frames. ], batch size: 190, lr: 7.85e-03, grad_scale: 1.0 2023-04-28 23:34:12,119 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8148, 4.2219, 4.4701, 3.4051, 3.8350, 4.3428, 3.9572, 2.5634], device='cuda:0'), covar=tensor([0.0322, 0.0037, 0.0023, 0.0198, 0.0065, 0.0052, 0.0045, 0.0328], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0063, 0.0064, 0.0120, 0.0069, 0.0077, 0.0069, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 23:34:31,228 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81299.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:34,102 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81301.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:34:35,504 INFO [train.py:904] (0/8) Epoch 9, batch 100, loss[loss=0.1729, simple_loss=0.2526, pruned_loss=0.04657, over 17203.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.287, pruned_loss=0.06503, over 1338518.08 frames. ], batch size: 44, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:34:46,992 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-28 23:34:54,540 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.676e+02 3.088e+02 4.301e+02 1.054e+03, threshold=6.177e+02, percent-clipped=2.0 2023-04-28 23:35:42,957 INFO [train.py:904] (0/8) Epoch 9, batch 150, loss[loss=0.2204, simple_loss=0.2917, pruned_loss=0.07452, over 16476.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2859, pruned_loss=0.0632, over 1782255.90 frames. ], batch size: 75, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:35:45,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1136, 5.0205, 4.9315, 4.5589, 4.4795, 4.9779, 5.0462, 4.5847], device='cuda:0'), covar=tensor([0.0529, 0.0412, 0.0273, 0.0245, 0.0991, 0.0317, 0.0246, 0.0617], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0249, 0.0248, 0.0221, 0.0269, 0.0253, 0.0170, 0.0285], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:35:55,871 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81360.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:35:58,301 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81362.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:40,118 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81392.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:36:54,432 INFO [train.py:904] (0/8) Epoch 9, batch 200, loss[loss=0.1951, simple_loss=0.2888, pruned_loss=0.05071, over 17281.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2866, pruned_loss=0.0642, over 2121337.68 frames. ], batch size: 52, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:37:13,049 INFO [optim.py:368] (0/8) 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 23:37:19,500 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6703, 4.6794, 4.5394, 4.1191, 4.6000, 1.8639, 4.3231, 4.2538], device='cuda:0'), covar=tensor([0.0092, 0.0073, 0.0120, 0.0246, 0.0070, 0.2018, 0.0107, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0113, 0.0100, 0.0146, 0.0136, 0.0117, 0.0170, 0.0131, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:37:23,756 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2214, 3.2497, 3.3433, 1.8624, 3.5329, 3.5735, 2.7296, 2.6222], device='cuda:0'), covar=tensor([0.0752, 0.0164, 0.0180, 0.1038, 0.0065, 0.0108, 0.0442, 0.0420], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0093, 0.0079, 0.0138, 0.0066, 0.0088, 0.0116, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 23:37:31,256 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81429.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:37:45,245 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81440.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:01,971 INFO [train.py:904] (0/8) Epoch 9, batch 250, loss[loss=0.192, simple_loss=0.2804, pruned_loss=0.05182, over 16708.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2843, pruned_loss=0.06426, over 2393715.85 frames. ], batch size: 57, lr: 7.84e-03, grad_scale: 1.0 2023-04-28 23:38:29,690 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81472.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:36,207 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81477.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:38:47,473 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-28 23:39:10,406 INFO [train.py:904] (0/8) Epoch 9, batch 300, loss[loss=0.2022, simple_loss=0.2757, pruned_loss=0.06434, over 12036.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2807, pruned_loss=0.06109, over 2592813.95 frames. ], batch size: 246, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:39:29,676 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.655e+02 2.475e+02 2.919e+02 3.755e+02 7.155e+02, threshold=5.837e+02, percent-clipped=3.0 2023-04-28 23:39:32,600 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1480, 3.9374, 4.1855, 4.3916, 4.4819, 4.0706, 4.2322, 4.4616], device='cuda:0'), covar=tensor([0.1254, 0.1117, 0.1384, 0.0621, 0.0523, 0.1114, 0.2132, 0.0520], device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0598, 0.0738, 0.0607, 0.0454, 0.0450, 0.0479, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:39:32,743 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5745, 3.5160, 2.8133, 2.2767, 2.4628, 2.1948, 3.6042, 3.3468], device='cuda:0'), covar=tensor([0.2387, 0.0738, 0.1321, 0.2129, 0.2228, 0.1743, 0.0520, 0.1175], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0253, 0.0278, 0.0266, 0.0264, 0.0215, 0.0257, 0.0276], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:40:17,845 INFO [train.py:904] (0/8) Epoch 9, batch 350, loss[loss=0.1931, simple_loss=0.2642, pruned_loss=0.06103, over 16305.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2791, pruned_loss=0.06006, over 2764002.15 frames. ], batch size: 165, lr: 7.83e-03, grad_scale: 1.0 2023-04-28 23:40:44,832 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7421, 3.3213, 2.6485, 5.1540, 4.4194, 4.7109, 1.6668, 3.5093], device='cuda:0'), covar=tensor([0.1288, 0.0550, 0.1107, 0.0132, 0.0264, 0.0329, 0.1384, 0.0641], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0152, 0.0174, 0.0121, 0.0186, 0.0205, 0.0173, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 23:41:23,835 INFO [train.py:904] (0/8) Epoch 9, batch 400, loss[loss=0.2381, simple_loss=0.3, pruned_loss=0.08813, over 16718.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2768, pruned_loss=0.05954, over 2887808.40 frames. ], batch size: 134, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:41:43,587 INFO [optim.py:368] (0/8) 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 23:41:51,212 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6349, 2.7949, 2.5850, 4.8618, 4.0332, 4.4216, 1.4681, 3.1152], device='cuda:0'), covar=tensor([0.1405, 0.0680, 0.1126, 0.0129, 0.0293, 0.0334, 0.1466, 0.0746], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0152, 0.0174, 0.0121, 0.0187, 0.0205, 0.0173, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-28 23:42:27,839 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2860, 5.2405, 5.0813, 4.7398, 4.5930, 5.1158, 5.2237, 4.7682], device='cuda:0'), covar=tensor([0.0595, 0.0346, 0.0257, 0.0245, 0.1181, 0.0367, 0.0228, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0267, 0.0263, 0.0236, 0.0289, 0.0271, 0.0181, 0.0304], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:42:33,329 INFO [train.py:904] (0/8) Epoch 9, batch 450, loss[loss=0.2165, simple_loss=0.2827, pruned_loss=0.07518, over 16876.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2743, pruned_loss=0.05825, over 2985594.94 frames. ], batch size: 116, lr: 7.83e-03, grad_scale: 2.0 2023-04-28 23:42:37,321 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=81657.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:42:58,056 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81670.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:43:42,340 INFO [train.py:904] (0/8) Epoch 9, batch 500, loss[loss=0.188, simple_loss=0.2771, pruned_loss=0.04944, over 17114.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2727, pruned_loss=0.05773, over 3052252.83 frames. ], batch size: 49, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:44:01,901 INFO [optim.py:368] (0/8) 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 23:44:22,683 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=81731.0, num_to_drop=1, layers_to_drop={1} 2023-04-28 23:44:50,830 INFO [train.py:904] (0/8) Epoch 9, batch 550, loss[loss=0.1635, simple_loss=0.241, pruned_loss=0.04303, over 16004.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2718, pruned_loss=0.05711, over 3110773.11 frames. ], batch size: 35, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:45:19,844 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81772.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:45:43,873 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3595, 4.4865, 4.6019, 2.4829, 4.8959, 4.9399, 3.5951, 3.8367], device='cuda:0'), covar=tensor([0.0589, 0.0137, 0.0214, 0.0988, 0.0051, 0.0067, 0.0315, 0.0301], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0096, 0.0084, 0.0143, 0.0069, 0.0093, 0.0120, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 23:45:53,202 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7444, 3.9112, 2.1011, 4.4982, 2.7279, 4.4555, 2.1771, 3.1555], device='cuda:0'), covar=tensor([0.0230, 0.0363, 0.1642, 0.0142, 0.0952, 0.0377, 0.1589, 0.0679], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0163, 0.0186, 0.0112, 0.0166, 0.0198, 0.0195, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-28 23:46:02,626 INFO [train.py:904] (0/8) Epoch 9, batch 600, loss[loss=0.2081, simple_loss=0.2681, pruned_loss=0.07411, over 16767.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2711, pruned_loss=0.05754, over 3163057.66 frames. ], batch size: 134, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:46:14,149 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-28 23:46:17,738 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6598, 3.9687, 4.1018, 2.2774, 4.2496, 4.3160, 3.2251, 3.1077], device='cuda:0'), covar=tensor([0.0763, 0.0145, 0.0157, 0.1089, 0.0064, 0.0106, 0.0338, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0097, 0.0084, 0.0145, 0.0069, 0.0094, 0.0121, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 23:46:21,578 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.753e+02 2.670e+02 3.296e+02 4.052e+02 8.860e+02, threshold=6.592e+02, percent-clipped=6.0 2023-04-28 23:46:25,734 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9383, 2.1143, 2.2893, 4.7570, 2.0028, 2.6693, 2.2812, 2.4122], device='cuda:0'), covar=tensor([0.0751, 0.3202, 0.2011, 0.0280, 0.3777, 0.2147, 0.2733, 0.3389], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0370, 0.0312, 0.0323, 0.0399, 0.0408, 0.0330, 0.0434], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:46:26,575 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=81820.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:47:09,571 INFO [train.py:904] (0/8) Epoch 9, batch 650, loss[loss=0.1701, simple_loss=0.2528, pruned_loss=0.04368, over 17223.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2701, pruned_loss=0.05646, over 3206595.40 frames. ], batch size: 44, lr: 7.82e-03, grad_scale: 2.0 2023-04-28 23:48:18,164 INFO [train.py:904] (0/8) Epoch 9, batch 700, loss[loss=0.2232, simple_loss=0.2916, pruned_loss=0.07744, over 16860.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2707, pruned_loss=0.05707, over 3235075.12 frames. ], batch size: 116, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:48:37,202 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.553e+02 3.037e+02 4.738e+02 2.852e+03, threshold=6.073e+02, percent-clipped=12.0 2023-04-28 23:49:03,482 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3921, 5.3082, 5.1414, 4.8067, 4.7656, 5.2096, 5.2768, 4.8407], device='cuda:0'), covar=tensor([0.0514, 0.0392, 0.0246, 0.0226, 0.1021, 0.0384, 0.0187, 0.0629], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0272, 0.0269, 0.0240, 0.0294, 0.0276, 0.0184, 0.0310], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:49:25,155 INFO [train.py:904] (0/8) Epoch 9, batch 750, loss[loss=0.1673, simple_loss=0.2504, pruned_loss=0.04213, over 16832.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2706, pruned_loss=0.05644, over 3247942.37 frames. ], batch size: 42, lr: 7.81e-03, grad_scale: 2.0 2023-04-28 23:49:29,184 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81955.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:49:31,596 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=81957.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:49:46,749 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-28 23:50:14,442 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=81987.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:32,343 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-82000.pt 2023-04-28 23:50:37,840 INFO [train.py:904] (0/8) Epoch 9, batch 800, loss[loss=0.1818, simple_loss=0.2658, pruned_loss=0.04892, over 16715.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2698, pruned_loss=0.05544, over 3265405.94 frames. ], batch size: 62, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:50:39,918 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82003.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:42,266 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82005.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:50:56,933 INFO [optim.py:368] (0/8) 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 23:51:10,872 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82026.0, num_to_drop=1, layers_to_drop={2} 2023-04-28 23:51:40,250 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82048.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:51:45,047 INFO [train.py:904] (0/8) Epoch 9, batch 850, loss[loss=0.1943, simple_loss=0.2631, pruned_loss=0.0628, over 16662.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2688, pruned_loss=0.05479, over 3278037.16 frames. ], batch size: 89, lr: 7.81e-03, grad_scale: 4.0 2023-04-28 23:51:49,626 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82055.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:52:52,822 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4008, 2.9452, 2.5778, 2.2168, 2.1584, 2.1587, 2.8069, 2.8297], device='cuda:0'), covar=tensor([0.2170, 0.0736, 0.1425, 0.1767, 0.1802, 0.1739, 0.0521, 0.0929], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0255, 0.0280, 0.0267, 0.0273, 0.0216, 0.0259, 0.0284], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:52:54,655 INFO [train.py:904] (0/8) Epoch 9, batch 900, loss[loss=0.1763, simple_loss=0.267, pruned_loss=0.04279, over 17282.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2683, pruned_loss=0.05438, over 3276614.67 frames. ], batch size: 52, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:53:13,856 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.661e+02 2.604e+02 3.157e+02 3.589e+02 5.638e+02, threshold=6.315e+02, percent-clipped=0.0 2023-04-28 23:53:14,352 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82116.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:54:03,809 INFO [train.py:904] (0/8) Epoch 9, batch 950, loss[loss=0.223, simple_loss=0.2824, pruned_loss=0.08177, over 16787.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2689, pruned_loss=0.05483, over 3291484.99 frames. ], batch size: 124, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:54:10,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1400, 5.7519, 5.9681, 5.6762, 5.8534, 6.3025, 6.0099, 5.6384], device='cuda:0'), covar=tensor([0.0814, 0.1752, 0.1603, 0.2035, 0.2428, 0.1011, 0.1104, 0.2407], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0465, 0.0497, 0.0409, 0.0551, 0.0523, 0.0395, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 23:54:15,821 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8306, 4.7945, 4.7067, 4.1726, 4.7757, 2.0560, 4.4440, 4.4730], device='cuda:0'), covar=tensor([0.0089, 0.0079, 0.0134, 0.0321, 0.0072, 0.2074, 0.0121, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0107, 0.0156, 0.0148, 0.0124, 0.0175, 0.0141, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-28 23:55:02,743 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7645, 3.9347, 4.1634, 2.0335, 4.4516, 4.4828, 3.1219, 3.5426], device='cuda:0'), covar=tensor([0.0681, 0.0183, 0.0241, 0.1107, 0.0055, 0.0114, 0.0384, 0.0303], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0097, 0.0085, 0.0141, 0.0068, 0.0095, 0.0119, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-28 23:55:11,100 INFO [train.py:904] (0/8) Epoch 9, batch 1000, loss[loss=0.1756, simple_loss=0.2671, pruned_loss=0.04209, over 17245.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2674, pruned_loss=0.05455, over 3306516.71 frames. ], batch size: 52, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:55:11,842 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-28 23:55:31,954 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.413e+02 2.342e+02 2.920e+02 3.464e+02 5.943e+02, threshold=5.839e+02, percent-clipped=0.0 2023-04-28 23:56:20,447 INFO [train.py:904] (0/8) Epoch 9, batch 1050, loss[loss=0.1622, simple_loss=0.2571, pruned_loss=0.03362, over 17086.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2674, pruned_loss=0.05453, over 3309532.44 frames. ], batch size: 47, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:21,079 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-28 23:57:28,548 INFO [train.py:904] (0/8) Epoch 9, batch 1100, loss[loss=0.1898, simple_loss=0.256, pruned_loss=0.06181, over 16850.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2667, pruned_loss=0.05338, over 3319209.87 frames. ], batch size: 116, lr: 7.80e-03, grad_scale: 4.0 2023-04-28 23:57:47,217 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.669e+02 2.514e+02 3.056e+02 3.616e+02 1.290e+03, threshold=6.113e+02, percent-clipped=7.0 2023-04-28 23:57:49,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2234, 5.1782, 5.0356, 4.6664, 4.5891, 5.0529, 5.0811, 4.7050], device='cuda:0'), covar=tensor([0.0510, 0.0308, 0.0234, 0.0256, 0.1090, 0.0372, 0.0256, 0.0602], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0284, 0.0279, 0.0251, 0.0307, 0.0288, 0.0192, 0.0322], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-28 23:58:01,928 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82326.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:15,347 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-28 23:58:24,259 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82343.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:24,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0680, 4.6258, 4.6245, 3.3077, 4.0026, 4.5006, 4.0132, 2.7607], device='cuda:0'), covar=tensor([0.0305, 0.0019, 0.0026, 0.0224, 0.0046, 0.0062, 0.0046, 0.0285], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0066, 0.0065, 0.0120, 0.0070, 0.0080, 0.0071, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-28 23:58:28,080 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82346.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:58:35,310 INFO [train.py:904] (0/8) Epoch 9, batch 1150, loss[loss=0.1825, simple_loss=0.2514, pruned_loss=0.05678, over 16497.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2653, pruned_loss=0.05255, over 3316055.93 frames. ], batch size: 75, lr: 7.79e-03, grad_scale: 4.0 2023-04-28 23:59:04,897 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82374.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:44,522 INFO [train.py:904] (0/8) Epoch 9, batch 1200, loss[loss=0.1797, simple_loss=0.251, pruned_loss=0.05425, over 16826.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2645, pruned_loss=0.05206, over 3327967.47 frames. ], batch size: 102, lr: 7.79e-03, grad_scale: 8.0 2023-04-28 23:59:45,387 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-28 23:59:50,742 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82407.0, num_to_drop=0, layers_to_drop=set() 2023-04-28 23:59:55,222 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82411.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:00:02,695 INFO [optim.py:368] (0/8) 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 00:00:50,247 INFO [train.py:904] (0/8) Epoch 9, batch 1250, loss[loss=0.1924, simple_loss=0.2815, pruned_loss=0.05166, over 17055.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2652, pruned_loss=0.05345, over 3320002.10 frames. ], batch size: 50, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:01:47,045 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82493.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:01:58,483 INFO [train.py:904] (0/8) Epoch 9, batch 1300, loss[loss=0.1799, simple_loss=0.272, pruned_loss=0.04387, over 17061.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.266, pruned_loss=0.05345, over 3327555.84 frames. ], batch size: 50, lr: 7.79e-03, grad_scale: 4.0 2023-04-29 00:02:18,040 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.404e+02 2.965e+02 3.987e+02 6.881e+02, threshold=5.930e+02, percent-clipped=4.0 2023-04-29 00:02:55,960 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0268, 4.4592, 4.5815, 3.3598, 3.8792, 4.3758, 3.9505, 2.7208], device='cuda:0'), covar=tensor([0.0311, 0.0027, 0.0019, 0.0210, 0.0054, 0.0064, 0.0047, 0.0288], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0067, 0.0065, 0.0121, 0.0070, 0.0080, 0.0072, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:03:05,226 INFO [train.py:904] (0/8) Epoch 9, batch 1350, loss[loss=0.1819, simple_loss=0.2825, pruned_loss=0.04065, over 17125.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.267, pruned_loss=0.05355, over 3316415.76 frames. ], batch size: 48, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:03:07,986 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82554.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:03:29,829 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:04:10,125 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7346, 3.8162, 2.0513, 4.1317, 2.7562, 4.0000, 2.1246, 2.9890], device='cuda:0'), covar=tensor([0.0165, 0.0260, 0.1502, 0.0214, 0.0722, 0.0520, 0.1393, 0.0590], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0164, 0.0186, 0.0117, 0.0166, 0.0205, 0.0196, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 00:04:12,576 INFO [train.py:904] (0/8) Epoch 9, batch 1400, loss[loss=0.1715, simple_loss=0.2571, pruned_loss=0.0429, over 17199.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2666, pruned_loss=0.0534, over 3321817.16 frames. ], batch size: 46, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:04:33,568 INFO [optim.py:368] (0/8) 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 00:04:45,756 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7380, 6.1341, 5.8272, 5.8704, 5.3464, 5.2900, 5.5735, 6.2072], device='cuda:0'), covar=tensor([0.1032, 0.0862, 0.0988, 0.0681, 0.0696, 0.0655, 0.0849, 0.0929], device='cuda:0'), in_proj_covar=tensor([0.0502, 0.0648, 0.0537, 0.0438, 0.0398, 0.0416, 0.0533, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:04:53,151 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=82630.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:09,566 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:05:22,662 INFO [train.py:904] (0/8) Epoch 9, batch 1450, loss[loss=0.1881, simple_loss=0.2553, pruned_loss=0.0604, over 16919.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2656, pruned_loss=0.05301, over 3321187.61 frames. ], batch size: 109, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:06:14,229 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8365, 4.2964, 3.2302, 2.1480, 2.9956, 2.4263, 4.6155, 4.0011], device='cuda:0'), covar=tensor([0.2602, 0.0654, 0.1347, 0.2357, 0.2586, 0.1750, 0.0382, 0.0916], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0258, 0.0279, 0.0268, 0.0278, 0.0217, 0.0260, 0.0287], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:06:16,063 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:29,720 INFO [train.py:904] (0/8) Epoch 9, batch 1500, loss[loss=0.1685, simple_loss=0.2567, pruned_loss=0.04012, over 17253.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2659, pruned_loss=0.05338, over 3318167.91 frames. ], batch size: 45, lr: 7.78e-03, grad_scale: 4.0 2023-04-29 00:06:29,983 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:42,820 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=82711.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:06:51,614 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.540e+02 3.048e+02 3.562e+02 8.559e+02, threshold=6.096e+02, percent-clipped=4.0 2023-04-29 00:06:52,134 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3365, 5.1096, 5.3147, 5.5333, 5.6929, 4.9557, 5.5823, 5.6208], device='cuda:0'), covar=tensor([0.1143, 0.0888, 0.1328, 0.0494, 0.0420, 0.0645, 0.0435, 0.0446], device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0641, 0.0807, 0.0659, 0.0495, 0.0488, 0.0514, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:07:39,176 INFO [train.py:904] (0/8) Epoch 9, batch 1550, loss[loss=0.2036, simple_loss=0.2762, pruned_loss=0.0655, over 16512.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2669, pruned_loss=0.05344, over 3322449.03 frames. ], batch size: 75, lr: 7.77e-03, grad_scale: 4.0 2023-04-29 00:07:49,802 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=82759.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:07:58,566 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-29 00:08:12,308 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6971, 3.8767, 2.6309, 2.2924, 2.6503, 2.1176, 3.8050, 3.5865], device='cuda:0'), covar=tensor([0.2465, 0.0617, 0.1689, 0.2250, 0.2313, 0.1947, 0.0561, 0.1123], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0255, 0.0276, 0.0265, 0.0274, 0.0214, 0.0257, 0.0285], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:08:19,726 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4335, 2.0560, 2.2039, 4.0749, 2.1368, 2.6574, 2.1795, 2.3149], device='cuda:0'), covar=tensor([0.0816, 0.2980, 0.1933, 0.0364, 0.2930, 0.1832, 0.2942, 0.2463], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0374, 0.0316, 0.0326, 0.0399, 0.0420, 0.0335, 0.0441], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:08:48,745 INFO [train.py:904] (0/8) Epoch 9, batch 1600, loss[loss=0.1653, simple_loss=0.2527, pruned_loss=0.03897, over 16850.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2685, pruned_loss=0.05416, over 3326749.01 frames. ], batch size: 42, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:09:09,712 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.591e+02 3.276e+02 4.037e+02 8.145e+02, threshold=6.551e+02, percent-clipped=5.0 2023-04-29 00:09:20,065 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9026, 3.3479, 3.1049, 1.9660, 2.6856, 2.3733, 3.3888, 3.4480], device='cuda:0'), covar=tensor([0.0298, 0.0621, 0.0615, 0.1663, 0.0760, 0.0851, 0.0577, 0.0753], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0139, 0.0156, 0.0142, 0.0134, 0.0124, 0.0136, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 00:09:52,317 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82849.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 00:09:56,031 INFO [train.py:904] (0/8) Epoch 9, batch 1650, loss[loss=0.1779, simple_loss=0.2604, pruned_loss=0.04768, over 17248.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2703, pruned_loss=0.0548, over 3326093.57 frames. ], batch size: 45, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:09:56,444 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4022, 5.4068, 5.2019, 4.5541, 5.2150, 2.1491, 4.9688, 5.3179], device='cuda:0'), covar=tensor([0.0066, 0.0057, 0.0121, 0.0360, 0.0072, 0.1990, 0.0104, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0111, 0.0161, 0.0155, 0.0130, 0.0176, 0.0145, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:10:22,001 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3177, 5.6799, 5.3172, 5.4886, 5.0431, 4.9483, 5.0975, 5.7413], device='cuda:0'), covar=tensor([0.1003, 0.0784, 0.1173, 0.0644, 0.0837, 0.0721, 0.0957, 0.0851], device='cuda:0'), in_proj_covar=tensor([0.0514, 0.0658, 0.0548, 0.0450, 0.0410, 0.0423, 0.0542, 0.0487], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:11:00,026 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 00:11:05,717 INFO [train.py:904] (0/8) Epoch 9, batch 1700, loss[loss=0.2144, simple_loss=0.3054, pruned_loss=0.06171, over 16749.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2721, pruned_loss=0.05575, over 3327826.16 frames. ], batch size: 62, lr: 7.77e-03, grad_scale: 8.0 2023-04-29 00:11:24,539 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.504e+02 2.790e+02 3.377e+02 4.170e+02 9.157e+02, threshold=6.754e+02, percent-clipped=3.0 2023-04-29 00:11:37,137 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=82925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:11:53,718 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5023, 2.5415, 2.0789, 2.2660, 2.8752, 2.7359, 3.5093, 3.1819], device='cuda:0'), covar=tensor([0.0066, 0.0292, 0.0355, 0.0337, 0.0182, 0.0237, 0.0146, 0.0165], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0196, 0.0189, 0.0193, 0.0192, 0.0196, 0.0198, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:12:09,324 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6013, 3.8880, 2.8121, 2.2170, 2.6802, 2.1646, 3.7630, 3.5557], device='cuda:0'), covar=tensor([0.2429, 0.0486, 0.1516, 0.2162, 0.2270, 0.1697, 0.0523, 0.1055], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0254, 0.0278, 0.0267, 0.0276, 0.0215, 0.0257, 0.0285], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:12:13,199 INFO [train.py:904] (0/8) Epoch 9, batch 1750, loss[loss=0.1669, simple_loss=0.2496, pruned_loss=0.04213, over 16823.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2722, pruned_loss=0.05579, over 3334983.59 frames. ], batch size: 102, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:12:28,826 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=82963.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:13:19,623 INFO [train.py:904] (0/8) Epoch 9, batch 1800, loss[loss=0.2271, simple_loss=0.3091, pruned_loss=0.07254, over 16518.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2725, pruned_loss=0.05525, over 3340170.61 frames. ], batch size: 68, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:13:19,981 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:13:24,743 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9509, 5.0476, 5.5622, 5.4591, 5.5225, 5.1561, 4.9226, 4.8197], device='cuda:0'), covar=tensor([0.0441, 0.0546, 0.0367, 0.0592, 0.0574, 0.0439, 0.1318, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0322, 0.0326, 0.0307, 0.0370, 0.0345, 0.0446, 0.0275], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 00:13:40,261 INFO [optim.py:368] (0/8) 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 00:13:50,162 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:18,822 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1213, 5.0652, 4.9871, 4.7364, 4.2984, 5.0380, 5.1374, 4.5327], device='cuda:0'), covar=tensor([0.0576, 0.0409, 0.0305, 0.0253, 0.1340, 0.0362, 0.0240, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0290, 0.0284, 0.0254, 0.0307, 0.0290, 0.0196, 0.0326], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:14:26,668 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83050.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:14:28,693 INFO [train.py:904] (0/8) Epoch 9, batch 1850, loss[loss=0.187, simple_loss=0.2769, pruned_loss=0.04851, over 17120.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2738, pruned_loss=0.05564, over 3337299.03 frames. ], batch size: 48, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:11,118 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 00:15:37,693 INFO [train.py:904] (0/8) Epoch 9, batch 1900, loss[loss=0.1915, simple_loss=0.2683, pruned_loss=0.05732, over 16436.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2737, pruned_loss=0.05563, over 3317281.78 frames. ], batch size: 146, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:15:54,384 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2367, 4.1186, 4.2935, 4.1991, 4.1599, 4.7553, 4.3015, 3.9470], device='cuda:0'), covar=tensor([0.1720, 0.1955, 0.1953, 0.2075, 0.3029, 0.1177, 0.1511, 0.2870], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0471, 0.0501, 0.0418, 0.0552, 0.0525, 0.0401, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:15:59,152 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.589e+02 2.476e+02 2.876e+02 3.479e+02 6.930e+02, threshold=5.751e+02, percent-clipped=2.0 2023-04-29 00:16:42,395 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83149.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 00:16:46,828 INFO [train.py:904] (0/8) Epoch 9, batch 1950, loss[loss=0.2028, simple_loss=0.3033, pruned_loss=0.0511, over 17269.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.274, pruned_loss=0.05552, over 3317467.79 frames. ], batch size: 52, lr: 7.76e-03, grad_scale: 8.0 2023-04-29 00:17:48,406 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:17:48,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6379, 2.9554, 2.3904, 4.0605, 3.3484, 3.9765, 1.3924, 2.7339], device='cuda:0'), covar=tensor([0.1274, 0.0477, 0.1038, 0.0143, 0.0183, 0.0363, 0.1381, 0.0779], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0152, 0.0174, 0.0126, 0.0196, 0.0209, 0.0171, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 00:17:49,293 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83197.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 00:17:55,085 INFO [train.py:904] (0/8) Epoch 9, batch 2000, loss[loss=0.2121, simple_loss=0.2953, pruned_loss=0.0644, over 17039.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2735, pruned_loss=0.05516, over 3322634.84 frames. ], batch size: 50, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:18:17,513 INFO [optim.py:368] (0/8) 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 00:18:26,906 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:04,149 INFO [train.py:904] (0/8) Epoch 9, batch 2050, loss[loss=0.2072, simple_loss=0.2792, pruned_loss=0.06764, over 16930.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2736, pruned_loss=0.05577, over 3306069.54 frames. ], batch size: 116, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:19:12,762 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:19:48,952 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:20:13,205 INFO [train.py:904] (0/8) Epoch 9, batch 2100, loss[loss=0.2316, simple_loss=0.3088, pruned_loss=0.07714, over 15586.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2753, pruned_loss=0.0566, over 3295456.70 frames. ], batch size: 191, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:20:20,170 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3436, 1.4956, 2.0491, 2.2716, 2.2927, 2.2675, 1.6551, 2.4822], device='cuda:0'), covar=tensor([0.0129, 0.0271, 0.0166, 0.0151, 0.0156, 0.0161, 0.0263, 0.0055], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0165, 0.0150, 0.0153, 0.0158, 0.0115, 0.0164, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 00:20:35,022 INFO [optim.py:368] (0/8) 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 00:20:37,095 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:11,837 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83345.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:21:20,744 INFO [train.py:904] (0/8) Epoch 9, batch 2150, loss[loss=0.1908, simple_loss=0.2766, pruned_loss=0.05248, over 17226.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2757, pruned_loss=0.05675, over 3309005.72 frames. ], batch size: 46, lr: 7.75e-03, grad_scale: 8.0 2023-04-29 00:21:48,437 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4317, 1.9694, 2.1903, 4.0700, 1.9912, 2.4482, 2.0868, 2.1963], device='cuda:0'), covar=tensor([0.0909, 0.3255, 0.1814, 0.0378, 0.3485, 0.2045, 0.2920, 0.2757], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0376, 0.0315, 0.0324, 0.0402, 0.0426, 0.0335, 0.0443], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:22:31,378 INFO [train.py:904] (0/8) Epoch 9, batch 2200, loss[loss=0.2067, simple_loss=0.2937, pruned_loss=0.05985, over 16701.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2778, pruned_loss=0.0589, over 3300729.50 frames. ], batch size: 57, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:22:54,055 INFO [optim.py:368] (0/8) 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 00:23:20,485 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.66 vs. limit=5.0 2023-04-29 00:23:21,222 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8095, 5.0174, 5.1737, 5.0211, 5.0712, 5.6430, 5.2211, 4.9268], device='cuda:0'), covar=tensor([0.1106, 0.1807, 0.2027, 0.2007, 0.2617, 0.1008, 0.1467, 0.2520], device='cuda:0'), in_proj_covar=tensor([0.0339, 0.0473, 0.0502, 0.0421, 0.0554, 0.0528, 0.0403, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:23:41,055 INFO [train.py:904] (0/8) Epoch 9, batch 2250, loss[loss=0.1853, simple_loss=0.2778, pruned_loss=0.04643, over 17040.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2769, pruned_loss=0.05801, over 3305581.77 frames. ], batch size: 50, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:24:16,114 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3311, 5.8315, 6.0303, 5.7329, 5.8253, 6.2403, 5.9067, 5.6072], device='cuda:0'), covar=tensor([0.0615, 0.1398, 0.1238, 0.1592, 0.2314, 0.0844, 0.1010, 0.1895], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0474, 0.0502, 0.0420, 0.0555, 0.0530, 0.0402, 0.0560], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:24:24,852 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1976, 4.4184, 4.5892, 2.3206, 4.8781, 4.8831, 3.5729, 3.9684], device='cuda:0'), covar=tensor([0.0662, 0.0140, 0.0186, 0.1086, 0.0056, 0.0100, 0.0305, 0.0276], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0098, 0.0087, 0.0143, 0.0070, 0.0099, 0.0121, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 00:24:49,201 INFO [train.py:904] (0/8) Epoch 9, batch 2300, loss[loss=0.2264, simple_loss=0.2872, pruned_loss=0.08281, over 16729.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2765, pruned_loss=0.05826, over 3317741.00 frames. ], batch size: 134, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:12,023 INFO [optim.py:368] (0/8) 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 00:25:52,553 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9788, 2.2554, 2.4105, 4.8901, 2.1771, 2.9611, 2.4431, 2.6019], device='cuda:0'), covar=tensor([0.0739, 0.3237, 0.1970, 0.0292, 0.3601, 0.1987, 0.2660, 0.3224], device='cuda:0'), in_proj_covar=tensor([0.0353, 0.0376, 0.0316, 0.0324, 0.0400, 0.0424, 0.0334, 0.0443], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:25:59,040 INFO [train.py:904] (0/8) Epoch 9, batch 2350, loss[loss=0.2222, simple_loss=0.2873, pruned_loss=0.07854, over 16390.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2772, pruned_loss=0.05849, over 3319573.22 frames. ], batch size: 146, lr: 7.74e-03, grad_scale: 8.0 2023-04-29 00:25:59,369 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83552.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:27:06,422 INFO [train.py:904] (0/8) Epoch 9, batch 2400, loss[loss=0.2203, simple_loss=0.2901, pruned_loss=0.07524, over 16887.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2777, pruned_loss=0.0583, over 3315574.29 frames. ], batch size: 116, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:27:21,196 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.03 vs. limit=2.0 2023-04-29 00:27:24,808 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-29 00:27:29,673 INFO [optim.py:368] (0/8) 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 00:27:31,379 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83619.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:27:59,102 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=83640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:28:14,971 INFO [train.py:904] (0/8) Epoch 9, batch 2450, loss[loss=0.2097, simple_loss=0.308, pruned_loss=0.05573, over 17251.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2794, pruned_loss=0.05815, over 3313874.67 frames. ], batch size: 52, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:28:16,662 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8793, 3.5656, 2.7999, 5.2113, 4.4586, 4.8067, 1.6613, 3.4283], device='cuda:0'), covar=tensor([0.1224, 0.0527, 0.1003, 0.0110, 0.0209, 0.0303, 0.1359, 0.0635], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0153, 0.0174, 0.0128, 0.0199, 0.0210, 0.0172, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 00:28:22,594 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3050, 4.2867, 4.2364, 3.5721, 4.2825, 1.5984, 4.0058, 3.9647], device='cuda:0'), covar=tensor([0.0102, 0.0085, 0.0153, 0.0358, 0.0084, 0.2550, 0.0128, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0111, 0.0162, 0.0154, 0.0130, 0.0176, 0.0146, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:28:25,674 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1200, 3.4481, 3.2854, 1.9908, 2.8309, 2.3813, 3.6327, 3.5462], device='cuda:0'), covar=tensor([0.0223, 0.0652, 0.0519, 0.1584, 0.0715, 0.0833, 0.0446, 0.0677], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0140, 0.0156, 0.0140, 0.0134, 0.0123, 0.0135, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 00:28:34,506 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83667.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:29:22,513 INFO [train.py:904] (0/8) Epoch 9, batch 2500, loss[loss=0.1935, simple_loss=0.2786, pruned_loss=0.05422, over 16554.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2787, pruned_loss=0.05787, over 3312900.65 frames. ], batch size: 68, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:29:43,292 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4858, 2.4698, 2.0168, 2.1603, 2.7884, 2.5973, 3.3934, 3.0738], device='cuda:0'), covar=tensor([0.0080, 0.0266, 0.0306, 0.0322, 0.0164, 0.0240, 0.0141, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0191, 0.0185, 0.0190, 0.0189, 0.0193, 0.0197, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:29:44,588 INFO [optim.py:368] (0/8) 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 00:30:28,784 INFO [train.py:904] (0/8) Epoch 9, batch 2550, loss[loss=0.2272, simple_loss=0.3032, pruned_loss=0.07564, over 16497.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2781, pruned_loss=0.05736, over 3321803.92 frames. ], batch size: 146, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:30:34,590 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4234, 4.4006, 4.4401, 3.9306, 4.4376, 1.8152, 4.1534, 4.1579], device='cuda:0'), covar=tensor([0.0090, 0.0079, 0.0117, 0.0281, 0.0077, 0.2090, 0.0119, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0110, 0.0161, 0.0154, 0.0129, 0.0175, 0.0146, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:30:41,832 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83760.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:31:04,985 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0122, 1.9059, 2.4349, 2.8548, 2.7990, 3.3007, 2.2724, 3.2642], device='cuda:0'), covar=tensor([0.0140, 0.0317, 0.0202, 0.0186, 0.0166, 0.0117, 0.0269, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0165, 0.0150, 0.0153, 0.0157, 0.0114, 0.0163, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 00:31:37,331 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7171, 3.8457, 3.0507, 2.1899, 2.5233, 2.1718, 3.8143, 3.3750], device='cuda:0'), covar=tensor([0.2133, 0.0548, 0.1291, 0.2230, 0.2329, 0.1766, 0.0471, 0.1239], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0256, 0.0278, 0.0268, 0.0281, 0.0214, 0.0262, 0.0289], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:31:38,618 INFO [train.py:904] (0/8) Epoch 9, batch 2600, loss[loss=0.1799, simple_loss=0.2587, pruned_loss=0.05061, over 16831.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2772, pruned_loss=0.0565, over 3325968.33 frames. ], batch size: 102, lr: 7.73e-03, grad_scale: 8.0 2023-04-29 00:31:42,369 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8575, 5.0284, 5.2391, 5.0026, 5.0303, 5.6685, 5.1464, 4.8731], device='cuda:0'), covar=tensor([0.1085, 0.1734, 0.1791, 0.1815, 0.2621, 0.1134, 0.1315, 0.2187], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0476, 0.0505, 0.0418, 0.0556, 0.0531, 0.0402, 0.0560], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:31:59,383 INFO [optim.py:368] (0/8) 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 00:32:04,267 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=83821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:32:10,164 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 00:32:45,554 INFO [train.py:904] (0/8) Epoch 9, batch 2650, loss[loss=0.2247, simple_loss=0.3013, pruned_loss=0.07403, over 12642.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2777, pruned_loss=0.05565, over 3327846.97 frames. ], batch size: 247, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:32:45,974 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:33:36,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7148, 2.8424, 2.3834, 3.9865, 3.3910, 4.0104, 1.4974, 2.8126], device='cuda:0'), covar=tensor([0.1201, 0.0529, 0.0973, 0.0129, 0.0188, 0.0344, 0.1271, 0.0710], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0154, 0.0175, 0.0129, 0.0200, 0.0212, 0.0173, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 00:33:45,971 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6786, 4.5945, 4.5360, 4.3488, 4.1555, 4.6227, 4.4825, 4.3490], device='cuda:0'), covar=tensor([0.0641, 0.0530, 0.0323, 0.0257, 0.0934, 0.0488, 0.0409, 0.0661], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0297, 0.0288, 0.0260, 0.0313, 0.0294, 0.0196, 0.0331], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:33:51,887 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83900.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:33:53,993 INFO [train.py:904] (0/8) Epoch 9, batch 2700, loss[loss=0.2044, simple_loss=0.278, pruned_loss=0.06541, over 16713.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2776, pruned_loss=0.05526, over 3337374.02 frames. ], batch size: 134, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:34:17,452 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.501e+02 2.856e+02 3.411e+02 7.585e+02, threshold=5.711e+02, percent-clipped=1.0 2023-04-29 00:34:47,085 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=83940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:34:58,182 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:02,462 INFO [train.py:904] (0/8) Epoch 9, batch 2750, loss[loss=0.1807, simple_loss=0.2692, pruned_loss=0.04612, over 17230.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2778, pruned_loss=0.05475, over 3341940.68 frames. ], batch size: 45, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:35:48,340 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=83985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:35:51,709 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=83988.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:36:08,870 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-84000.pt 2023-04-29 00:36:15,275 INFO [train.py:904] (0/8) Epoch 9, batch 2800, loss[loss=0.1797, simple_loss=0.2758, pruned_loss=0.04182, over 17069.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2782, pruned_loss=0.05489, over 3343538.37 frames. ], batch size: 53, lr: 7.72e-03, grad_scale: 8.0 2023-04-29 00:36:25,570 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84010.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 00:36:36,300 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.751e+02 3.290e+02 3.928e+02 7.223e+02, threshold=6.581e+02, percent-clipped=1.0 2023-04-29 00:37:16,000 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:37:23,284 INFO [train.py:904] (0/8) Epoch 9, batch 2850, loss[loss=0.2332, simple_loss=0.3067, pruned_loss=0.07984, over 16511.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2778, pruned_loss=0.05541, over 3333657.78 frames. ], batch size: 68, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:37:56,046 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7245, 4.2194, 4.3367, 3.0545, 3.8536, 4.3260, 3.8847, 2.7325], device='cuda:0'), covar=tensor([0.0373, 0.0038, 0.0029, 0.0242, 0.0063, 0.0057, 0.0055, 0.0284], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0069, 0.0067, 0.0123, 0.0074, 0.0084, 0.0075, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:38:25,961 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:32,600 INFO [train.py:904] (0/8) Epoch 9, batch 2900, loss[loss=0.279, simple_loss=0.3165, pruned_loss=0.1208, over 11754.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2776, pruned_loss=0.0561, over 3328665.30 frames. ], batch size: 246, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:38:52,450 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:38:54,445 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.515e+02 3.057e+02 3.611e+02 6.139e+02, threshold=6.114e+02, percent-clipped=0.0 2023-04-29 00:39:30,601 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 00:39:43,185 INFO [train.py:904] (0/8) Epoch 9, batch 2950, loss[loss=0.2013, simple_loss=0.2725, pruned_loss=0.06505, over 16701.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2771, pruned_loss=0.05698, over 3320262.60 frames. ], batch size: 134, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:39:51,144 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:21,136 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-29 00:40:48,002 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:40:50,397 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3322, 2.0143, 2.2127, 4.0326, 2.0807, 2.5842, 2.1101, 2.2833], device='cuda:0'), covar=tensor([0.0883, 0.3092, 0.1857, 0.0410, 0.2969, 0.1881, 0.2831, 0.2476], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0378, 0.0316, 0.0326, 0.0400, 0.0428, 0.0337, 0.0447], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:40:52,822 INFO [train.py:904] (0/8) Epoch 9, batch 3000, loss[loss=0.1881, simple_loss=0.2681, pruned_loss=0.0541, over 16417.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2758, pruned_loss=0.05629, over 3323856.49 frames. ], batch size: 146, lr: 7.71e-03, grad_scale: 8.0 2023-04-29 00:40:52,823 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 00:41:02,064 INFO [train.py:938] (0/8) Epoch 9, validation: loss=0.1444, simple_loss=0.2507, pruned_loss=0.019, over 944034.00 frames. 2023-04-29 00:41:02,064 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 00:41:23,166 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.086e+02 2.665e+02 3.010e+02 4.048e+02 8.692e+02, threshold=6.019e+02, percent-clipped=1.0 2023-04-29 00:41:33,545 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3182, 5.2224, 5.1956, 4.8542, 4.7633, 5.1594, 5.1729, 4.7694], device='cuda:0'), covar=tensor([0.0489, 0.0355, 0.0210, 0.0214, 0.0913, 0.0315, 0.0246, 0.0619], device='cuda:0'), in_proj_covar=tensor([0.0246, 0.0297, 0.0285, 0.0260, 0.0312, 0.0296, 0.0196, 0.0331], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:41:47,898 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:09,947 INFO [train.py:904] (0/8) Epoch 9, batch 3050, loss[loss=0.1994, simple_loss=0.2855, pruned_loss=0.05664, over 16713.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2756, pruned_loss=0.05601, over 3332680.33 frames. ], batch size: 62, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:42:20,952 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:42:41,854 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2729, 4.2831, 4.7182, 4.6713, 4.7142, 4.3492, 4.4027, 4.2261], device='cuda:0'), covar=tensor([0.0303, 0.0499, 0.0313, 0.0398, 0.0410, 0.0329, 0.0788, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0326, 0.0329, 0.0314, 0.0372, 0.0347, 0.0458, 0.0279], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 00:42:45,158 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4375, 5.7743, 5.5221, 5.5963, 5.1968, 4.9094, 5.2270, 5.9355], device='cuda:0'), covar=tensor([0.1140, 0.0913, 0.1009, 0.0603, 0.0786, 0.0622, 0.0875, 0.0777], device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0664, 0.0554, 0.0451, 0.0409, 0.0421, 0.0546, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:43:11,195 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84297.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:43:16,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5755, 4.7498, 4.5358, 4.4290, 3.6947, 4.7323, 4.7127, 4.3017], device='cuda:0'), covar=tensor([0.1038, 0.0880, 0.0510, 0.0387, 0.1969, 0.0561, 0.0473, 0.0781], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0299, 0.0288, 0.0262, 0.0315, 0.0297, 0.0198, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:43:17,171 INFO [train.py:904] (0/8) Epoch 9, batch 3100, loss[loss=0.1949, simple_loss=0.2701, pruned_loss=0.05984, over 16511.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2757, pruned_loss=0.05637, over 3335663.94 frames. ], batch size: 75, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:43:22,116 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84305.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 00:43:39,991 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.846e+02 2.519e+02 3.109e+02 3.888e+02 1.112e+03, threshold=6.217e+02, percent-clipped=5.0 2023-04-29 00:43:40,516 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4016, 2.0427, 2.8639, 3.0710, 3.0378, 3.5383, 2.4413, 3.4622], device='cuda:0'), covar=tensor([0.0100, 0.0287, 0.0183, 0.0154, 0.0150, 0.0116, 0.0282, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0165, 0.0152, 0.0155, 0.0157, 0.0115, 0.0164, 0.0106], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 00:44:06,186 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:12,253 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84341.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:44:28,379 INFO [train.py:904] (0/8) Epoch 9, batch 3150, loss[loss=0.1896, simple_loss=0.2621, pruned_loss=0.05854, over 16895.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2742, pruned_loss=0.05597, over 3334389.79 frames. ], batch size: 96, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:44:34,352 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 00:45:30,270 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:36,488 INFO [train.py:904] (0/8) Epoch 9, batch 3200, loss[loss=0.205, simple_loss=0.294, pruned_loss=0.05805, over 17012.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2734, pruned_loss=0.05494, over 3341185.21 frames. ], batch size: 55, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:45:56,471 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:45:59,093 INFO [optim.py:368] (0/8) 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 00:46:00,559 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4247, 2.9911, 2.6075, 2.2874, 2.2923, 2.1602, 2.9930, 2.9618], device='cuda:0'), covar=tensor([0.2068, 0.0671, 0.1250, 0.1566, 0.1785, 0.1555, 0.0460, 0.0828], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0256, 0.0279, 0.0268, 0.0282, 0.0215, 0.0262, 0.0292], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:46:45,891 INFO [train.py:904] (0/8) Epoch 9, batch 3250, loss[loss=0.1838, simple_loss=0.2731, pruned_loss=0.04725, over 17104.00 frames. ], tot_loss[loss=0.1915, simple_loss=0.2735, pruned_loss=0.05479, over 3339483.66 frames. ], batch size: 53, lr: 7.70e-03, grad_scale: 8.0 2023-04-29 00:46:46,756 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-29 00:46:47,431 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84453.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:02,633 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84464.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:47:55,754 INFO [train.py:904] (0/8) Epoch 9, batch 3300, loss[loss=0.1677, simple_loss=0.2561, pruned_loss=0.03965, over 17212.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2748, pruned_loss=0.05547, over 3338542.02 frames. ], batch size: 45, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:47:56,155 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4074, 5.3865, 5.1354, 4.5887, 5.1669, 1.9500, 4.9219, 5.2271], device='cuda:0'), covar=tensor([0.0057, 0.0056, 0.0131, 0.0330, 0.0074, 0.2020, 0.0098, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0113, 0.0164, 0.0156, 0.0132, 0.0176, 0.0148, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:48:18,093 INFO [optim.py:368] (0/8) 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 00:48:27,887 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:48:56,111 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7622, 1.7886, 2.3202, 2.7616, 2.6234, 3.1314, 1.9767, 3.1628], device='cuda:0'), covar=tensor([0.0144, 0.0317, 0.0220, 0.0203, 0.0186, 0.0110, 0.0296, 0.0081], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0165, 0.0152, 0.0155, 0.0159, 0.0116, 0.0164, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 00:49:04,554 INFO [train.py:904] (0/8) Epoch 9, batch 3350, loss[loss=0.1939, simple_loss=0.2758, pruned_loss=0.05595, over 16266.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2762, pruned_loss=0.05607, over 3335772.13 frames. ], batch size: 165, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:49:10,257 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:49:10,424 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5478, 3.7184, 4.0097, 2.0979, 4.1755, 4.2131, 3.0975, 3.0339], device='cuda:0'), covar=tensor([0.0778, 0.0170, 0.0138, 0.1107, 0.0065, 0.0116, 0.0422, 0.0414], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0098, 0.0088, 0.0141, 0.0070, 0.0101, 0.0122, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 00:49:51,999 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7595, 3.9544, 2.2643, 4.4081, 2.8172, 4.4832, 2.3274, 3.1984], device='cuda:0'), covar=tensor([0.0236, 0.0329, 0.1443, 0.0164, 0.0775, 0.0328, 0.1375, 0.0571], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0164, 0.0183, 0.0121, 0.0165, 0.0207, 0.0190, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 00:49:53,265 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:01,517 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:16,096 INFO [train.py:904] (0/8) Epoch 9, batch 3400, loss[loss=0.2071, simple_loss=0.2904, pruned_loss=0.06192, over 12164.00 frames. ], tot_loss[loss=0.1927, simple_loss=0.2755, pruned_loss=0.05498, over 3328952.27 frames. ], batch size: 248, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:50:20,181 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:50:38,681 INFO [optim.py:368] (0/8) 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 00:50:59,358 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84633.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:10,829 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:51:24,590 INFO [train.py:904] (0/8) Epoch 9, batch 3450, loss[loss=0.1864, simple_loss=0.2673, pruned_loss=0.05275, over 16549.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2746, pruned_loss=0.0541, over 3329112.26 frames. ], batch size: 68, lr: 7.69e-03, grad_scale: 8.0 2023-04-29 00:51:26,680 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:16,071 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84689.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:20,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:23,592 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84694.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:52:35,107 INFO [train.py:904] (0/8) Epoch 9, batch 3500, loss[loss=0.193, simple_loss=0.2831, pruned_loss=0.05138, over 16649.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2736, pruned_loss=0.05409, over 3326758.50 frames. ], batch size: 57, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:52:56,978 INFO [optim.py:368] (0/8) 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 00:53:44,892 INFO [train.py:904] (0/8) Epoch 9, batch 3550, loss[loss=0.1808, simple_loss=0.2712, pruned_loss=0.04522, over 17184.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.272, pruned_loss=0.05388, over 3326973.43 frames. ], batch size: 46, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:53:46,580 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:48,659 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 00:54:53,027 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84800.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:54,909 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84801.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:54:55,847 INFO [train.py:904] (0/8) Epoch 9, batch 3600, loss[loss=0.1972, simple_loss=0.2727, pruned_loss=0.06081, over 16713.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2711, pruned_loss=0.05364, over 3316529.92 frames. ], batch size: 134, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:55:14,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3901, 4.0609, 3.8858, 2.1180, 3.2539, 2.8442, 3.7732, 3.9842], device='cuda:0'), covar=tensor([0.0276, 0.0612, 0.0486, 0.1550, 0.0696, 0.0759, 0.0641, 0.0802], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0145, 0.0157, 0.0142, 0.0136, 0.0125, 0.0137, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 00:55:17,925 INFO [optim.py:368] (0/8) 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 00:55:43,627 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1251, 2.1603, 2.6826, 3.0077, 2.9171, 3.4781, 2.4672, 3.4572], device='cuda:0'), covar=tensor([0.0128, 0.0272, 0.0160, 0.0157, 0.0154, 0.0105, 0.0242, 0.0083], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0166, 0.0151, 0.0154, 0.0159, 0.0116, 0.0164, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 00:55:53,035 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-29 00:56:07,149 INFO [train.py:904] (0/8) Epoch 9, batch 3650, loss[loss=0.1918, simple_loss=0.2661, pruned_loss=0.05874, over 16691.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2692, pruned_loss=0.05389, over 3308410.85 frames. ], batch size: 134, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:56:12,755 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84855.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:21,101 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=84861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:44,699 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8841, 2.0318, 2.2799, 3.1351, 2.1171, 2.2725, 2.3028, 2.1552], device='cuda:0'), covar=tensor([0.0855, 0.2591, 0.1578, 0.0545, 0.3208, 0.1821, 0.2104, 0.2985], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0382, 0.0320, 0.0332, 0.0406, 0.0433, 0.0342, 0.0450], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 00:56:50,506 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:56:54,746 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8088, 1.7483, 2.3538, 2.7663, 2.7415, 2.4310, 1.8037, 2.8996], device='cuda:0'), covar=tensor([0.0105, 0.0304, 0.0214, 0.0156, 0.0154, 0.0177, 0.0308, 0.0089], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0167, 0.0153, 0.0155, 0.0161, 0.0117, 0.0165, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 00:57:07,351 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:21,445 INFO [train.py:904] (0/8) Epoch 9, batch 3700, loss[loss=0.1932, simple_loss=0.2606, pruned_loss=0.06293, over 16537.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2683, pruned_loss=0.05585, over 3292726.98 frames. ], batch size: 75, lr: 7.68e-03, grad_scale: 8.0 2023-04-29 00:57:23,508 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84903.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:57:46,410 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.577e+02 2.579e+02 2.947e+02 3.773e+02 6.988e+02, threshold=5.894e+02, percent-clipped=1.0 2023-04-29 00:58:18,418 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=84940.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:24,038 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7398, 4.9504, 5.0945, 4.9743, 4.9378, 5.5363, 5.0504, 4.7401], device='cuda:0'), covar=tensor([0.1180, 0.1613, 0.1426, 0.1728, 0.2246, 0.0886, 0.1334, 0.2333], device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0469, 0.0498, 0.0418, 0.0547, 0.0520, 0.0399, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 00:58:34,890 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=84951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:58:35,661 INFO [train.py:904] (0/8) Epoch 9, batch 3750, loss[loss=0.198, simple_loss=0.2665, pruned_loss=0.06475, over 16274.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2689, pruned_loss=0.05725, over 3277330.26 frames. ], batch size: 165, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 00:59:28,360 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=84989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:34,011 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=84992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 00:59:47,563 INFO [train.py:904] (0/8) Epoch 9, batch 3800, loss[loss=0.2056, simple_loss=0.2753, pruned_loss=0.06797, over 16879.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2702, pruned_loss=0.05907, over 3282605.01 frames. ], batch size: 96, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:00:03,373 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85012.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:00:11,018 INFO [optim.py:368] (0/8) 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 01:00:35,392 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5161, 2.0866, 2.2771, 4.2954, 2.0789, 2.5837, 2.1440, 2.3793], device='cuda:0'), covar=tensor([0.0921, 0.3159, 0.1920, 0.0339, 0.3304, 0.1978, 0.3104, 0.2393], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0385, 0.0322, 0.0330, 0.0404, 0.0435, 0.0343, 0.0452], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:00:43,936 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:01:00,361 INFO [train.py:904] (0/8) Epoch 9, batch 3850, loss[loss=0.186, simple_loss=0.257, pruned_loss=0.05748, over 16914.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2699, pruned_loss=0.05948, over 3283600.02 frames. ], batch size: 109, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:01:10,233 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7333, 4.7947, 4.8883, 4.8735, 4.7813, 5.3947, 5.0107, 4.6648], device='cuda:0'), covar=tensor([0.1273, 0.1633, 0.1552, 0.1889, 0.2654, 0.0968, 0.1271, 0.2297], device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0463, 0.0495, 0.0412, 0.0540, 0.0514, 0.0395, 0.0550], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 01:01:59,629 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3457, 2.2433, 1.7216, 1.9602, 2.5937, 2.3739, 2.7014, 2.7432], device='cuda:0'), covar=tensor([0.0104, 0.0212, 0.0302, 0.0281, 0.0113, 0.0189, 0.0134, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0189, 0.0186, 0.0185, 0.0185, 0.0190, 0.0194, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:02:10,453 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7461, 4.8151, 4.9481, 4.8791, 4.8962, 5.4564, 5.0618, 4.7343], device='cuda:0'), covar=tensor([0.1403, 0.1821, 0.1703, 0.1769, 0.2457, 0.0883, 0.1435, 0.2360], device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0468, 0.0502, 0.0416, 0.0545, 0.0520, 0.0401, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 01:02:13,004 INFO [train.py:904] (0/8) Epoch 9, batch 3900, loss[loss=0.2101, simple_loss=0.2801, pruned_loss=0.07006, over 16421.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.2697, pruned_loss=0.06046, over 3287758.33 frames. ], batch size: 146, lr: 7.67e-03, grad_scale: 8.0 2023-04-29 01:02:21,725 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7158, 2.7830, 1.8767, 2.7583, 2.2383, 2.8575, 1.9672, 2.4019], device='cuda:0'), covar=tensor([0.0205, 0.0275, 0.1240, 0.0158, 0.0613, 0.0287, 0.1198, 0.0508], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0166, 0.0185, 0.0120, 0.0166, 0.0209, 0.0193, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 01:02:21,729 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:29,572 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7096, 5.0180, 4.7748, 4.8367, 4.4908, 4.4454, 4.5002, 5.0619], device='cuda:0'), covar=tensor([0.0889, 0.0794, 0.0963, 0.0608, 0.0709, 0.0953, 0.0905, 0.0782], device='cuda:0'), in_proj_covar=tensor([0.0508, 0.0649, 0.0533, 0.0443, 0.0405, 0.0413, 0.0537, 0.0482], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:02:32,598 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:02:36,237 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.478e+02 2.844e+02 3.462e+02 6.399e+02, threshold=5.687e+02, percent-clipped=3.0 2023-04-29 01:03:24,370 INFO [train.py:904] (0/8) Epoch 9, batch 3950, loss[loss=0.2178, simple_loss=0.2891, pruned_loss=0.07325, over 15567.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2694, pruned_loss=0.06087, over 3272261.43 frames. ], batch size: 191, lr: 7.66e-03, grad_scale: 8.0 2023-04-29 01:03:32,144 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85156.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:03:32,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8153, 4.7714, 4.7383, 4.5240, 4.4096, 4.7880, 4.6013, 4.5533], device='cuda:0'), covar=tensor([0.0614, 0.0509, 0.0252, 0.0260, 0.0828, 0.0377, 0.0381, 0.0610], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0292, 0.0285, 0.0257, 0.0308, 0.0294, 0.0193, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 01:03:41,839 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7398, 3.7536, 4.0871, 2.0877, 4.2558, 4.2599, 3.2876, 3.2042], device='cuda:0'), covar=tensor([0.0653, 0.0166, 0.0125, 0.1060, 0.0046, 0.0106, 0.0306, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0097, 0.0085, 0.0138, 0.0069, 0.0099, 0.0120, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 01:03:50,068 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85168.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:04:00,708 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:07,282 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85181.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:04:37,119 INFO [train.py:904] (0/8) Epoch 9, batch 4000, loss[loss=0.1936, simple_loss=0.2787, pruned_loss=0.05422, over 15502.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2695, pruned_loss=0.0612, over 3272725.27 frames. ], batch size: 191, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:05:00,775 INFO [optim.py:368] (0/8) 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 01:05:09,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3864, 5.7421, 5.4409, 5.5504, 5.0362, 4.8724, 5.1698, 5.8086], device='cuda:0'), covar=tensor([0.1022, 0.0771, 0.0895, 0.0632, 0.0838, 0.0647, 0.0775, 0.0715], device='cuda:0'), in_proj_covar=tensor([0.0510, 0.0650, 0.0535, 0.0442, 0.0405, 0.0411, 0.0539, 0.0482], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:05:17,724 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:05:46,423 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5316, 3.7433, 2.8713, 2.2961, 2.5915, 2.2703, 3.8665, 3.5479], device='cuda:0'), covar=tensor([0.2609, 0.0684, 0.1620, 0.2021, 0.2400, 0.1720, 0.0528, 0.0826], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0257, 0.0281, 0.0273, 0.0287, 0.0218, 0.0265, 0.0295], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 01:05:51,200 INFO [train.py:904] (0/8) Epoch 9, batch 4050, loss[loss=0.1927, simple_loss=0.2775, pruned_loss=0.05399, over 16225.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2694, pruned_loss=0.05992, over 3280548.03 frames. ], batch size: 35, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:05:58,191 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6445, 2.7111, 1.8415, 2.7035, 2.1520, 2.7930, 1.9433, 2.3203], device='cuda:0'), covar=tensor([0.0223, 0.0330, 0.1261, 0.0145, 0.0617, 0.0311, 0.1180, 0.0561], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0166, 0.0186, 0.0120, 0.0167, 0.0208, 0.0192, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 01:06:22,119 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 01:06:45,618 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 9, batch 4100, loss[loss=0.2278, simple_loss=0.3123, pruned_loss=0.07159, over 16626.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.271, pruned_loss=0.0589, over 3285434.54 frames. ], batch size: 134, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:07:12,418 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:07:28,045 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.051e+02 2.369e+02 2.790e+02 6.834e+02, threshold=4.737e+02, percent-clipped=1.0 2023-04-29 01:07:57,467 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:08:20,251 INFO [train.py:904] (0/8) Epoch 9, batch 4150, loss[loss=0.2824, simple_loss=0.3402, pruned_loss=0.1123, over 11512.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2794, pruned_loss=0.06266, over 3239211.96 frames. ], batch size: 247, lr: 7.66e-03, grad_scale: 16.0 2023-04-29 01:08:55,686 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 01:09:37,257 INFO [train.py:904] (0/8) Epoch 9, batch 4200, loss[loss=0.2432, simple_loss=0.3111, pruned_loss=0.08764, over 11717.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2863, pruned_loss=0.0643, over 3221061.71 frames. ], batch size: 246, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:02,498 INFO [optim.py:368] (0/8) 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 01:10:17,685 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:10:51,171 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3272, 5.1470, 5.2283, 5.4635, 5.6246, 4.9477, 5.6320, 5.6301], device='cuda:0'), covar=tensor([0.1172, 0.0857, 0.1458, 0.0614, 0.0507, 0.0583, 0.0451, 0.0491], device='cuda:0'), in_proj_covar=tensor([0.0496, 0.0601, 0.0746, 0.0610, 0.0466, 0.0470, 0.0481, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:10:52,281 INFO [train.py:904] (0/8) Epoch 9, batch 4250, loss[loss=0.2222, simple_loss=0.3047, pruned_loss=0.06981, over 16358.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.289, pruned_loss=0.06339, over 3215631.70 frames. ], batch size: 35, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:10:59,132 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:08,341 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85463.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:11:11,305 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6435, 4.4002, 4.0986, 4.6721, 4.8188, 4.4759, 4.7528, 4.9116], device='cuda:0'), covar=tensor([0.1013, 0.1038, 0.2419, 0.1131, 0.0880, 0.1033, 0.1055, 0.0858], device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0598, 0.0742, 0.0608, 0.0463, 0.0468, 0.0479, 0.0534], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:11:20,278 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:39,062 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:11:48,291 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:07,386 INFO [train.py:904] (0/8) Epoch 9, batch 4300, loss[loss=0.2062, simple_loss=0.2954, pruned_loss=0.05847, over 16684.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2898, pruned_loss=0.06239, over 3203991.81 frames. ], batch size: 134, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:12:11,408 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85504.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:12:30,790 INFO [optim.py:368] (0/8) 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 01:12:49,221 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85530.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:10,275 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:13:12,762 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3950, 4.2191, 4.4669, 4.5940, 4.7189, 4.3213, 4.6644, 4.7373], device='cuda:0'), covar=tensor([0.1177, 0.0895, 0.1066, 0.0475, 0.0402, 0.0844, 0.0492, 0.0433], device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0600, 0.0742, 0.0607, 0.0465, 0.0469, 0.0481, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:13:21,363 INFO [train.py:904] (0/8) Epoch 9, batch 4350, loss[loss=0.235, simple_loss=0.3023, pruned_loss=0.08388, over 11748.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2933, pruned_loss=0.06381, over 3178530.76 frames. ], batch size: 248, lr: 7.65e-03, grad_scale: 16.0 2023-04-29 01:14:18,959 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:34,744 INFO [train.py:904] (0/8) Epoch 9, batch 4400, loss[loss=0.2027, simple_loss=0.2906, pruned_loss=0.05742, over 16933.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2958, pruned_loss=0.0648, over 3200017.97 frames. ], batch size: 41, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:14:41,661 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85607.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:14:56,973 INFO [optim.py:368] (0/8) 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 01:15:06,898 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:15:46,028 INFO [train.py:904] (0/8) Epoch 9, batch 4450, loss[loss=0.2128, simple_loss=0.3045, pruned_loss=0.06056, over 16919.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2987, pruned_loss=0.06579, over 3197853.89 frames. ], batch size: 96, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:15:50,416 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85655.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:16:19,396 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8415, 3.9477, 4.2914, 1.9415, 4.7528, 4.7486, 3.3205, 3.5771], device='cuda:0'), covar=tensor([0.0729, 0.0175, 0.0144, 0.1120, 0.0029, 0.0046, 0.0313, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0096, 0.0084, 0.0138, 0.0067, 0.0096, 0.0118, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 01:16:28,247 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4061, 2.4096, 1.9540, 2.1999, 2.8005, 2.4449, 3.1811, 3.0281], device='cuda:0'), covar=tensor([0.0046, 0.0254, 0.0346, 0.0300, 0.0154, 0.0256, 0.0118, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0190, 0.0188, 0.0186, 0.0186, 0.0192, 0.0191, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:16:33,298 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85686.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:16:56,349 INFO [train.py:904] (0/8) Epoch 9, batch 4500, loss[loss=0.2037, simple_loss=0.2922, pruned_loss=0.05758, over 16248.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2986, pruned_loss=0.06589, over 3208584.49 frames. ], batch size: 165, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:17:04,839 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7725, 3.7608, 3.0790, 2.2949, 2.8421, 2.4555, 4.2058, 3.7610], device='cuda:0'), covar=tensor([0.2490, 0.0701, 0.1469, 0.1961, 0.2050, 0.1642, 0.0433, 0.0755], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0255, 0.0280, 0.0273, 0.0285, 0.0216, 0.0264, 0.0287], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:17:10,614 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0993, 3.2464, 3.4327, 1.4942, 3.7466, 3.8257, 2.8667, 2.7841], device='cuda:0'), covar=tensor([0.0910, 0.0229, 0.0211, 0.1342, 0.0059, 0.0078, 0.0382, 0.0451], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0095, 0.0084, 0.0137, 0.0067, 0.0095, 0.0117, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 01:17:20,318 INFO [optim.py:368] (0/8) 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 01:17:32,425 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6840, 1.3934, 2.1848, 2.6247, 2.5384, 2.8285, 1.4301, 2.7468], device='cuda:0'), covar=tensor([0.0105, 0.0350, 0.0168, 0.0141, 0.0138, 0.0085, 0.0402, 0.0077], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0162, 0.0148, 0.0151, 0.0155, 0.0114, 0.0162, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 01:17:32,498 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4197, 3.3157, 2.6451, 2.0590, 2.3675, 2.1260, 3.5722, 3.2742], device='cuda:0'), covar=tensor([0.2738, 0.0833, 0.1680, 0.2075, 0.2128, 0.1786, 0.0514, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0256, 0.0281, 0.0273, 0.0286, 0.0217, 0.0265, 0.0287], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:18:07,078 INFO [train.py:904] (0/8) Epoch 9, batch 4550, loss[loss=0.2346, simple_loss=0.3106, pruned_loss=0.0793, over 16796.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2993, pruned_loss=0.06652, over 3213153.36 frames. ], batch size: 83, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:18:20,801 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 01:18:23,676 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:35,038 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=85771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:18:44,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6154, 2.9855, 2.9848, 5.0386, 3.9254, 4.4039, 1.5978, 3.1769], device='cuda:0'), covar=tensor([0.1344, 0.0670, 0.0951, 0.0085, 0.0377, 0.0317, 0.1490, 0.0769], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0153, 0.0175, 0.0126, 0.0201, 0.0206, 0.0174, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 01:18:54,381 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:20,633 INFO [train.py:904] (0/8) Epoch 9, batch 4600, loss[loss=0.2022, simple_loss=0.2841, pruned_loss=0.06013, over 16722.00 frames. ], tot_loss[loss=0.2161, simple_loss=0.2996, pruned_loss=0.06627, over 3229319.22 frames. ], batch size: 62, lr: 7.64e-03, grad_scale: 16.0 2023-04-29 01:19:32,657 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:42,429 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.222e+02 2.575e+02 3.036e+02 5.036e+02, threshold=5.150e+02, percent-clipped=0.0 2023-04-29 01:19:44,976 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=85819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:19:58,165 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7717, 3.7163, 3.8417, 3.9946, 4.0477, 3.6704, 4.0132, 4.0878], device='cuda:0'), covar=tensor([0.1108, 0.0821, 0.1054, 0.0468, 0.0412, 0.1597, 0.0559, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0601, 0.0745, 0.0608, 0.0462, 0.0468, 0.0478, 0.0533], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:20:12,957 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85839.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:20:30,862 INFO [train.py:904] (0/8) Epoch 9, batch 4650, loss[loss=0.2125, simple_loss=0.2878, pruned_loss=0.06857, over 17015.00 frames. ], tot_loss[loss=0.2151, simple_loss=0.2985, pruned_loss=0.06589, over 3233026.32 frames. ], batch size: 55, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:21:04,782 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:20,349 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:21:42,734 INFO [train.py:904] (0/8) Epoch 9, batch 4700, loss[loss=0.1859, simple_loss=0.2727, pruned_loss=0.0496, over 16546.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2953, pruned_loss=0.06442, over 3228908.31 frames. ], batch size: 68, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:22:06,316 INFO [optim.py:368] (0/8) 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 01:22:25,235 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85930.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:22:32,897 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:22:55,510 INFO [train.py:904] (0/8) Epoch 9, batch 4750, loss[loss=0.1844, simple_loss=0.2662, pruned_loss=0.0513, over 16910.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2914, pruned_loss=0.06213, over 3227192.74 frames. ], batch size: 109, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:23:04,373 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=85957.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:23:06,764 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4002, 1.5446, 1.9379, 2.4304, 2.4761, 2.6421, 1.6606, 2.5741], device='cuda:0'), covar=tensor([0.0134, 0.0386, 0.0244, 0.0221, 0.0191, 0.0157, 0.0346, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0164, 0.0150, 0.0154, 0.0158, 0.0115, 0.0164, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 01:23:14,548 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2430, 5.0339, 5.1983, 5.4980, 5.6175, 4.9472, 5.5836, 5.6055], device='cuda:0'), covar=tensor([0.1220, 0.0852, 0.1244, 0.0431, 0.0371, 0.0624, 0.0339, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0595, 0.0737, 0.0602, 0.0461, 0.0463, 0.0475, 0.0527], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:23:19,935 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1240, 3.3886, 3.5520, 3.5152, 3.5333, 3.3283, 3.3698, 3.3863], device='cuda:0'), covar=tensor([0.0346, 0.0497, 0.0445, 0.0468, 0.0396, 0.0400, 0.0763, 0.0433], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0302, 0.0312, 0.0299, 0.0351, 0.0324, 0.0430, 0.0259], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 01:23:39,947 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=85981.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:23:54,620 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=85991.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:24:07,806 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-86000.pt 2023-04-29 01:24:13,783 INFO [train.py:904] (0/8) Epoch 9, batch 4800, loss[loss=0.2032, simple_loss=0.2954, pruned_loss=0.05554, over 16408.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2877, pruned_loss=0.06026, over 3223584.32 frames. ], batch size: 146, lr: 7.63e-03, grad_scale: 16.0 2023-04-29 01:24:14,251 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:24:17,609 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9017, 4.1915, 3.9606, 4.0518, 3.6911, 3.8024, 3.8551, 4.1334], device='cuda:0'), covar=tensor([0.0935, 0.0827, 0.0849, 0.0580, 0.0704, 0.1448, 0.0793, 0.0995], device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0607, 0.0508, 0.0414, 0.0381, 0.0394, 0.0507, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:24:37,226 INFO [optim.py:368] (0/8) 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 01:24:38,403 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:24:59,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1023, 4.4506, 3.9600, 4.3574, 4.0259, 3.9517, 4.0684, 4.4068], device='cuda:0'), covar=tensor([0.1925, 0.1383, 0.2065, 0.0970, 0.1308, 0.2373, 0.1688, 0.1562], device='cuda:0'), in_proj_covar=tensor([0.0483, 0.0607, 0.0509, 0.0413, 0.0381, 0.0394, 0.0508, 0.0456], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:25:28,107 INFO [train.py:904] (0/8) Epoch 9, batch 4850, loss[loss=0.1885, simple_loss=0.2839, pruned_loss=0.04655, over 16853.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2876, pruned_loss=0.05864, over 3236523.08 frames. ], batch size: 96, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:25:28,532 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:37,960 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 01:25:45,089 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:25:54,085 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:16,479 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:26:42,277 INFO [train.py:904] (0/8) Epoch 9, batch 4900, loss[loss=0.1905, simple_loss=0.2781, pruned_loss=0.05148, over 16717.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2869, pruned_loss=0.05753, over 3220445.37 frames. ], batch size: 83, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:26:51,504 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4049, 4.4026, 4.2537, 4.0469, 3.9045, 4.3233, 4.1508, 4.0429], device='cuda:0'), covar=tensor([0.0521, 0.0402, 0.0247, 0.0229, 0.0868, 0.0408, 0.0491, 0.0550], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0269, 0.0263, 0.0237, 0.0284, 0.0269, 0.0180, 0.0302], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:26:59,208 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 01:27:08,497 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3558, 3.2312, 2.6760, 2.0322, 2.2108, 2.1272, 3.2693, 3.1446], device='cuda:0'), covar=tensor([0.2435, 0.0700, 0.1503, 0.2136, 0.2034, 0.1661, 0.0537, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0252, 0.0277, 0.0271, 0.0280, 0.0214, 0.0261, 0.0282], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:27:13,965 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5050, 3.5752, 3.2987, 3.1407, 3.0844, 3.4148, 3.2302, 3.2812], device='cuda:0'), covar=tensor([0.0500, 0.0372, 0.0235, 0.0224, 0.0570, 0.0320, 0.1164, 0.0411], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0269, 0.0262, 0.0237, 0.0284, 0.0269, 0.0179, 0.0301], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:27:23,788 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:25,908 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:37,218 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86139.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:27:55,679 INFO [train.py:904] (0/8) Epoch 9, batch 4950, loss[loss=0.221, simple_loss=0.308, pruned_loss=0.06702, over 16708.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2869, pruned_loss=0.05731, over 3212892.98 frames. ], batch size: 134, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:28:22,588 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5235, 2.0689, 1.6251, 1.9955, 2.4876, 2.2189, 2.5558, 2.6673], device='cuda:0'), covar=tensor([0.0089, 0.0305, 0.0397, 0.0317, 0.0165, 0.0259, 0.0116, 0.0186], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0188, 0.0186, 0.0183, 0.0184, 0.0189, 0.0186, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:28:33,048 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2319, 4.2915, 4.0664, 3.8755, 3.6823, 4.1943, 4.0093, 3.9246], device='cuda:0'), covar=tensor([0.0515, 0.0355, 0.0274, 0.0247, 0.0960, 0.0331, 0.0512, 0.0525], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0271, 0.0264, 0.0238, 0.0286, 0.0270, 0.0180, 0.0302], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:28:45,918 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86186.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:28:46,988 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:01,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3195, 3.1621, 2.5981, 1.9968, 2.2970, 2.0641, 3.2263, 3.0924], device='cuda:0'), covar=tensor([0.2408, 0.0722, 0.1429, 0.1820, 0.1693, 0.1594, 0.0538, 0.0821], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0251, 0.0275, 0.0268, 0.0276, 0.0212, 0.0260, 0.0279], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:29:08,683 INFO [train.py:904] (0/8) Epoch 9, batch 5000, loss[loss=0.2098, simple_loss=0.2968, pruned_loss=0.06135, over 15332.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2888, pruned_loss=0.05808, over 3211566.65 frames. ], batch size: 190, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:29:32,040 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.365e+02 2.931e+02 3.414e+02 6.733e+02, threshold=5.862e+02, percent-clipped=2.0 2023-04-29 01:29:49,858 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86231.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:29:54,567 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:30:11,768 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 9, batch 5050, loss[loss=0.2264, simple_loss=0.3075, pruned_loss=0.07269, over 16448.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2892, pruned_loss=0.05791, over 3230165.19 frames. ], batch size: 35, lr: 7.62e-03, grad_scale: 16.0 2023-04-29 01:30:23,877 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.64 vs. limit=2.0 2023-04-29 01:31:01,281 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:07,941 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86286.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 01:31:29,542 INFO [train.py:904] (0/8) Epoch 9, batch 5100, loss[loss=0.1987, simple_loss=0.2868, pruned_loss=0.05528, over 16452.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2871, pruned_loss=0.05687, over 3231379.05 frames. ], batch size: 146, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:31:36,853 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:39,981 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5487, 2.4289, 1.9783, 2.2414, 2.9550, 2.6261, 3.4177, 3.1724], device='cuda:0'), covar=tensor([0.0039, 0.0266, 0.0337, 0.0272, 0.0141, 0.0226, 0.0079, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0189, 0.0188, 0.0185, 0.0185, 0.0191, 0.0187, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:31:45,612 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86313.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:31:52,766 INFO [optim.py:368] (0/8) 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 01:32:08,057 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:32:24,799 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4254, 3.4011, 2.6077, 2.0972, 2.4689, 2.2915, 3.5896, 3.3635], device='cuda:0'), covar=tensor([0.2445, 0.0841, 0.1558, 0.2180, 0.1776, 0.1525, 0.0502, 0.0865], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0253, 0.0277, 0.0269, 0.0278, 0.0213, 0.0262, 0.0280], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:32:41,547 INFO [train.py:904] (0/8) Epoch 9, batch 5150, loss[loss=0.1919, simple_loss=0.2968, pruned_loss=0.04344, over 16807.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2875, pruned_loss=0.05631, over 3219050.23 frames. ], batch size: 96, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:32:50,292 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86358.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:33:54,439 INFO [train.py:904] (0/8) Epoch 9, batch 5200, loss[loss=0.216, simple_loss=0.3009, pruned_loss=0.06552, over 16285.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2871, pruned_loss=0.05652, over 3206371.57 frames. ], batch size: 165, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:34:02,839 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:17,302 INFO [optim.py:368] (0/8) 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 01:34:27,764 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:50,113 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:34:57,908 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:35:06,338 INFO [train.py:904] (0/8) Epoch 9, batch 5250, loss[loss=0.2031, simple_loss=0.2954, pruned_loss=0.05542, over 16891.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2849, pruned_loss=0.05621, over 3212855.96 frames. ], batch size: 96, lr: 7.61e-03, grad_scale: 16.0 2023-04-29 01:35:24,404 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.80 vs. limit=5.0 2023-04-29 01:36:02,605 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 01:36:09,177 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.52 vs. limit=5.0 2023-04-29 01:36:17,623 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86501.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:36:19,011 INFO [train.py:904] (0/8) Epoch 9, batch 5300, loss[loss=0.1855, simple_loss=0.2614, pruned_loss=0.05477, over 16365.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2815, pruned_loss=0.05529, over 3214187.18 frames. ], batch size: 146, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:36:25,652 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86506.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:36:42,027 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.402e+02 2.448e+02 2.767e+02 3.292e+02 5.111e+02, threshold=5.534e+02, percent-clipped=0.0 2023-04-29 01:37:01,926 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86531.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:37:32,905 INFO [train.py:904] (0/8) Epoch 9, batch 5350, loss[loss=0.1909, simple_loss=0.2767, pruned_loss=0.05257, over 16259.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2794, pruned_loss=0.05432, over 3212354.34 frames. ], batch size: 35, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:11,782 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:13,777 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:38:23,208 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86586.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:38:45,875 INFO [train.py:904] (0/8) Epoch 9, batch 5400, loss[loss=0.2142, simple_loss=0.3036, pruned_loss=0.06242, over 16941.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2825, pruned_loss=0.05504, over 3218123.34 frames. ], batch size: 109, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:38:46,271 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86602.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:01,702 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86613.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:39:09,038 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.152e+02 2.423e+02 2.907e+02 4.710e+02, threshold=4.847e+02, percent-clipped=0.0 2023-04-29 01:39:31,535 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86634.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:39:41,812 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:00,477 INFO [train.py:904] (0/8) Epoch 9, batch 5450, loss[loss=0.291, simple_loss=0.3519, pruned_loss=0.115, over 15268.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2862, pruned_loss=0.05734, over 3213953.83 frames. ], batch size: 190, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:40:11,025 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86658.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:40:15,342 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 9, batch 5500, loss[loss=0.2311, simple_loss=0.3161, pruned_loss=0.07307, over 16425.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2944, pruned_loss=0.06304, over 3181158.32 frames. ], batch size: 75, lr: 7.60e-03, grad_scale: 16.0 2023-04-29 01:41:24,320 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:27,613 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86708.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:41:43,583 INFO [optim.py:368] (0/8) 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 01:41:50,065 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 01:41:54,558 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:42:03,980 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 01:42:36,879 INFO [train.py:904] (0/8) Epoch 9, batch 5550, loss[loss=0.228, simple_loss=0.3182, pruned_loss=0.06892, over 16702.00 frames. ], tot_loss[loss=0.221, simple_loss=0.303, pruned_loss=0.06943, over 3149853.43 frames. ], batch size: 89, lr: 7.59e-03, grad_scale: 16.0 2023-04-29 01:42:43,066 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:09,703 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=86773.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:42,292 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5860, 4.8977, 4.6353, 4.6830, 4.3720, 4.3213, 4.4535, 4.9593], device='cuda:0'), covar=tensor([0.0914, 0.0797, 0.1064, 0.0690, 0.0715, 0.0976, 0.0842, 0.0826], device='cuda:0'), in_proj_covar=tensor([0.0484, 0.0612, 0.0507, 0.0414, 0.0378, 0.0392, 0.0503, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:43:44,647 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86796.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:43:53,946 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86801.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 01:43:54,679 INFO [train.py:904] (0/8) Epoch 9, batch 5600, loss[loss=0.3221, simple_loss=0.3831, pruned_loss=0.1306, over 15327.00 frames. ], tot_loss[loss=0.228, simple_loss=0.3081, pruned_loss=0.07397, over 3128659.98 frames. ], batch size: 190, lr: 7.59e-03, grad_scale: 8.0 2023-04-29 01:44:14,072 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:44:15,771 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7541, 3.7700, 3.8535, 3.7140, 3.7983, 4.1874, 3.8917, 3.6142], device='cuda:0'), covar=tensor([0.1951, 0.1910, 0.1790, 0.2245, 0.2534, 0.1466, 0.1472, 0.2435], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0443, 0.0471, 0.0389, 0.0519, 0.0496, 0.0382, 0.0527], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 01:44:21,511 INFO [optim.py:368] (0/8) 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 01:44:36,699 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-29 01:45:17,297 INFO [train.py:904] (0/8) Epoch 9, batch 5650, loss[loss=0.2038, simple_loss=0.289, pruned_loss=0.05932, over 16505.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3135, pruned_loss=0.07829, over 3111286.54 frames. ], batch size: 35, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:45:19,143 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0253, 4.0261, 4.4023, 2.2867, 4.8770, 4.8020, 3.1825, 3.4572], device='cuda:0'), covar=tensor([0.0631, 0.0163, 0.0132, 0.1002, 0.0029, 0.0063, 0.0354, 0.0385], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0097, 0.0085, 0.0138, 0.0067, 0.0096, 0.0120, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 01:45:53,458 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:45:58,534 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=86878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:33,114 INFO [train.py:904] (0/8) Epoch 9, batch 5700, loss[loss=0.2367, simple_loss=0.3185, pruned_loss=0.07742, over 16689.00 frames. ], tot_loss[loss=0.2389, simple_loss=0.3163, pruned_loss=0.0807, over 3095661.47 frames. ], batch size: 124, lr: 7.59e-03, grad_scale: 4.0 2023-04-29 01:46:33,473 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=86902.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:46:44,107 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 01:46:59,885 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.388e+02 3.612e+02 4.435e+02 5.746e+02 9.391e+02, threshold=8.870e+02, percent-clipped=0.0 2023-04-29 01:47:25,193 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=86936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:30,399 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=86939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:47:48,294 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 9, batch 5750, loss[loss=0.2134, simple_loss=0.301, pruned_loss=0.06289, over 16457.00 frames. ], tot_loss[loss=0.2418, simple_loss=0.319, pruned_loss=0.0823, over 3067512.61 frames. ], batch size: 75, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:48:04,122 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-29 01:48:38,684 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6779, 2.7352, 2.3644, 4.3944, 3.1282, 4.1945, 1.4456, 3.0648], device='cuda:0'), covar=tensor([0.1381, 0.0708, 0.1227, 0.0139, 0.0262, 0.0348, 0.1644, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0152, 0.0176, 0.0126, 0.0199, 0.0205, 0.0173, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 01:48:52,723 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 01:49:12,521 INFO [train.py:904] (0/8) Epoch 9, batch 5800, loss[loss=0.2341, simple_loss=0.3042, pruned_loss=0.08205, over 12305.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3183, pruned_loss=0.08112, over 3056788.04 frames. ], batch size: 247, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:49:13,816 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87002.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:49:34,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-29 01:49:40,723 INFO [optim.py:368] (0/8) 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 01:50:30,451 INFO [train.py:904] (0/8) Epoch 9, batch 5850, loss[loss=0.2363, simple_loss=0.3197, pruned_loss=0.07648, over 16712.00 frames. ], tot_loss[loss=0.2374, simple_loss=0.3159, pruned_loss=0.07944, over 3041424.76 frames. ], batch size: 89, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:50:47,498 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87063.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:51:37,568 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.75 vs. limit=5.0 2023-04-29 01:51:42,106 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:51:50,031 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87101.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 01:51:50,888 INFO [train.py:904] (0/8) Epoch 9, batch 5900, loss[loss=0.2147, simple_loss=0.3124, pruned_loss=0.05855, over 16262.00 frames. ], tot_loss[loss=0.2367, simple_loss=0.3153, pruned_loss=0.07904, over 3045550.31 frames. ], batch size: 35, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:52:00,299 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7770, 1.2933, 1.6758, 1.6658, 1.8425, 1.8266, 1.4553, 1.7715], device='cuda:0'), covar=tensor([0.0163, 0.0280, 0.0126, 0.0180, 0.0162, 0.0103, 0.0293, 0.0064], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0162, 0.0145, 0.0148, 0.0155, 0.0112, 0.0164, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 01:52:22,907 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.063e+02 3.002e+02 3.675e+02 4.500e+02 8.851e+02, threshold=7.350e+02, percent-clipped=1.0 2023-04-29 01:53:01,035 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87144.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:09,064 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87149.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:13,008 INFO [train.py:904] (0/8) Epoch 9, batch 5950, loss[loss=0.2229, simple_loss=0.3146, pruned_loss=0.06555, over 16888.00 frames. ], tot_loss[loss=0.2351, simple_loss=0.3158, pruned_loss=0.07725, over 3064528.81 frames. ], batch size: 90, lr: 7.58e-03, grad_scale: 4.0 2023-04-29 01:53:42,058 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:53:43,587 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 01:54:17,388 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4463, 3.3868, 3.6582, 1.7546, 3.8980, 3.9415, 2.9244, 2.8328], device='cuda:0'), covar=tensor([0.0674, 0.0165, 0.0155, 0.1163, 0.0044, 0.0091, 0.0347, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0097, 0.0084, 0.0138, 0.0066, 0.0094, 0.0119, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-29 01:54:33,106 INFO [train.py:904] (0/8) Epoch 9, batch 6000, loss[loss=0.2271, simple_loss=0.3033, pruned_loss=0.07543, over 16933.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3144, pruned_loss=0.0763, over 3081448.26 frames. ], batch size: 116, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:54:33,107 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 01:54:44,301 INFO [train.py:938] (0/8) Epoch 9, validation: loss=0.1674, simple_loss=0.2809, pruned_loss=0.02692, over 944034.00 frames. 2023-04-29 01:54:44,302 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 01:55:11,809 INFO [optim.py:368] (0/8) 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 01:55:25,056 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6614, 4.7063, 4.5219, 4.3586, 4.0538, 4.6377, 4.5233, 4.2751], device='cuda:0'), covar=tensor([0.0602, 0.0454, 0.0279, 0.0265, 0.1059, 0.0382, 0.0394, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0276, 0.0264, 0.0240, 0.0287, 0.0277, 0.0182, 0.0309], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:55:28,024 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87230.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:34,792 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:55:37,451 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87236.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:56:03,559 INFO [train.py:904] (0/8) Epoch 9, batch 6050, loss[loss=0.2216, simple_loss=0.3133, pruned_loss=0.06496, over 16801.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.3122, pruned_loss=0.07533, over 3075093.44 frames. ], batch size: 83, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:56:16,819 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-04-29 01:56:52,087 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87284.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:06,981 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87291.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 01:57:23,346 INFO [train.py:904] (0/8) Epoch 9, batch 6100, loss[loss=0.2126, simple_loss=0.2977, pruned_loss=0.0637, over 16856.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3113, pruned_loss=0.07334, over 3104031.96 frames. ], batch size: 116, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:57:52,505 INFO [optim.py:368] (0/8) 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 01:58:42,527 INFO [train.py:904] (0/8) Epoch 9, batch 6150, loss[loss=0.2114, simple_loss=0.2969, pruned_loss=0.063, over 16885.00 frames. ], tot_loss[loss=0.2269, simple_loss=0.3089, pruned_loss=0.07249, over 3111152.80 frames. ], batch size: 116, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 01:58:48,475 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6809, 3.8293, 4.1044, 1.7884, 4.3608, 4.4628, 3.3157, 3.1951], device='cuda:0'), covar=tensor([0.0711, 0.0156, 0.0142, 0.1246, 0.0068, 0.0072, 0.0300, 0.0436], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0098, 0.0085, 0.0139, 0.0067, 0.0096, 0.0120, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 01:58:52,915 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87358.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 01:58:54,251 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2639, 4.2830, 4.0965, 3.9432, 3.7692, 4.2099, 4.0022, 3.8746], device='cuda:0'), covar=tensor([0.0462, 0.0343, 0.0271, 0.0239, 0.0809, 0.0370, 0.0584, 0.0606], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0273, 0.0264, 0.0240, 0.0287, 0.0277, 0.0180, 0.0310], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 01:58:58,689 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87362.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:00:00,637 INFO [train.py:904] (0/8) Epoch 9, batch 6200, loss[loss=0.2726, simple_loss=0.3254, pruned_loss=0.1098, over 11815.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3082, pruned_loss=0.07289, over 3100769.93 frames. ], batch size: 248, lr: 7.57e-03, grad_scale: 8.0 2023-04-29 02:00:28,306 INFO [optim.py:368] (0/8) 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 02:00:34,937 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87423.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:18,833 INFO [train.py:904] (0/8) Epoch 9, batch 6250, loss[loss=0.2273, simple_loss=0.2983, pruned_loss=0.07813, over 12022.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3073, pruned_loss=0.07267, over 3096343.45 frames. ], batch size: 247, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:01:32,076 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:01:45,464 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:12,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4541, 4.2022, 4.4103, 4.6327, 4.7538, 4.3051, 4.7127, 4.7567], device='cuda:0'), covar=tensor([0.1368, 0.1077, 0.1459, 0.0581, 0.0531, 0.0931, 0.0654, 0.0549], device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0606, 0.0739, 0.0610, 0.0466, 0.0463, 0.0489, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:02:20,054 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 02:02:27,600 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 02:02:33,532 INFO [train.py:904] (0/8) Epoch 9, batch 6300, loss[loss=0.2274, simple_loss=0.3087, pruned_loss=0.07308, over 16709.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.308, pruned_loss=0.07235, over 3111382.67 frames. ], batch size: 62, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:02:38,434 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:02:59,296 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:04,116 INFO [optim.py:368] (0/8) 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 02:03:06,352 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87522.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:24,793 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87534.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:03:51,236 INFO [train.py:904] (0/8) Epoch 9, batch 6350, loss[loss=0.2141, simple_loss=0.2921, pruned_loss=0.06808, over 16591.00 frames. ], tot_loss[loss=0.2273, simple_loss=0.3083, pruned_loss=0.07317, over 3118311.21 frames. ], batch size: 57, lr: 7.56e-03, grad_scale: 4.0 2023-04-29 02:04:12,971 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87566.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 02:04:37,475 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87582.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:40,720 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87584.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:44,435 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87586.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:04:49,465 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:08,854 INFO [train.py:904] (0/8) Epoch 9, batch 6400, loss[loss=0.232, simple_loss=0.3081, pruned_loss=0.07792, over 16890.00 frames. ], tot_loss[loss=0.2301, simple_loss=0.3099, pruned_loss=0.07517, over 3093024.98 frames. ], batch size: 116, lr: 7.56e-03, grad_scale: 8.0 2023-04-29 02:05:13,511 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=87605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:05:31,958 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7822, 5.0992, 4.8365, 4.8153, 4.5816, 4.4963, 4.4903, 5.1969], device='cuda:0'), covar=tensor([0.1040, 0.0840, 0.1070, 0.0690, 0.0760, 0.0886, 0.1011, 0.0771], device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0626, 0.0526, 0.0427, 0.0388, 0.0406, 0.0523, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:05:37,571 INFO [optim.py:368] (0/8) 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 02:05:59,409 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.82 vs. limit=5.0 2023-04-29 02:06:15,038 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87645.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:06:21,655 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:06:24,602 INFO [train.py:904] (0/8) Epoch 9, batch 6450, loss[loss=0.2045, simple_loss=0.2879, pruned_loss=0.06051, over 16679.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3093, pruned_loss=0.0741, over 3098706.03 frames. ], batch size: 124, lr: 7.55e-03, grad_scale: 4.0 2023-04-29 02:06:34,320 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87658.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:06:46,428 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=87666.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:07:41,898 INFO [train.py:904] (0/8) Epoch 9, batch 6500, loss[loss=0.2349, simple_loss=0.3173, pruned_loss=0.07631, over 16358.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3069, pruned_loss=0.07287, over 3112288.27 frames. ], batch size: 146, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:07:48,427 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87706.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:08:06,446 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:08:13,255 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.348e+02 3.236e+02 4.021e+02 5.179e+02 1.078e+03, threshold=8.043e+02, percent-clipped=2.0 2023-04-29 02:09:00,847 INFO [train.py:904] (0/8) Epoch 9, batch 6550, loss[loss=0.2239, simple_loss=0.3248, pruned_loss=0.06147, over 16533.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.3099, pruned_loss=0.07382, over 3123293.21 frames. ], batch size: 75, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:09:16,169 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6158, 4.0287, 3.7050, 2.0365, 3.0390, 2.6614, 3.7712, 3.8815], device='cuda:0'), covar=tensor([0.0228, 0.0517, 0.0548, 0.1773, 0.0749, 0.0853, 0.0589, 0.0722], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0138, 0.0156, 0.0140, 0.0133, 0.0124, 0.0135, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 02:09:47,370 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 02:10:18,528 INFO [train.py:904] (0/8) Epoch 9, batch 6600, loss[loss=0.2374, simple_loss=0.3118, pruned_loss=0.08147, over 16409.00 frames. ], tot_loss[loss=0.2306, simple_loss=0.3123, pruned_loss=0.07444, over 3109850.61 frames. ], batch size: 68, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:10:41,983 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87817.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:10:51,648 INFO [optim.py:368] (0/8) 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 02:11:26,768 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 02:11:36,982 INFO [train.py:904] (0/8) Epoch 9, batch 6650, loss[loss=0.268, simple_loss=0.3254, pruned_loss=0.1053, over 11378.00 frames. ], tot_loss[loss=0.2314, simple_loss=0.3125, pruned_loss=0.0752, over 3101312.96 frames. ], batch size: 248, lr: 7.55e-03, grad_scale: 2.0 2023-04-29 02:11:51,008 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87861.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:12:28,644 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=87886.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:12:32,012 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 02:12:53,900 INFO [train.py:904] (0/8) Epoch 9, batch 6700, loss[loss=0.2487, simple_loss=0.3271, pruned_loss=0.08515, over 16698.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3112, pruned_loss=0.07518, over 3107142.64 frames. ], batch size: 89, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:13:26,707 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.105e+02 3.383e+02 4.194e+02 5.361e+02 9.838e+02, threshold=8.388e+02, percent-clipped=4.0 2023-04-29 02:13:43,772 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=87934.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:13:52,500 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87940.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:14:00,205 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:14:10,418 INFO [train.py:904] (0/8) Epoch 9, batch 6750, loss[loss=0.2095, simple_loss=0.2869, pruned_loss=0.06601, over 16743.00 frames. ], tot_loss[loss=0.2304, simple_loss=0.31, pruned_loss=0.07535, over 3096068.91 frames. ], batch size: 89, lr: 7.54e-03, grad_scale: 2.0 2023-04-29 02:14:24,282 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=87961.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:15:23,587 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-88000.pt 2023-04-29 02:15:29,305 INFO [train.py:904] (0/8) Epoch 9, batch 6800, loss[loss=0.2549, simple_loss=0.3173, pruned_loss=0.09627, over 11471.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3094, pruned_loss=0.07527, over 3092055.25 frames. ], batch size: 246, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:15:54,174 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:16:02,238 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 3.136e+02 3.707e+02 4.804e+02 8.151e+02, threshold=7.414e+02, percent-clipped=0.0 2023-04-29 02:16:45,307 INFO [train.py:904] (0/8) Epoch 9, batch 6850, loss[loss=0.2127, simple_loss=0.3179, pruned_loss=0.05376, over 16906.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3109, pruned_loss=0.07539, over 3105349.97 frames. ], batch size: 96, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:17:06,653 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:17:16,081 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4254, 2.0595, 2.1339, 4.1760, 1.8392, 2.6046, 2.0991, 2.2423], device='cuda:0'), covar=tensor([0.0891, 0.3119, 0.2111, 0.0341, 0.3899, 0.1995, 0.2992, 0.2750], device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0375, 0.0313, 0.0319, 0.0405, 0.0424, 0.0335, 0.0440], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:17:52,164 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9210, 5.1770, 4.9313, 4.9612, 4.6842, 4.6254, 4.6497, 5.2456], device='cuda:0'), covar=tensor([0.0914, 0.0797, 0.1000, 0.0614, 0.0773, 0.0829, 0.1026, 0.0870], device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0625, 0.0523, 0.0423, 0.0386, 0.0406, 0.0518, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:17:59,109 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88100.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:01,010 INFO [train.py:904] (0/8) Epoch 9, batch 6900, loss[loss=0.3014, simple_loss=0.3516, pruned_loss=0.1256, over 11517.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3128, pruned_loss=0.0744, over 3127218.58 frames. ], batch size: 247, lr: 7.54e-03, grad_scale: 4.0 2023-04-29 02:18:24,677 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:18:33,104 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.937e+02 3.211e+02 3.844e+02 4.463e+02 9.504e+02, threshold=7.687e+02, percent-clipped=4.0 2023-04-29 02:18:38,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8539, 3.8262, 4.2832, 2.0105, 4.5585, 4.4818, 3.1059, 3.3466], device='cuda:0'), covar=tensor([0.0661, 0.0166, 0.0133, 0.1052, 0.0034, 0.0080, 0.0355, 0.0370], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0099, 0.0086, 0.0140, 0.0067, 0.0097, 0.0120, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 02:18:49,144 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-29 02:19:03,951 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7852, 2.2846, 2.3546, 4.4189, 2.1984, 2.9464, 2.3553, 2.5228], device='cuda:0'), covar=tensor([0.0743, 0.2990, 0.1875, 0.0309, 0.3413, 0.1730, 0.2644, 0.2572], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0372, 0.0311, 0.0316, 0.0402, 0.0420, 0.0332, 0.0436], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:19:18,075 INFO [train.py:904] (0/8) Epoch 9, batch 6950, loss[loss=0.2214, simple_loss=0.3019, pruned_loss=0.07044, over 16548.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3149, pruned_loss=0.07695, over 3096015.60 frames. ], batch size: 146, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:19:32,139 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88161.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:19:32,193 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:19:38,180 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:20:05,623 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 02:20:31,303 INFO [train.py:904] (0/8) Epoch 9, batch 7000, loss[loss=0.2282, simple_loss=0.3144, pruned_loss=0.07103, over 16300.00 frames. ], tot_loss[loss=0.2335, simple_loss=0.3148, pruned_loss=0.07608, over 3106035.96 frames. ], batch size: 165, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:20:43,256 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:03,410 INFO [optim.py:368] (0/8) 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 02:21:10,617 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 02:21:29,047 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88240.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:35,997 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88245.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:21:44,998 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7299, 3.6345, 3.8096, 3.6780, 3.7742, 4.1855, 3.9082, 3.6192], device='cuda:0'), covar=tensor([0.2122, 0.2313, 0.2002, 0.2373, 0.2851, 0.1605, 0.1464, 0.2532], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0454, 0.0487, 0.0402, 0.0529, 0.0515, 0.0392, 0.0543], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 02:21:47,200 INFO [train.py:904] (0/8) Epoch 9, batch 7050, loss[loss=0.2334, simple_loss=0.3145, pruned_loss=0.07617, over 16702.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.316, pruned_loss=0.07673, over 3087631.57 frames. ], batch size: 124, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:22:00,429 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88261.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 02:22:40,452 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88288.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:22:45,429 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=88291.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:22:47,885 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:02,094 INFO [train.py:904] (0/8) Epoch 9, batch 7100, loss[loss=0.2288, simple_loss=0.3107, pruned_loss=0.07348, over 16538.00 frames. ], tot_loss[loss=0.2344, simple_loss=0.3148, pruned_loss=0.07697, over 3060270.57 frames. ], batch size: 146, lr: 7.53e-03, grad_scale: 4.0 2023-04-29 02:23:12,562 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:23:33,102 INFO [optim.py:368] (0/8) 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 02:24:00,854 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4892, 1.4907, 2.0497, 2.5472, 2.5225, 2.8451, 1.6957, 2.6671], device='cuda:0'), covar=tensor([0.0148, 0.0347, 0.0211, 0.0167, 0.0165, 0.0100, 0.0328, 0.0079], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0160, 0.0143, 0.0144, 0.0152, 0.0112, 0.0161, 0.0103], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 02:24:15,983 INFO [train.py:904] (0/8) Epoch 9, batch 7150, loss[loss=0.2314, simple_loss=0.3134, pruned_loss=0.0747, over 15288.00 frames. ], tot_loss[loss=0.2317, simple_loss=0.312, pruned_loss=0.07566, over 3085851.53 frames. ], batch size: 190, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:24:16,636 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=88352.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:25:10,853 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6360, 3.7027, 2.0441, 4.2210, 2.6713, 4.1227, 2.1639, 2.8608], device='cuda:0'), covar=tensor([0.0175, 0.0302, 0.1582, 0.0077, 0.0749, 0.0329, 0.1421, 0.0675], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0161, 0.0186, 0.0112, 0.0165, 0.0201, 0.0193, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 02:25:28,110 INFO [train.py:904] (0/8) Epoch 9, batch 7200, loss[loss=0.2131, simple_loss=0.2949, pruned_loss=0.0657, over 11829.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3093, pruned_loss=0.07377, over 3070921.65 frames. ], batch size: 246, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:00,149 INFO [optim.py:368] (0/8) 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 02:26:45,476 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 02:26:47,217 INFO [train.py:904] (0/8) Epoch 9, batch 7250, loss[loss=0.2492, simple_loss=0.3137, pruned_loss=0.09231, over 11880.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3072, pruned_loss=0.07253, over 3067177.78 frames. ], batch size: 248, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:26:53,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88456.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:28:00,997 INFO [train.py:904] (0/8) Epoch 9, batch 7300, loss[loss=0.2326, simple_loss=0.3237, pruned_loss=0.07077, over 16447.00 frames. ], tot_loss[loss=0.2253, simple_loss=0.3062, pruned_loss=0.07218, over 3066379.38 frames. ], batch size: 146, lr: 7.52e-03, grad_scale: 8.0 2023-04-29 02:28:33,454 INFO [optim.py:368] (0/8) 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 02:29:14,112 INFO [train.py:904] (0/8) Epoch 9, batch 7350, loss[loss=0.2496, simple_loss=0.315, pruned_loss=0.0921, over 11227.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3071, pruned_loss=0.07323, over 3055459.45 frames. ], batch size: 246, lr: 7.52e-03, grad_scale: 4.0 2023-04-29 02:29:29,720 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3899, 4.5672, 4.4619, 2.8417, 3.9163, 4.3651, 3.9391, 2.6489], device='cuda:0'), covar=tensor([0.0375, 0.0014, 0.0022, 0.0254, 0.0049, 0.0065, 0.0040, 0.0270], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0063, 0.0065, 0.0123, 0.0072, 0.0083, 0.0072, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 02:30:06,931 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8933, 2.2389, 1.6814, 1.9374, 2.6384, 2.3561, 2.8452, 2.8604], device='cuda:0'), covar=tensor([0.0069, 0.0280, 0.0391, 0.0352, 0.0151, 0.0273, 0.0134, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0186, 0.0185, 0.0183, 0.0184, 0.0187, 0.0185, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:30:27,932 INFO [train.py:904] (0/8) Epoch 9, batch 7400, loss[loss=0.236, simple_loss=0.3158, pruned_loss=0.0781, over 16444.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3071, pruned_loss=0.07294, over 3071348.64 frames. ], batch size: 146, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:31:01,752 INFO [optim.py:368] (0/8) 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 02:31:30,558 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5419, 3.5812, 2.9156, 2.1759, 2.6024, 2.2718, 3.8065, 3.5452], device='cuda:0'), covar=tensor([0.2757, 0.0840, 0.1643, 0.2274, 0.2223, 0.1844, 0.0524, 0.0945], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0258, 0.0283, 0.0272, 0.0283, 0.0217, 0.0265, 0.0286], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:31:37,833 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=88647.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:31:43,879 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2116, 4.2416, 4.6275, 2.4559, 5.0048, 4.9709, 3.4265, 3.8388], device='cuda:0'), covar=tensor([0.0627, 0.0157, 0.0146, 0.0984, 0.0036, 0.0056, 0.0358, 0.0304], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0100, 0.0086, 0.0141, 0.0068, 0.0096, 0.0120, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 02:31:44,630 INFO [train.py:904] (0/8) Epoch 9, batch 7450, loss[loss=0.225, simple_loss=0.3102, pruned_loss=0.06989, over 16419.00 frames. ], tot_loss[loss=0.2284, simple_loss=0.3089, pruned_loss=0.07398, over 3084568.72 frames. ], batch size: 35, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:31:45,407 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2789, 3.9852, 4.0203, 2.5934, 3.5659, 3.9568, 3.6567, 2.1239], device='cuda:0'), covar=tensor([0.0408, 0.0032, 0.0025, 0.0283, 0.0062, 0.0073, 0.0050, 0.0333], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0063, 0.0065, 0.0123, 0.0072, 0.0083, 0.0072, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 02:32:38,996 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0276, 3.0972, 1.8500, 3.2762, 2.3407, 3.2613, 1.9066, 2.4725], device='cuda:0'), covar=tensor([0.0231, 0.0350, 0.1527, 0.0156, 0.0790, 0.0553, 0.1510, 0.0736], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0160, 0.0186, 0.0112, 0.0166, 0.0200, 0.0193, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 02:33:03,524 INFO [train.py:904] (0/8) Epoch 9, batch 7500, loss[loss=0.2662, simple_loss=0.3221, pruned_loss=0.1052, over 11378.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3091, pruned_loss=0.07352, over 3064684.06 frames. ], batch size: 250, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:33:36,931 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 3.551e+02 4.142e+02 5.326e+02 1.060e+03, threshold=8.283e+02, percent-clipped=5.0 2023-04-29 02:34:18,169 INFO [train.py:904] (0/8) Epoch 9, batch 7550, loss[loss=0.2235, simple_loss=0.2984, pruned_loss=0.07427, over 16595.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3093, pruned_loss=0.07466, over 3053453.99 frames. ], batch size: 57, lr: 7.51e-03, grad_scale: 4.0 2023-04-29 02:34:23,605 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88756.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:35:00,731 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4499, 2.1130, 2.1996, 4.1157, 1.9913, 2.6590, 2.2452, 2.2918], device='cuda:0'), covar=tensor([0.0858, 0.3178, 0.2128, 0.0377, 0.3968, 0.2019, 0.2797, 0.2996], device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0380, 0.0317, 0.0321, 0.0411, 0.0430, 0.0340, 0.0443], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:35:33,606 INFO [train.py:904] (0/8) Epoch 9, batch 7600, loss[loss=0.2109, simple_loss=0.2924, pruned_loss=0.06473, over 15362.00 frames. ], tot_loss[loss=0.2289, simple_loss=0.3084, pruned_loss=0.07468, over 3043128.85 frames. ], batch size: 190, lr: 7.51e-03, grad_scale: 8.0 2023-04-29 02:35:37,267 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88804.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:35:52,471 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5616, 2.7324, 2.3653, 3.8571, 2.7962, 3.9371, 1.3350, 2.8151], device='cuda:0'), covar=tensor([0.1373, 0.0618, 0.1116, 0.0139, 0.0210, 0.0353, 0.1522, 0.0764], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0156, 0.0179, 0.0127, 0.0203, 0.0206, 0.0177, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 02:35:55,012 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3608, 3.6959, 3.5451, 2.1175, 3.0762, 2.7421, 3.5983, 3.7455], device='cuda:0'), covar=tensor([0.0239, 0.0600, 0.0554, 0.1687, 0.0759, 0.0798, 0.0641, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0140, 0.0158, 0.0143, 0.0135, 0.0126, 0.0138, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 02:35:57,808 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4979, 4.5331, 4.3151, 3.7178, 4.3378, 1.4914, 4.1407, 4.1760], device='cuda:0'), covar=tensor([0.0077, 0.0061, 0.0138, 0.0338, 0.0078, 0.2443, 0.0109, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0102, 0.0150, 0.0144, 0.0118, 0.0166, 0.0133, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:36:05,555 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.836e+02 3.313e+02 3.789e+02 4.639e+02 1.052e+03, threshold=7.577e+02, percent-clipped=2.0 2023-04-29 02:36:45,926 INFO [train.py:904] (0/8) Epoch 9, batch 7650, loss[loss=0.2287, simple_loss=0.3193, pruned_loss=0.06906, over 16802.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3087, pruned_loss=0.07501, over 3062774.30 frames. ], batch size: 83, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:37:06,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0627, 3.2972, 3.5127, 3.5057, 3.4992, 3.3143, 3.3427, 3.3839], device='cuda:0'), covar=tensor([0.0428, 0.0619, 0.0415, 0.0472, 0.0470, 0.0476, 0.0818, 0.0484], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0306, 0.0313, 0.0302, 0.0349, 0.0327, 0.0433, 0.0265], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 02:37:45,804 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6300, 2.5888, 1.7132, 2.7473, 2.1622, 2.7777, 1.8873, 2.3006], device='cuda:0'), covar=tensor([0.0211, 0.0348, 0.1333, 0.0161, 0.0613, 0.0440, 0.1221, 0.0555], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0161, 0.0186, 0.0112, 0.0165, 0.0201, 0.0193, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 02:38:01,891 INFO [train.py:904] (0/8) Epoch 9, batch 7700, loss[loss=0.2499, simple_loss=0.3124, pruned_loss=0.09374, over 11537.00 frames. ], tot_loss[loss=0.23, simple_loss=0.3091, pruned_loss=0.07544, over 3048186.08 frames. ], batch size: 248, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:38:34,569 INFO [optim.py:368] (0/8) 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 02:39:08,857 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=88947.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 02:39:16,828 INFO [train.py:904] (0/8) Epoch 9, batch 7750, loss[loss=0.1887, simple_loss=0.2901, pruned_loss=0.04369, over 16805.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3089, pruned_loss=0.07476, over 3075815.68 frames. ], batch size: 102, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:39:28,563 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4954, 4.4735, 4.2569, 4.0459, 3.8900, 4.3688, 4.2533, 4.0344], device='cuda:0'), covar=tensor([0.0509, 0.0390, 0.0272, 0.0268, 0.0954, 0.0417, 0.0406, 0.0633], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0269, 0.0256, 0.0236, 0.0278, 0.0270, 0.0180, 0.0301], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:39:42,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5450, 4.6889, 4.8536, 4.7700, 4.7598, 5.3245, 4.8374, 4.5977], device='cuda:0'), covar=tensor([0.0995, 0.1865, 0.1652, 0.1687, 0.2250, 0.0901, 0.1321, 0.2256], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0455, 0.0489, 0.0405, 0.0529, 0.0518, 0.0395, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 02:40:20,779 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=88995.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:40:29,344 INFO [train.py:904] (0/8) Epoch 9, batch 7800, loss[loss=0.2973, simple_loss=0.3467, pruned_loss=0.1239, over 11589.00 frames. ], tot_loss[loss=0.233, simple_loss=0.3114, pruned_loss=0.07727, over 3057606.76 frames. ], batch size: 247, lr: 7.50e-03, grad_scale: 8.0 2023-04-29 02:41:02,679 INFO [optim.py:368] (0/8) 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 02:41:07,785 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:41:44,469 INFO [train.py:904] (0/8) Epoch 9, batch 7850, loss[loss=0.2185, simple_loss=0.3131, pruned_loss=0.06194, over 16936.00 frames. ], tot_loss[loss=0.2332, simple_loss=0.312, pruned_loss=0.07719, over 3049504.01 frames. ], batch size: 96, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:42:01,863 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7793, 1.6636, 1.5539, 1.4757, 1.7945, 1.6227, 1.7008, 1.9093], device='cuda:0'), covar=tensor([0.0078, 0.0159, 0.0230, 0.0221, 0.0120, 0.0168, 0.0121, 0.0118], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0186, 0.0186, 0.0184, 0.0186, 0.0188, 0.0187, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:42:27,621 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 02:42:38,982 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:42:57,238 INFO [train.py:904] (0/8) Epoch 9, batch 7900, loss[loss=0.2707, simple_loss=0.3547, pruned_loss=0.09341, over 15094.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3103, pruned_loss=0.07555, over 3063884.37 frames. ], batch size: 190, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:43:13,386 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6591, 2.4380, 2.5335, 4.6305, 2.1624, 2.9765, 2.5044, 2.6780], device='cuda:0'), covar=tensor([0.0841, 0.3020, 0.1886, 0.0298, 0.3594, 0.1917, 0.2589, 0.2606], device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0378, 0.0315, 0.0319, 0.0408, 0.0426, 0.0338, 0.0439], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:43:20,947 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5057, 3.4962, 3.4534, 2.8291, 3.4175, 2.0192, 3.1301, 2.8135], device='cuda:0'), covar=tensor([0.0118, 0.0086, 0.0127, 0.0227, 0.0076, 0.1909, 0.0112, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0115, 0.0102, 0.0149, 0.0143, 0.0118, 0.0165, 0.0133, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:43:28,446 INFO [optim.py:368] (0/8) 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 02:44:01,780 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0914, 3.3357, 3.5641, 3.5261, 3.5148, 3.3236, 3.3565, 3.4450], device='cuda:0'), covar=tensor([0.0439, 0.0629, 0.0390, 0.0472, 0.0503, 0.0459, 0.0812, 0.0433], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0303, 0.0310, 0.0297, 0.0349, 0.0325, 0.0429, 0.0260], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 02:44:12,931 INFO [train.py:904] (0/8) Epoch 9, batch 7950, loss[loss=0.2273, simple_loss=0.3048, pruned_loss=0.0749, over 16694.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3105, pruned_loss=0.07566, over 3065167.63 frames. ], batch size: 134, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:44:39,072 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89169.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:44:43,976 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8113, 4.6125, 4.8156, 5.0225, 5.1719, 4.6228, 5.1321, 5.1153], device='cuda:0'), covar=tensor([0.1272, 0.0881, 0.1139, 0.0522, 0.0418, 0.0626, 0.0434, 0.0465], device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0599, 0.0728, 0.0601, 0.0463, 0.0460, 0.0487, 0.0539], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:44:48,914 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 02:45:24,796 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4248, 1.9315, 1.4799, 1.6764, 2.2873, 2.0124, 2.4811, 2.5290], device='cuda:0'), covar=tensor([0.0105, 0.0291, 0.0438, 0.0377, 0.0208, 0.0295, 0.0191, 0.0172], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0187, 0.0185, 0.0184, 0.0185, 0.0187, 0.0188, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:45:26,770 INFO [train.py:904] (0/8) Epoch 9, batch 8000, loss[loss=0.2638, simple_loss=0.3213, pruned_loss=0.1032, over 11496.00 frames. ], tot_loss[loss=0.2313, simple_loss=0.3109, pruned_loss=0.07587, over 3077461.07 frames. ], batch size: 247, lr: 7.49e-03, grad_scale: 8.0 2023-04-29 02:45:59,521 INFO [optim.py:368] (0/8) 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 02:46:08,551 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89230.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:46:40,084 INFO [train.py:904] (0/8) Epoch 9, batch 8050, loss[loss=0.2544, simple_loss=0.3334, pruned_loss=0.08776, over 17183.00 frames. ], tot_loss[loss=0.2294, simple_loss=0.3096, pruned_loss=0.07457, over 3100205.74 frames. ], batch size: 46, lr: 7.49e-03, grad_scale: 4.0 2023-04-29 02:47:55,807 INFO [train.py:904] (0/8) Epoch 9, batch 8100, loss[loss=0.2183, simple_loss=0.2938, pruned_loss=0.07144, over 17015.00 frames. ], tot_loss[loss=0.2288, simple_loss=0.309, pruned_loss=0.07433, over 3085204.90 frames. ], batch size: 55, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:48:29,434 INFO [optim.py:368] (0/8) 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 02:49:10,981 INFO [train.py:904] (0/8) Epoch 9, batch 8150, loss[loss=0.2091, simple_loss=0.2887, pruned_loss=0.06473, over 16682.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3064, pruned_loss=0.07314, over 3102265.52 frames. ], batch size: 134, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:49:45,563 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:49:59,367 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89383.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:50:15,324 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 02:50:28,023 INFO [train.py:904] (0/8) Epoch 9, batch 8200, loss[loss=0.1837, simple_loss=0.2722, pruned_loss=0.0476, over 16727.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3032, pruned_loss=0.07194, over 3122484.91 frames. ], batch size: 89, lr: 7.48e-03, grad_scale: 4.0 2023-04-29 02:51:05,355 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89425.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:06,154 INFO [optim.py:368] (0/8) 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 02:51:20,879 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:51:27,457 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89439.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:51:47,439 INFO [train.py:904] (0/8) Epoch 9, batch 8250, loss[loss=0.2004, simple_loss=0.2788, pruned_loss=0.06097, over 11972.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3026, pruned_loss=0.06984, over 3092729.95 frames. ], batch size: 248, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:52:30,632 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3022, 1.9611, 2.0198, 3.7843, 1.9384, 2.4310, 2.0555, 2.0973], device='cuda:0'), covar=tensor([0.0767, 0.3396, 0.2192, 0.0365, 0.3822, 0.2139, 0.3080, 0.3030], device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0375, 0.0312, 0.0314, 0.0404, 0.0421, 0.0335, 0.0436], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:52:43,291 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:52:52,355 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2106, 3.0691, 3.1119, 1.8930, 3.3370, 3.3880, 2.8428, 2.7442], device='cuda:0'), covar=tensor([0.0669, 0.0170, 0.0204, 0.1015, 0.0067, 0.0113, 0.0314, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0097, 0.0084, 0.0139, 0.0067, 0.0093, 0.0117, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-29 02:53:03,778 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 02:53:05,812 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89500.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 02:53:08,819 INFO [train.py:904] (0/8) Epoch 9, batch 8300, loss[loss=0.2024, simple_loss=0.2769, pruned_loss=0.06391, over 11862.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2993, pruned_loss=0.06674, over 3074901.73 frames. ], batch size: 247, lr: 7.48e-03, grad_scale: 2.0 2023-04-29 02:53:21,209 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89509.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:47,014 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89525.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:53:48,529 INFO [optim.py:368] (0/8) 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 02:53:50,994 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5341, 3.1517, 3.1488, 1.9867, 2.6991, 2.2375, 3.0081, 3.2052], device='cuda:0'), covar=tensor([0.0390, 0.0677, 0.0468, 0.1623, 0.0804, 0.0884, 0.0815, 0.0740], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0135, 0.0153, 0.0138, 0.0131, 0.0123, 0.0133, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 02:54:31,543 INFO [train.py:904] (0/8) Epoch 9, batch 8350, loss[loss=0.1962, simple_loss=0.2924, pruned_loss=0.05001, over 16280.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2983, pruned_loss=0.06449, over 3071301.02 frames. ], batch size: 165, lr: 7.47e-03, grad_scale: 2.0 2023-04-29 02:55:01,157 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89570.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:55:52,206 INFO [train.py:904] (0/8) Epoch 9, batch 8400, loss[loss=0.1924, simple_loss=0.2717, pruned_loss=0.05655, over 12235.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2956, pruned_loss=0.06256, over 3052052.08 frames. ], batch size: 248, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:56:31,321 INFO [optim.py:368] (0/8) 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 02:57:13,783 INFO [train.py:904] (0/8) Epoch 9, batch 8450, loss[loss=0.1894, simple_loss=0.2834, pruned_loss=0.04769, over 16609.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2938, pruned_loss=0.06066, over 3059702.11 frames. ], batch size: 134, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:57:22,333 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2128, 3.0698, 3.1745, 1.8381, 3.3649, 3.4287, 2.8179, 2.7693], device='cuda:0'), covar=tensor([0.0692, 0.0192, 0.0164, 0.1070, 0.0061, 0.0109, 0.0345, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0095, 0.0083, 0.0136, 0.0066, 0.0091, 0.0115, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-29 02:57:59,641 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:04,760 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89683.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:58:34,972 INFO [train.py:904] (0/8) Epoch 9, batch 8500, loss[loss=0.2049, simple_loss=0.2729, pruned_loss=0.06841, over 11774.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2898, pruned_loss=0.05781, over 3068622.21 frames. ], batch size: 248, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 02:59:08,350 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1091, 2.6039, 2.6812, 1.9415, 2.8513, 2.8980, 2.5276, 2.4879], device='cuda:0'), covar=tensor([0.0590, 0.0171, 0.0159, 0.0853, 0.0074, 0.0139, 0.0341, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0094, 0.0082, 0.0135, 0.0065, 0.0090, 0.0114, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0003], device='cuda:0') 2023-04-29 02:59:09,645 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0171, 1.7856, 1.5976, 1.4872, 1.8889, 1.5874, 1.7452, 1.9318], device='cuda:0'), covar=tensor([0.0081, 0.0176, 0.0248, 0.0225, 0.0121, 0.0166, 0.0106, 0.0129], device='cuda:0'), in_proj_covar=tensor([0.0121, 0.0188, 0.0186, 0.0187, 0.0187, 0.0187, 0.0185, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 02:59:14,104 INFO [optim.py:368] (0/8) 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 02:59:20,474 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89730.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:22,205 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:24,972 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7682, 4.7214, 5.1334, 5.1121, 5.1089, 4.8549, 4.8702, 4.5769], device='cuda:0'), covar=tensor([0.0207, 0.0444, 0.0264, 0.0367, 0.0353, 0.0229, 0.0654, 0.0308], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0300, 0.0303, 0.0290, 0.0341, 0.0319, 0.0417, 0.0257], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 02:59:38,945 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 02:59:58,595 INFO [train.py:904] (0/8) Epoch 9, batch 8550, loss[loss=0.2222, simple_loss=0.3133, pruned_loss=0.06555, over 16133.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2877, pruned_loss=0.05663, over 3071458.11 frames. ], batch size: 165, lr: 7.47e-03, grad_scale: 4.0 2023-04-29 03:00:54,511 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:01:24,832 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89795.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:01:37,962 INFO [train.py:904] (0/8) Epoch 9, batch 8600, loss[loss=0.1842, simple_loss=0.2759, pruned_loss=0.04627, over 16400.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2878, pruned_loss=0.05592, over 3047498.44 frames. ], batch size: 68, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:02:25,202 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=89825.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:02:26,427 INFO [optim.py:368] (0/8) 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 03:03:15,794 INFO [train.py:904] (0/8) Epoch 9, batch 8650, loss[loss=0.1848, simple_loss=0.2702, pruned_loss=0.04968, over 12026.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2858, pruned_loss=0.0539, over 3068868.83 frames. ], batch size: 246, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:03:41,246 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0404, 4.1559, 2.4542, 4.6817, 2.9771, 4.5844, 2.3594, 3.2239], device='cuda:0'), covar=tensor([0.0161, 0.0231, 0.1304, 0.0076, 0.0740, 0.0360, 0.1335, 0.0564], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0154, 0.0179, 0.0109, 0.0160, 0.0191, 0.0188, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 03:03:47,289 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:47,434 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=89865.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:03:55,513 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 03:04:05,357 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=89873.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:05:02,356 INFO [train.py:904] (0/8) Epoch 9, batch 8700, loss[loss=0.1867, simple_loss=0.2802, pruned_loss=0.04663, over 15400.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2827, pruned_loss=0.05269, over 3067154.62 frames. ], batch size: 191, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:05:45,070 INFO [optim.py:368] (0/8) 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 03:05:45,890 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=89926.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:06:09,302 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 03:06:36,272 INFO [train.py:904] (0/8) Epoch 9, batch 8750, loss[loss=0.206, simple_loss=0.302, pruned_loss=0.05496, over 16739.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2822, pruned_loss=0.05197, over 3068430.72 frames. ], batch size: 124, lr: 7.46e-03, grad_scale: 4.0 2023-04-29 03:07:41,145 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.48 vs. limit=5.0 2023-04-29 03:08:25,358 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-90000.pt 2023-04-29 03:08:32,649 INFO [train.py:904] (0/8) Epoch 9, batch 8800, loss[loss=0.2005, simple_loss=0.2811, pruned_loss=0.05998, over 12493.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2808, pruned_loss=0.05137, over 3049804.05 frames. ], batch size: 248, lr: 7.46e-03, grad_scale: 8.0 2023-04-29 03:09:21,970 INFO [optim.py:368] (0/8) 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 03:09:31,740 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90030.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:44,388 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:09:54,873 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3477, 2.9272, 2.5681, 2.1641, 2.1353, 2.0604, 2.9041, 2.7355], device='cuda:0'), covar=tensor([0.2277, 0.0670, 0.1366, 0.1894, 0.2047, 0.1710, 0.0474, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0239, 0.0266, 0.0256, 0.0253, 0.0204, 0.0248, 0.0263], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:10:17,863 INFO [train.py:904] (0/8) Epoch 9, batch 8850, loss[loss=0.1807, simple_loss=0.2672, pruned_loss=0.04709, over 12395.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2829, pruned_loss=0.05063, over 3034642.21 frames. ], batch size: 248, lr: 7.45e-03, grad_scale: 8.0 2023-04-29 03:10:19,939 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9534, 3.9615, 4.2748, 4.3039, 4.2972, 4.0291, 4.0653, 3.9908], device='cuda:0'), covar=tensor([0.0247, 0.0374, 0.0408, 0.0324, 0.0300, 0.0305, 0.0699, 0.0345], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0293, 0.0298, 0.0290, 0.0335, 0.0312, 0.0411, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 03:11:11,513 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:18,217 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:11:49,568 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90095.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 03:12:02,868 INFO [train.py:904] (0/8) Epoch 9, batch 8900, loss[loss=0.2081, simple_loss=0.2972, pruned_loss=0.05952, over 16286.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2833, pruned_loss=0.04958, over 3048121.05 frames. ], batch size: 165, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:12:57,515 INFO [optim.py:368] (0/8) 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 03:13:07,421 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90129.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:10,461 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:13:48,656 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90143.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 03:14:08,722 INFO [train.py:904] (0/8) Epoch 9, batch 8950, loss[loss=0.1766, simple_loss=0.2641, pruned_loss=0.04453, over 16795.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2833, pruned_loss=0.05005, over 3067664.76 frames. ], batch size: 124, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:14:38,368 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:14:57,498 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.37 vs. limit=5.0 2023-04-29 03:15:31,245 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 03:15:34,977 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90191.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:15:57,373 INFO [train.py:904] (0/8) Epoch 9, batch 9000, loss[loss=0.1755, simple_loss=0.2635, pruned_loss=0.04379, over 17231.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2799, pruned_loss=0.04825, over 3080547.85 frames. ], batch size: 44, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:15:57,374 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 03:16:07,533 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 03:16:31,274 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:16:41,797 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2143, 4.0232, 4.2791, 4.4080, 4.5390, 4.0165, 4.5352, 4.5101], device='cuda:0'), covar=tensor([0.1440, 0.1160, 0.1352, 0.0679, 0.0530, 0.1224, 0.0519, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0458, 0.0566, 0.0687, 0.0573, 0.0440, 0.0441, 0.0460, 0.0516], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:16:47,986 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90221.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 03:16:58,708 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.701e+02 2.410e+02 2.881e+02 3.616e+02 7.746e+02, threshold=5.761e+02, percent-clipped=2.0 2023-04-29 03:17:05,036 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8929, 3.8926, 4.2965, 4.2788, 4.2543, 4.0006, 4.0561, 3.9699], device='cuda:0'), covar=tensor([0.0269, 0.0490, 0.0362, 0.0385, 0.0365, 0.0351, 0.0642, 0.0333], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0293, 0.0301, 0.0292, 0.0338, 0.0315, 0.0412, 0.0255], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 03:17:08,217 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:17:26,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6056, 3.6788, 3.4273, 3.2050, 3.2355, 3.5573, 3.3422, 3.3636], device='cuda:0'), covar=tensor([0.0504, 0.0426, 0.0257, 0.0221, 0.0608, 0.0360, 0.0909, 0.0445], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0261, 0.0252, 0.0233, 0.0269, 0.0262, 0.0172, 0.0291], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:17:51,891 INFO [train.py:904] (0/8) Epoch 9, batch 9050, loss[loss=0.2323, simple_loss=0.3043, pruned_loss=0.08011, over 12879.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2803, pruned_loss=0.0487, over 3078151.71 frames. ], batch size: 248, lr: 7.45e-03, grad_scale: 4.0 2023-04-29 03:18:09,681 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3546, 5.2500, 5.1135, 4.7919, 4.7018, 5.2263, 5.1475, 4.8362], device='cuda:0'), covar=tensor([0.0402, 0.0372, 0.0256, 0.0212, 0.0809, 0.0366, 0.0182, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0261, 0.0252, 0.0233, 0.0268, 0.0262, 0.0172, 0.0291], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:18:52,712 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-29 03:19:00,405 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8556, 5.4400, 5.5894, 5.3906, 5.4944, 5.9589, 5.5221, 5.2438], device='cuda:0'), covar=tensor([0.0884, 0.1338, 0.1489, 0.1765, 0.1891, 0.0718, 0.1220, 0.2029], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0437, 0.0466, 0.0383, 0.0502, 0.0486, 0.0380, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:19:19,161 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:19:36,856 INFO [train.py:904] (0/8) Epoch 9, batch 9100, loss[loss=0.1978, simple_loss=0.2879, pruned_loss=0.05384, over 16910.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2798, pruned_loss=0.049, over 3077496.11 frames. ], batch size: 116, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:20:34,115 INFO [optim.py:368] (0/8) 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 03:20:58,238 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:33,850 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90351.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:21:35,366 INFO [train.py:904] (0/8) Epoch 9, batch 9150, loss[loss=0.1791, simple_loss=0.2719, pruned_loss=0.04315, over 16259.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2801, pruned_loss=0.04872, over 3065296.91 frames. ], batch size: 165, lr: 7.44e-03, grad_scale: 4.0 2023-04-29 03:21:57,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8759, 2.2184, 2.1557, 2.9126, 1.8603, 3.2737, 1.6026, 2.5944], device='cuda:0'), covar=tensor([0.1284, 0.0651, 0.1147, 0.0143, 0.0092, 0.0473, 0.1395, 0.0798], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0150, 0.0175, 0.0123, 0.0185, 0.0201, 0.0174, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 03:22:44,967 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:23:22,761 INFO [train.py:904] (0/8) Epoch 9, batch 9200, loss[loss=0.1654, simple_loss=0.247, pruned_loss=0.04194, over 12444.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2759, pruned_loss=0.04765, over 3064073.59 frames. ], batch size: 248, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:23:42,377 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90412.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:24:07,628 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 9, batch 9250, loss[loss=0.186, simple_loss=0.277, pruned_loss=0.0475, over 16231.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2752, pruned_loss=0.04741, over 3076144.96 frames. ], batch size: 165, lr: 7.44e-03, grad_scale: 8.0 2023-04-29 03:25:14,820 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90459.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:12,494 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:38,507 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90496.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:26:49,741 INFO [train.py:904] (0/8) Epoch 9, batch 9300, loss[loss=0.1754, simple_loss=0.2704, pruned_loss=0.04016, over 16309.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2733, pruned_loss=0.0466, over 3068923.61 frames. ], batch size: 146, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:27:35,071 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:27:36,367 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90521.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 03:27:36,528 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0625, 3.7139, 3.4761, 1.9778, 2.8724, 2.4977, 3.3253, 3.6482], device='cuda:0'), covar=tensor([0.0281, 0.0608, 0.0587, 0.1924, 0.0764, 0.0937, 0.0775, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0129, 0.0151, 0.0138, 0.0129, 0.0121, 0.0131, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 03:27:46,920 INFO [optim.py:368] (0/8) 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 03:27:53,607 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 03:28:36,082 INFO [train.py:904] (0/8) Epoch 9, batch 9350, loss[loss=0.1908, simple_loss=0.2791, pruned_loss=0.05122, over 16204.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2733, pruned_loss=0.04675, over 3073981.94 frames. ], batch size: 165, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:28:48,624 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90557.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:13,846 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90569.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:29:42,199 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6592, 2.6604, 1.8281, 2.8571, 2.1832, 2.8023, 1.9951, 2.3915], device='cuda:0'), covar=tensor([0.0198, 0.0354, 0.1303, 0.0181, 0.0746, 0.0464, 0.1296, 0.0553], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0153, 0.0178, 0.0108, 0.0158, 0.0186, 0.0186, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 03:29:50,768 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90588.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:30:18,695 INFO [train.py:904] (0/8) Epoch 9, batch 9400, loss[loss=0.1492, simple_loss=0.2403, pruned_loss=0.02904, over 12931.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2733, pruned_loss=0.04623, over 3071382.90 frames. ], batch size: 246, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:30:25,312 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.72 vs. limit=5.0 2023-04-29 03:31:09,568 INFO [optim.py:368] (0/8) 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 03:32:00,910 INFO [train.py:904] (0/8) Epoch 9, batch 9450, loss[loss=0.1983, simple_loss=0.2888, pruned_loss=0.05387, over 12683.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2747, pruned_loss=0.04634, over 3063483.94 frames. ], batch size: 250, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:33:38,701 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0251, 5.4106, 5.1969, 5.2188, 4.8472, 4.8363, 4.8210, 5.4685], device='cuda:0'), covar=tensor([0.0873, 0.0772, 0.0850, 0.0599, 0.0658, 0.0688, 0.0803, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0595, 0.0485, 0.0406, 0.0369, 0.0387, 0.0496, 0.0443], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:33:43,978 INFO [train.py:904] (0/8) Epoch 9, batch 9500, loss[loss=0.1661, simple_loss=0.2679, pruned_loss=0.03211, over 16841.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2742, pruned_loss=0.046, over 3069231.18 frames. ], batch size: 102, lr: 7.43e-03, grad_scale: 8.0 2023-04-29 03:33:58,232 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:34:35,340 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.298e+02 2.954e+02 3.660e+02 6.291e+02, threshold=5.908e+02, percent-clipped=3.0 2023-04-29 03:35:30,049 INFO [train.py:904] (0/8) Epoch 9, batch 9550, loss[loss=0.1747, simple_loss=0.2703, pruned_loss=0.03954, over 16842.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2738, pruned_loss=0.04593, over 3082397.10 frames. ], batch size: 96, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:36:43,135 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:36:50,276 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6996, 3.7124, 4.0459, 4.0516, 4.0595, 3.7710, 3.8462, 3.7927], device='cuda:0'), covar=tensor([0.0320, 0.0590, 0.0514, 0.0481, 0.0435, 0.0427, 0.0795, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0290, 0.0296, 0.0287, 0.0334, 0.0312, 0.0403, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 03:37:11,192 INFO [train.py:904] (0/8) Epoch 9, batch 9600, loss[loss=0.2253, simple_loss=0.3168, pruned_loss=0.06685, over 15431.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2758, pruned_loss=0.04717, over 3073306.70 frames. ], batch size: 191, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:37:12,257 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.04 vs. limit=2.0 2023-04-29 03:37:37,099 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90815.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:37:50,006 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-04-29 03:37:54,425 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2596, 3.3547, 1.7362, 3.6454, 2.3938, 3.5819, 2.0735, 2.7743], device='cuda:0'), covar=tensor([0.0251, 0.0338, 0.1777, 0.0126, 0.0894, 0.0512, 0.1496, 0.0703], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0153, 0.0180, 0.0108, 0.0159, 0.0189, 0.0188, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 03:37:59,123 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90834.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:38:19,777 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3500, 4.3168, 4.1919, 3.9200, 3.8207, 4.2906, 4.0476, 3.9177], device='cuda:0'), covar=tensor([0.0511, 0.0474, 0.0286, 0.0271, 0.0902, 0.0405, 0.0572, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0255, 0.0247, 0.0229, 0.0263, 0.0258, 0.0170, 0.0285], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-29 03:38:55,312 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=90849.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:39:01,499 INFO [train.py:904] (0/8) Epoch 9, batch 9650, loss[loss=0.1889, simple_loss=0.2797, pruned_loss=0.04906, over 16198.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.278, pruned_loss=0.04788, over 3065239.83 frames. ], batch size: 165, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:39:03,439 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=90852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:39:29,976 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9133, 1.7262, 1.5062, 1.4865, 1.8944, 1.5340, 1.7698, 1.9339], device='cuda:0'), covar=tensor([0.0076, 0.0188, 0.0279, 0.0263, 0.0147, 0.0215, 0.0100, 0.0137], device='cuda:0'), in_proj_covar=tensor([0.0118, 0.0188, 0.0183, 0.0182, 0.0182, 0.0185, 0.0176, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:39:43,386 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 03:40:14,577 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6036, 4.6140, 4.4674, 4.1728, 4.1093, 4.5565, 4.3531, 4.1978], device='cuda:0'), covar=tensor([0.0478, 0.0315, 0.0280, 0.0253, 0.0847, 0.0348, 0.0404, 0.0618], device='cuda:0'), in_proj_covar=tensor([0.0211, 0.0253, 0.0246, 0.0227, 0.0262, 0.0256, 0.0169, 0.0284], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-29 03:40:18,966 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=90888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:40:48,114 INFO [train.py:904] (0/8) Epoch 9, batch 9700, loss[loss=0.1872, simple_loss=0.278, pruned_loss=0.04823, over 16625.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2769, pruned_loss=0.04763, over 3076653.42 frames. ], batch size: 134, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:41:04,571 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=90910.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:41:40,521 INFO [optim.py:368] (0/8) 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 03:41:59,659 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=90936.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:42:31,492 INFO [train.py:904] (0/8) Epoch 9, batch 9750, loss[loss=0.1854, simple_loss=0.2819, pruned_loss=0.04448, over 16913.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2759, pruned_loss=0.04781, over 3079288.70 frames. ], batch size: 116, lr: 7.42e-03, grad_scale: 8.0 2023-04-29 03:44:10,884 INFO [train.py:904] (0/8) Epoch 9, batch 9800, loss[loss=0.1893, simple_loss=0.2873, pruned_loss=0.04567, over 16242.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2758, pruned_loss=0.0467, over 3096800.92 frames. ], batch size: 165, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:44:21,872 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:44:57,917 INFO [optim.py:368] (0/8) 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 03:45:32,276 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.05 vs. limit=5.0 2023-04-29 03:45:40,774 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 03:45:43,481 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.37 vs. limit=5.0 2023-04-29 03:45:57,966 INFO [train.py:904] (0/8) Epoch 9, batch 9850, loss[loss=0.171, simple_loss=0.2745, pruned_loss=0.03378, over 17000.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2767, pruned_loss=0.04636, over 3086631.41 frames. ], batch size: 41, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:46:04,902 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:46:10,638 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8662, 1.7599, 1.5229, 1.4365, 1.9349, 1.5450, 1.6911, 1.9576], device='cuda:0'), covar=tensor([0.0065, 0.0218, 0.0284, 0.0271, 0.0147, 0.0213, 0.0095, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0116, 0.0188, 0.0182, 0.0182, 0.0182, 0.0185, 0.0176, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:47:48,707 INFO [train.py:904] (0/8) Epoch 9, batch 9900, loss[loss=0.1938, simple_loss=0.2702, pruned_loss=0.05867, over 12613.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2769, pruned_loss=0.04632, over 3064949.21 frames. ], batch size: 246, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:48:20,189 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:48:47,936 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 9, batch 9950, loss[loss=0.1923, simple_loss=0.2798, pruned_loss=0.05243, over 16910.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2789, pruned_loss=0.04655, over 3081038.46 frames. ], batch size: 116, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:49:49,260 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:50:01,413 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0917, 3.7857, 3.7941, 2.4490, 3.5701, 3.7464, 3.6664, 2.2284], device='cuda:0'), covar=tensor([0.0442, 0.0024, 0.0027, 0.0296, 0.0050, 0.0055, 0.0039, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0062, 0.0064, 0.0120, 0.0071, 0.0079, 0.0071, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 03:50:14,118 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91163.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:51:15,549 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0269, 3.4211, 3.3878, 2.3122, 3.2241, 3.3950, 3.3547, 2.0316], device='cuda:0'), covar=tensor([0.0397, 0.0028, 0.0033, 0.0265, 0.0060, 0.0049, 0.0046, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0062, 0.0064, 0.0120, 0.0071, 0.0079, 0.0071, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 03:51:44,353 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91200.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:51:47,245 INFO [train.py:904] (0/8) Epoch 9, batch 10000, loss[loss=0.1953, simple_loss=0.29, pruned_loss=0.05027, over 16249.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2772, pruned_loss=0.04601, over 3088676.86 frames. ], batch size: 165, lr: 7.41e-03, grad_scale: 8.0 2023-04-29 03:51:53,911 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91205.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:52:35,746 INFO [optim.py:368] (0/8) 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 03:52:37,293 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 03:53:27,240 INFO [train.py:904] (0/8) Epoch 9, batch 10050, loss[loss=0.1725, simple_loss=0.2637, pruned_loss=0.04062, over 16465.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2776, pruned_loss=0.04605, over 3075961.56 frames. ], batch size: 68, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:54:26,569 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 03:54:30,282 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 9, batch 10100, loss[loss=0.1602, simple_loss=0.2503, pruned_loss=0.03512, over 16764.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2779, pruned_loss=0.0465, over 3059867.81 frames. ], batch size: 83, lr: 7.40e-03, grad_scale: 8.0 2023-04-29 03:55:51,502 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9427, 3.1532, 3.2080, 2.1205, 2.9400, 3.2049, 3.0853, 1.7492], device='cuda:0'), covar=tensor([0.0380, 0.0037, 0.0042, 0.0305, 0.0081, 0.0065, 0.0064, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0062, 0.0064, 0.0121, 0.0072, 0.0079, 0.0071, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 03:55:52,062 INFO [optim.py:368] (0/8) 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 03:55:57,015 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9649, 4.0087, 3.7969, 3.5706, 3.5321, 3.9332, 3.6379, 3.6140], device='cuda:0'), covar=tensor([0.0502, 0.0460, 0.0300, 0.0279, 0.0849, 0.0403, 0.0881, 0.0578], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0253, 0.0245, 0.0225, 0.0262, 0.0255, 0.0168, 0.0283], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-29 03:56:13,317 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6793, 2.4412, 2.4155, 3.6357, 2.4907, 3.7204, 1.5182, 2.8538], device='cuda:0'), covar=tensor([0.1326, 0.0647, 0.1034, 0.0108, 0.0100, 0.0338, 0.1406, 0.0653], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0149, 0.0173, 0.0121, 0.0176, 0.0199, 0.0174, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 03:56:14,394 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 03:56:21,173 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-9.pt 2023-04-29 03:56:45,262 INFO [train.py:904] (0/8) Epoch 10, batch 0, loss[loss=0.2403, simple_loss=0.3226, pruned_loss=0.07902, over 16622.00 frames. ], tot_loss[loss=0.2403, simple_loss=0.3226, pruned_loss=0.07902, over 16622.00 frames. ], batch size: 62, lr: 7.04e-03, grad_scale: 8.0 2023-04-29 03:56:45,263 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 03:56:50,359 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9824, 5.2983, 5.1051, 5.1886, 4.8885, 4.9245, 4.7259, 5.3346], device='cuda:0'), covar=tensor([0.0818, 0.0679, 0.0690, 0.0509, 0.0780, 0.0306, 0.0830, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0467, 0.0596, 0.0485, 0.0407, 0.0373, 0.0388, 0.0496, 0.0443], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:56:52,895 INFO [train.py:938] (0/8) Epoch 10, validation: loss=0.158, simple_loss=0.2614, pruned_loss=0.02732, over 944034.00 frames. 2023-04-29 03:56:52,896 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 03:57:03,870 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2099, 4.2152, 4.3876, 4.1607, 4.2445, 4.7558, 4.4447, 4.0701], device='cuda:0'), covar=tensor([0.1487, 0.1942, 0.1545, 0.1980, 0.2563, 0.1098, 0.1329, 0.2306], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0438, 0.0465, 0.0385, 0.0501, 0.0488, 0.0381, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-29 03:57:38,010 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1257, 5.6668, 5.7731, 5.5973, 5.7339, 6.1546, 5.6680, 5.5807], device='cuda:0'), covar=tensor([0.0747, 0.1811, 0.1769, 0.1786, 0.2414, 0.0910, 0.1318, 0.2113], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0447, 0.0472, 0.0393, 0.0513, 0.0497, 0.0387, 0.0516], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 03:58:02,514 INFO [train.py:904] (0/8) Epoch 10, batch 50, loss[loss=0.1755, simple_loss=0.2543, pruned_loss=0.04839, over 16868.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.2964, pruned_loss=0.07001, over 743348.18 frames. ], batch size: 42, lr: 7.04e-03, grad_scale: 2.0 2023-04-29 03:58:39,941 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.984e+02 3.603e+02 4.569e+02 8.591e+02, threshold=7.207e+02, percent-clipped=1.0 2023-04-29 03:59:08,810 INFO [train.py:904] (0/8) Epoch 10, batch 100, loss[loss=0.2094, simple_loss=0.2834, pruned_loss=0.06769, over 16694.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.289, pruned_loss=0.06587, over 1307055.60 frames. ], batch size: 83, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 03:59:38,008 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 03:59:58,503 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2265, 4.1575, 4.6972, 4.6600, 4.6736, 4.2759, 4.3478, 4.1997], device='cuda:0'), covar=tensor([0.0352, 0.0568, 0.0369, 0.0456, 0.0450, 0.0404, 0.0844, 0.0465], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0303, 0.0307, 0.0293, 0.0342, 0.0325, 0.0417, 0.0261], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 04:00:16,895 INFO [train.py:904] (0/8) Epoch 10, batch 150, loss[loss=0.2406, simple_loss=0.3014, pruned_loss=0.08987, over 16725.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2853, pruned_loss=0.06444, over 1749245.22 frames. ], batch size: 134, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:00:22,866 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:00:56,396 INFO [optim.py:368] (0/8) 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 04:01:12,448 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 200, loss[loss=0.216, simple_loss=0.2849, pruned_loss=0.0735, over 16921.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.2842, pruned_loss=0.06361, over 2093954.00 frames. ], batch size: 109, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:01:28,069 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:01:44,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2232, 3.2393, 3.4430, 2.3881, 3.2546, 3.5590, 3.2782, 1.7597], device='cuda:0'), covar=tensor([0.0381, 0.0149, 0.0054, 0.0294, 0.0112, 0.0083, 0.0086, 0.0458], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0068, 0.0066, 0.0124, 0.0075, 0.0081, 0.0074, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:02:34,681 INFO [train.py:904] (0/8) Epoch 10, batch 250, loss[loss=0.1727, simple_loss=0.2567, pruned_loss=0.04437, over 15921.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2822, pruned_loss=0.06261, over 2365341.22 frames. ], batch size: 35, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:02:36,505 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91603.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:02:38,043 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 04:03:11,343 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.504e+02 3.085e+02 3.691e+02 6.376e+02, threshold=6.169e+02, percent-clipped=0.0 2023-04-29 04:03:28,034 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91641.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:03:37,121 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2346, 2.1985, 1.7590, 1.9352, 2.5526, 2.3305, 2.6232, 2.6551], device='cuda:0'), covar=tensor([0.0121, 0.0242, 0.0345, 0.0321, 0.0151, 0.0244, 0.0165, 0.0177], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0198, 0.0194, 0.0194, 0.0195, 0.0197, 0.0196, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:03:42,043 INFO [train.py:904] (0/8) Epoch 10, batch 300, loss[loss=0.1652, simple_loss=0.2559, pruned_loss=0.03726, over 17230.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.279, pruned_loss=0.06045, over 2580827.54 frames. ], batch size: 44, lr: 7.03e-03, grad_scale: 1.0 2023-04-29 04:04:39,263 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0805, 1.8006, 2.3913, 2.9997, 2.8596, 3.3952, 2.3391, 3.2471], device='cuda:0'), covar=tensor([0.0149, 0.0364, 0.0231, 0.0180, 0.0181, 0.0111, 0.0290, 0.0094], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0167, 0.0151, 0.0152, 0.0161, 0.0118, 0.0167, 0.0104], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 04:04:40,881 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0570, 5.0160, 5.5429, 5.5108, 5.5315, 5.1288, 5.1509, 4.8797], device='cuda:0'), covar=tensor([0.0312, 0.0479, 0.0355, 0.0416, 0.0387, 0.0318, 0.0826, 0.0364], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0315, 0.0317, 0.0303, 0.0357, 0.0337, 0.0435, 0.0271], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 04:04:51,292 INFO [train.py:904] (0/8) Epoch 10, batch 350, loss[loss=0.1825, simple_loss=0.2678, pruned_loss=0.04862, over 17230.00 frames. ], tot_loss[loss=0.1953, simple_loss=0.275, pruned_loss=0.05777, over 2747555.28 frames. ], batch size: 45, lr: 7.02e-03, grad_scale: 1.0 2023-04-29 04:05:01,876 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 04:05:04,687 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5594, 3.8738, 3.8348, 2.1901, 3.1241, 2.6039, 3.9364, 3.9017], device='cuda:0'), covar=tensor([0.0225, 0.0655, 0.0520, 0.1579, 0.0693, 0.0883, 0.0537, 0.0920], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0135, 0.0155, 0.0140, 0.0132, 0.0124, 0.0133, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 04:05:25,374 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2959, 4.0773, 4.3429, 4.5082, 4.5918, 4.1538, 4.3871, 4.5620], device='cuda:0'), covar=tensor([0.1309, 0.0980, 0.1132, 0.0550, 0.0515, 0.1085, 0.2007, 0.0520], device='cuda:0'), in_proj_covar=tensor([0.0506, 0.0620, 0.0753, 0.0626, 0.0475, 0.0476, 0.0500, 0.0558], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:05:28,620 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.553e+02 2.384e+02 2.940e+02 3.585e+02 5.710e+02, threshold=5.881e+02, percent-clipped=0.0 2023-04-29 04:05:59,422 INFO [train.py:904] (0/8) Epoch 10, batch 400, loss[loss=0.1923, simple_loss=0.2622, pruned_loss=0.06118, over 16558.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2734, pruned_loss=0.05742, over 2877634.50 frames. ], batch size: 75, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:11,423 INFO [train.py:904] (0/8) Epoch 10, batch 450, loss[loss=0.1878, simple_loss=0.2808, pruned_loss=0.04746, over 17020.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2716, pruned_loss=0.05554, over 2972701.02 frames. ], batch size: 50, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:07:50,762 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.266e+02 2.904e+02 3.599e+02 6.333e+02, threshold=5.808e+02, percent-clipped=1.0 2023-04-29 04:08:10,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4893, 3.4914, 3.6514, 1.9415, 3.7466, 3.7627, 3.0261, 2.8050], device='cuda:0'), covar=tensor([0.0678, 0.0151, 0.0138, 0.1084, 0.0068, 0.0158, 0.0359, 0.0408], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0098, 0.0084, 0.0141, 0.0069, 0.0098, 0.0120, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 04:08:20,259 INFO [train.py:904] (0/8) Epoch 10, batch 500, loss[loss=0.1852, simple_loss=0.2752, pruned_loss=0.04761, over 17057.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2696, pruned_loss=0.05463, over 3052752.20 frames. ], batch size: 53, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:08:42,958 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 04:09:15,174 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91891.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:09:21,650 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1628, 4.1474, 2.4075, 4.7693, 2.9390, 4.6608, 2.2513, 3.4604], device='cuda:0'), covar=tensor([0.0175, 0.0311, 0.1464, 0.0140, 0.0849, 0.0367, 0.1604, 0.0579], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0164, 0.0187, 0.0121, 0.0167, 0.0201, 0.0194, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 04:09:23,885 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=91898.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:09:29,464 INFO [train.py:904] (0/8) Epoch 10, batch 550, loss[loss=0.183, simple_loss=0.2734, pruned_loss=0.04632, over 16625.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.269, pruned_loss=0.05417, over 3112813.00 frames. ], batch size: 57, lr: 7.02e-03, grad_scale: 2.0 2023-04-29 04:10:07,829 INFO [optim.py:368] (0/8) 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 04:10:22,775 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=91941.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:32,597 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2262, 3.4394, 3.9172, 2.7311, 3.6447, 3.9224, 3.6655, 2.0626], device='cuda:0'), covar=tensor([0.0424, 0.0222, 0.0033, 0.0267, 0.0057, 0.0064, 0.0072, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0069, 0.0066, 0.0124, 0.0075, 0.0083, 0.0075, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:10:38,349 INFO [train.py:904] (0/8) Epoch 10, batch 600, loss[loss=0.2033, simple_loss=0.2588, pruned_loss=0.07397, over 16859.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2684, pruned_loss=0.05481, over 3166487.85 frames. ], batch size: 102, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:10:38,830 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=91952.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:10:42,839 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91955.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:18,743 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=91981.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:11:30,055 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=91989.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:11:45,347 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-92000.pt 2023-04-29 04:11:51,305 INFO [train.py:904] (0/8) Epoch 10, batch 650, loss[loss=0.1952, simple_loss=0.2742, pruned_loss=0.05804, over 15475.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2664, pruned_loss=0.05392, over 3202140.28 frames. ], batch size: 191, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:11:56,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1483, 5.1499, 4.9293, 4.3840, 4.9404, 1.8764, 4.6883, 4.8448], device='cuda:0'), covar=tensor([0.0068, 0.0053, 0.0136, 0.0342, 0.0089, 0.2257, 0.0108, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0109, 0.0159, 0.0148, 0.0128, 0.0177, 0.0145, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:12:12,115 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:12:21,960 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2866, 3.2424, 3.4455, 2.5809, 3.1432, 3.5126, 3.2828, 2.1291], device='cuda:0'), covar=tensor([0.0362, 0.0088, 0.0045, 0.0246, 0.0083, 0.0079, 0.0066, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0069, 0.0067, 0.0124, 0.0076, 0.0083, 0.0075, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:12:31,807 INFO [optim.py:368] (0/8) 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 04:12:49,419 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92042.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:13:03,072 INFO [train.py:904] (0/8) Epoch 10, batch 700, loss[loss=0.1953, simple_loss=0.2597, pruned_loss=0.0655, over 16712.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.266, pruned_loss=0.05341, over 3237815.54 frames. ], batch size: 89, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:14:12,203 INFO [train.py:904] (0/8) Epoch 10, batch 750, loss[loss=0.1961, simple_loss=0.2723, pruned_loss=0.06, over 15356.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2667, pruned_loss=0.05325, over 3256430.07 frames. ], batch size: 190, lr: 7.01e-03, grad_scale: 2.0 2023-04-29 04:14:52,015 INFO [optim.py:368] (0/8) 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 04:15:01,151 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0711, 5.4733, 5.6058, 5.3742, 5.4230, 6.0043, 5.5498, 5.2672], device='cuda:0'), covar=tensor([0.0802, 0.1686, 0.1573, 0.2066, 0.2690, 0.1007, 0.1289, 0.2445], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0487, 0.0511, 0.0421, 0.0559, 0.0534, 0.0413, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:15:22,695 INFO [train.py:904] (0/8) Epoch 10, batch 800, loss[loss=0.1821, simple_loss=0.2694, pruned_loss=0.04737, over 17121.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2668, pruned_loss=0.05282, over 3271877.62 frames. ], batch size: 47, lr: 7.01e-03, grad_scale: 4.0 2023-04-29 04:16:27,571 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92198.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:16:32,820 INFO [train.py:904] (0/8) Epoch 10, batch 850, loss[loss=0.1857, simple_loss=0.2777, pruned_loss=0.04678, over 16777.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2661, pruned_loss=0.05296, over 3272146.20 frames. ], batch size: 57, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:17:10,139 INFO [optim.py:368] (0/8) 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 04:17:33,338 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92246.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:17:34,337 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:17:41,168 INFO [train.py:904] (0/8) Epoch 10, batch 900, loss[loss=0.1881, simple_loss=0.2586, pruned_loss=0.05879, over 16750.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2651, pruned_loss=0.05211, over 3279683.68 frames. ], batch size: 83, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:18:40,190 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 04:18:50,850 INFO [train.py:904] (0/8) Epoch 10, batch 950, loss[loss=0.1599, simple_loss=0.2436, pruned_loss=0.03812, over 16826.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2651, pruned_loss=0.05236, over 3286764.77 frames. ], batch size: 42, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:19:04,211 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92311.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:19:29,766 INFO [optim.py:368] (0/8) 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 04:19:39,512 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92337.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:19:58,801 INFO [train.py:904] (0/8) Epoch 10, batch 1000, loss[loss=0.2006, simple_loss=0.2658, pruned_loss=0.0677, over 12487.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2645, pruned_loss=0.05219, over 3295031.64 frames. ], batch size: 246, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:21:09,137 INFO [train.py:904] (0/8) Epoch 10, batch 1050, loss[loss=0.172, simple_loss=0.2614, pruned_loss=0.04124, over 17121.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2645, pruned_loss=0.052, over 3291244.71 frames. ], batch size: 47, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:21:34,057 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4158, 2.1899, 1.5707, 1.9977, 2.6506, 2.4925, 2.7707, 2.7600], device='cuda:0'), covar=tensor([0.0142, 0.0277, 0.0433, 0.0375, 0.0145, 0.0249, 0.0180, 0.0201], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0198, 0.0194, 0.0194, 0.0197, 0.0198, 0.0201, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:21:42,319 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2594, 5.6071, 5.3581, 5.4353, 5.0469, 4.9152, 5.0953, 5.6978], device='cuda:0'), covar=tensor([0.1027, 0.0831, 0.0950, 0.0616, 0.0794, 0.0734, 0.0965, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0666, 0.0548, 0.0458, 0.0416, 0.0425, 0.0560, 0.0503], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:21:48,371 INFO [optim.py:368] (0/8) 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 04:22:01,516 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1880, 3.3347, 3.5969, 2.4685, 3.2951, 3.5893, 3.4296, 2.0987], device='cuda:0'), covar=tensor([0.0357, 0.0095, 0.0033, 0.0255, 0.0079, 0.0065, 0.0057, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0069, 0.0067, 0.0125, 0.0077, 0.0084, 0.0075, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:22:07,291 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5143, 3.6695, 3.8723, 2.1303, 4.0386, 4.0856, 3.2356, 2.9469], device='cuda:0'), covar=tensor([0.0723, 0.0166, 0.0175, 0.1000, 0.0066, 0.0127, 0.0329, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0097, 0.0085, 0.0140, 0.0069, 0.0098, 0.0119, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 04:22:18,523 INFO [train.py:904] (0/8) Epoch 10, batch 1100, loss[loss=0.204, simple_loss=0.2696, pruned_loss=0.06915, over 16740.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2631, pruned_loss=0.05144, over 3297537.80 frames. ], batch size: 124, lr: 7.00e-03, grad_scale: 4.0 2023-04-29 04:22:48,274 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1680, 4.2041, 2.1422, 4.8153, 3.0358, 4.7834, 2.2580, 3.3132], device='cuda:0'), covar=tensor([0.0175, 0.0270, 0.1593, 0.0136, 0.0699, 0.0251, 0.1579, 0.0600], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0163, 0.0186, 0.0123, 0.0165, 0.0202, 0.0191, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 04:22:52,062 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 04:23:02,215 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 04:23:28,223 INFO [train.py:904] (0/8) Epoch 10, batch 1150, loss[loss=0.1817, simple_loss=0.277, pruned_loss=0.04318, over 17205.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.263, pruned_loss=0.05031, over 3306248.61 frames. ], batch size: 44, lr: 6.99e-03, grad_scale: 4.0 2023-04-29 04:23:52,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6343, 4.6552, 5.1014, 5.0424, 5.0944, 4.7320, 4.6723, 4.5048], device='cuda:0'), covar=tensor([0.0312, 0.0516, 0.0392, 0.0470, 0.0387, 0.0348, 0.0862, 0.0445], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0339, 0.0337, 0.0325, 0.0380, 0.0356, 0.0463, 0.0285], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 04:24:08,394 INFO [optim.py:368] (0/8) 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 04:24:25,707 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7645, 3.8936, 3.0011, 2.1912, 2.6683, 2.2069, 3.9001, 3.6920], device='cuda:0'), covar=tensor([0.2295, 0.0537, 0.1379, 0.2451, 0.2172, 0.1762, 0.0482, 0.0993], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0257, 0.0281, 0.0273, 0.0274, 0.0219, 0.0268, 0.0293], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:24:32,654 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:24:38,819 INFO [train.py:904] (0/8) Epoch 10, batch 1200, loss[loss=0.1664, simple_loss=0.2648, pruned_loss=0.03401, over 17133.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2625, pruned_loss=0.04979, over 3303560.13 frames. ], batch size: 46, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:25:05,518 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7890, 2.4994, 2.3457, 3.4740, 2.9238, 3.6555, 1.4664, 2.7333], device='cuda:0'), covar=tensor([0.1230, 0.0602, 0.1017, 0.0174, 0.0173, 0.0373, 0.1364, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0152, 0.0175, 0.0133, 0.0191, 0.0208, 0.0173, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 04:25:18,667 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 04:25:39,052 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92595.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:25:47,775 INFO [train.py:904] (0/8) Epoch 10, batch 1250, loss[loss=0.1511, simple_loss=0.2374, pruned_loss=0.03243, over 17207.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2631, pruned_loss=0.05101, over 3299692.95 frames. ], batch size: 44, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:25:49,279 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8116, 4.7615, 4.6745, 4.4012, 4.3101, 4.7539, 4.6494, 4.4692], device='cuda:0'), covar=tensor([0.0662, 0.0574, 0.0293, 0.0282, 0.1052, 0.0447, 0.0401, 0.0676], device='cuda:0'), in_proj_covar=tensor([0.0247, 0.0303, 0.0288, 0.0266, 0.0314, 0.0305, 0.0198, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:26:01,779 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:20,447 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:26:27,277 INFO [optim.py:368] (0/8) 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 04:26:37,459 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=92637.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 04:26:50,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9826, 1.7654, 2.2276, 2.8598, 2.6194, 3.4037, 2.1615, 3.2142], device='cuda:0'), covar=tensor([0.0154, 0.0361, 0.0246, 0.0193, 0.0219, 0.0114, 0.0306, 0.0099], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0170, 0.0155, 0.0158, 0.0166, 0.0121, 0.0171, 0.0111], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 04:26:58,811 INFO [train.py:904] (0/8) Epoch 10, batch 1300, loss[loss=0.191, simple_loss=0.2831, pruned_loss=0.04939, over 17253.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2635, pruned_loss=0.05111, over 3308978.52 frames. ], batch size: 52, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:27:07,421 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=92685.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:27:45,214 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:07,567 INFO [train.py:904] (0/8) Epoch 10, batch 1350, loss[loss=0.1946, simple_loss=0.2618, pruned_loss=0.06376, over 16878.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2636, pruned_loss=0.05081, over 3304818.80 frames. ], batch size: 96, lr: 6.99e-03, grad_scale: 8.0 2023-04-29 04:28:34,021 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:28:47,046 INFO [optim.py:368] (0/8) 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 04:29:18,982 INFO [train.py:904] (0/8) Epoch 10, batch 1400, loss[loss=0.1805, simple_loss=0.2528, pruned_loss=0.05405, over 16820.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2637, pruned_loss=0.05137, over 3305864.15 frames. ], batch size: 102, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:30:00,008 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:30:00,385 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 04:30:11,183 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8893, 4.9344, 5.5056, 5.4574, 5.4782, 5.1405, 5.0634, 4.7638], device='cuda:0'), covar=tensor([0.0357, 0.0472, 0.0324, 0.0460, 0.0385, 0.0285, 0.0863, 0.0422], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0340, 0.0341, 0.0325, 0.0382, 0.0356, 0.0466, 0.0286], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 04:30:27,024 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4611, 3.4792, 3.8958, 2.6465, 3.4008, 3.8205, 3.7171, 2.3444], device='cuda:0'), covar=tensor([0.0365, 0.0145, 0.0038, 0.0285, 0.0077, 0.0069, 0.0060, 0.0327], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0070, 0.0068, 0.0125, 0.0076, 0.0084, 0.0076, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:30:28,853 INFO [train.py:904] (0/8) Epoch 10, batch 1450, loss[loss=0.1946, simple_loss=0.261, pruned_loss=0.06407, over 16682.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.263, pruned_loss=0.05162, over 3309314.96 frames. ], batch size: 89, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:30:38,754 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5311, 3.7192, 3.8375, 1.9361, 3.9944, 4.0044, 3.1670, 2.8738], device='cuda:0'), covar=tensor([0.0743, 0.0137, 0.0157, 0.1122, 0.0073, 0.0130, 0.0354, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0098, 0.0087, 0.0141, 0.0070, 0.0100, 0.0121, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 04:30:47,558 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7553, 2.8190, 2.7005, 4.6031, 3.9049, 4.3691, 1.5228, 3.1600], device='cuda:0'), covar=tensor([0.1393, 0.0720, 0.1204, 0.0233, 0.0287, 0.0442, 0.1553, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0156, 0.0179, 0.0137, 0.0196, 0.0213, 0.0179, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 04:31:07,981 INFO [optim.py:368] (0/8) 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 04:31:09,644 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:31:30,555 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3949, 4.1051, 3.9101, 4.5248, 4.6282, 4.2166, 4.3613, 4.5626], device='cuda:0'), covar=tensor([0.1141, 0.1050, 0.2234, 0.0858, 0.0758, 0.1157, 0.1595, 0.0894], device='cuda:0'), in_proj_covar=tensor([0.0537, 0.0659, 0.0807, 0.0673, 0.0501, 0.0506, 0.0526, 0.0596], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:31:38,230 INFO [train.py:904] (0/8) Epoch 10, batch 1500, loss[loss=0.2291, simple_loss=0.2901, pruned_loss=0.08411, over 16871.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2626, pruned_loss=0.05136, over 3306400.04 frames. ], batch size: 116, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:32:00,080 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 04:32:34,087 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92892.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:45,032 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:32:48,550 INFO [train.py:904] (0/8) Epoch 10, batch 1550, loss[loss=0.2245, simple_loss=0.3029, pruned_loss=0.07305, over 15389.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2642, pruned_loss=0.0526, over 3301854.02 frames. ], batch size: 190, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:33:26,143 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.809e+02 2.944e+02 3.490e+02 4.097e+02 8.318e+02, threshold=6.980e+02, percent-clipped=5.0 2023-04-29 04:33:27,751 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92931.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:33:56,443 INFO [train.py:904] (0/8) Epoch 10, batch 1600, loss[loss=0.1783, simple_loss=0.2542, pruned_loss=0.05121, over 16748.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2661, pruned_loss=0.0535, over 3299561.87 frames. ], batch size: 102, lr: 6.98e-03, grad_scale: 8.0 2023-04-29 04:34:07,718 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:29,006 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=92975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:37,305 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=92981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:34:52,627 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=92992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:06,233 INFO [train.py:904] (0/8) Epoch 10, batch 1650, loss[loss=0.1849, simple_loss=0.2582, pruned_loss=0.05575, over 16725.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2672, pruned_loss=0.05378, over 3305703.84 frames. ], batch size: 89, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:35:18,212 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93010.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:45,201 INFO [optim.py:368] (0/8) 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 04:35:53,057 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:35:53,102 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:15,632 INFO [train.py:904] (0/8) Epoch 10, batch 1700, loss[loss=0.2033, simple_loss=0.2808, pruned_loss=0.06296, over 16781.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2687, pruned_loss=0.05446, over 3306074.42 frames. ], batch size: 83, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:36:42,609 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93071.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:36:46,280 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 04:36:48,270 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:37:17,867 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93097.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 04:37:23,702 INFO [train.py:904] (0/8) Epoch 10, batch 1750, loss[loss=0.1992, simple_loss=0.2939, pruned_loss=0.05229, over 17254.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2699, pruned_loss=0.0544, over 3307327.44 frames. ], batch size: 52, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:37:42,786 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:38:01,416 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 2.469e+02 2.889e+02 3.646e+02 7.131e+02, threshold=5.778e+02, percent-clipped=4.0 2023-04-29 04:38:32,570 INFO [train.py:904] (0/8) Epoch 10, batch 1800, loss[loss=0.1762, simple_loss=0.2576, pruned_loss=0.04746, over 16849.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2712, pruned_loss=0.05456, over 3317145.92 frames. ], batch size: 42, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:38:58,696 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 04:39:06,173 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:20,455 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93187.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:39:42,554 INFO [train.py:904] (0/8) Epoch 10, batch 1850, loss[loss=0.1917, simple_loss=0.2702, pruned_loss=0.05661, over 16896.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2722, pruned_loss=0.05497, over 3315869.51 frames. ], batch size: 96, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:39:48,630 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 04:40:21,085 INFO [optim.py:368] (0/8) 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 04:40:52,062 INFO [train.py:904] (0/8) Epoch 10, batch 1900, loss[loss=0.1648, simple_loss=0.2495, pruned_loss=0.04007, over 16998.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2715, pruned_loss=0.05423, over 3320336.32 frames. ], batch size: 41, lr: 6.97e-03, grad_scale: 8.0 2023-04-29 04:40:56,699 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:25,827 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4949, 3.9986, 3.9933, 2.1626, 3.2921, 2.5919, 3.8731, 3.9991], device='cuda:0'), covar=tensor([0.0291, 0.0644, 0.0434, 0.1609, 0.0662, 0.0833, 0.0610, 0.0980], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0141, 0.0156, 0.0141, 0.0134, 0.0124, 0.0136, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 04:41:33,309 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:41:41,189 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:42:02,729 INFO [train.py:904] (0/8) Epoch 10, batch 1950, loss[loss=0.2014, simple_loss=0.287, pruned_loss=0.05787, over 16698.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2709, pruned_loss=0.05329, over 3320268.56 frames. ], batch size: 57, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:42:06,554 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 04:42:39,966 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:42:41,421 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.458e+02 3.015e+02 3.561e+02 8.313e+02, threshold=6.031e+02, percent-clipped=4.0 2023-04-29 04:42:42,938 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93331.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:12,559 INFO [train.py:904] (0/8) Epoch 10, batch 2000, loss[loss=0.1872, simple_loss=0.2776, pruned_loss=0.04844, over 17071.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.2696, pruned_loss=0.05189, over 3328835.20 frames. ], batch size: 53, lr: 6.96e-03, grad_scale: 8.0 2023-04-29 04:43:31,565 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:43:44,503 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:44:08,383 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93392.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 04:44:16,273 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8800, 4.8231, 4.7581, 4.2907, 4.8222, 1.9794, 4.5860, 4.6527], device='cuda:0'), covar=tensor([0.0080, 0.0075, 0.0128, 0.0259, 0.0073, 0.2084, 0.0103, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0117, 0.0168, 0.0158, 0.0137, 0.0181, 0.0155, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:44:21,496 INFO [train.py:904] (0/8) Epoch 10, batch 2050, loss[loss=0.1896, simple_loss=0.2862, pruned_loss=0.04657, over 17247.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2702, pruned_loss=0.05241, over 3318751.61 frames. ], batch size: 52, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:44:31,756 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2601, 5.6248, 5.3439, 5.3922, 5.0418, 4.9284, 5.0811, 5.7646], device='cuda:0'), covar=tensor([0.1022, 0.0863, 0.1051, 0.0625, 0.0876, 0.0771, 0.1055, 0.0869], device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0681, 0.0561, 0.0467, 0.0429, 0.0435, 0.0568, 0.0516], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:44:50,365 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6475, 4.5913, 4.6088, 4.1308, 4.5967, 1.8848, 4.3948, 4.4322], device='cuda:0'), covar=tensor([0.0106, 0.0095, 0.0132, 0.0273, 0.0080, 0.2064, 0.0130, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0117, 0.0168, 0.0158, 0.0137, 0.0180, 0.0154, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:44:51,423 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93424.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:45:00,992 INFO [optim.py:368] (0/8) 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 04:45:29,951 INFO [train.py:904] (0/8) Epoch 10, batch 2100, loss[loss=0.2421, simple_loss=0.3178, pruned_loss=0.08317, over 12021.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.272, pruned_loss=0.05344, over 3310991.36 frames. ], batch size: 246, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:45:38,990 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 04:45:56,196 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=93471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:18,663 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93487.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:46:40,134 INFO [train.py:904] (0/8) Epoch 10, batch 2150, loss[loss=0.187, simple_loss=0.2653, pruned_loss=0.05432, over 16478.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2723, pruned_loss=0.05329, over 3313002.73 frames. ], batch size: 75, lr: 6.96e-03, grad_scale: 4.0 2023-04-29 04:47:18,312 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.473e+02 3.061e+02 3.473e+02 5.653e+02, threshold=6.122e+02, percent-clipped=0.0 2023-04-29 04:47:24,639 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:47:46,714 INFO [train.py:904] (0/8) Epoch 10, batch 2200, loss[loss=0.2778, simple_loss=0.3456, pruned_loss=0.105, over 11955.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2734, pruned_loss=0.05406, over 3312359.39 frames. ], batch size: 247, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:47:52,041 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:34,793 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93587.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:48:52,090 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0351, 3.9616, 4.4462, 4.4266, 4.4381, 4.1102, 4.1563, 4.0330], device='cuda:0'), covar=tensor([0.0348, 0.0628, 0.0415, 0.0446, 0.0449, 0.0362, 0.0856, 0.0567], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0344, 0.0349, 0.0330, 0.0389, 0.0361, 0.0470, 0.0290], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 04:48:54,030 INFO [train.py:904] (0/8) Epoch 10, batch 2250, loss[loss=0.1579, simple_loss=0.2394, pruned_loss=0.03824, over 17021.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2741, pruned_loss=0.05476, over 3311665.22 frames. ], batch size: 41, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:48:55,412 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:49:03,795 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-29 04:49:33,830 INFO [optim.py:368] (0/8) 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 04:49:34,285 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93635.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:04,002 INFO [train.py:904] (0/8) Epoch 10, batch 2300, loss[loss=0.1623, simple_loss=0.2571, pruned_loss=0.03371, over 17126.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2739, pruned_loss=0.05438, over 3308887.25 frames. ], batch size: 47, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:50:11,560 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5612, 4.4567, 4.4301, 4.2189, 4.1211, 4.4913, 4.2631, 4.1798], device='cuda:0'), covar=tensor([0.0572, 0.0586, 0.0254, 0.0243, 0.0753, 0.0438, 0.0512, 0.0642], device='cuda:0'), in_proj_covar=tensor([0.0254, 0.0310, 0.0296, 0.0274, 0.0323, 0.0310, 0.0205, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 04:50:22,600 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:50:57,946 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:09,779 INFO [train.py:904] (0/8) Epoch 10, batch 2350, loss[loss=0.1831, simple_loss=0.261, pruned_loss=0.05253, over 16783.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2741, pruned_loss=0.05479, over 3319074.19 frames. ], batch size: 39, lr: 6.95e-03, grad_scale: 4.0 2023-04-29 04:51:27,717 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93714.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:51:40,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0243, 4.3751, 3.3158, 2.4260, 3.0171, 2.6177, 4.6956, 4.1175], device='cuda:0'), covar=tensor([0.2187, 0.0540, 0.1385, 0.2071, 0.2268, 0.1594, 0.0333, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0259, 0.0283, 0.0273, 0.0280, 0.0220, 0.0265, 0.0296], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:51:49,912 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.552e+02 2.363e+02 2.805e+02 3.305e+02 9.718e+02, threshold=5.610e+02, percent-clipped=1.0 2023-04-29 04:52:03,151 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:15,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4217, 2.0102, 2.1530, 4.1156, 2.0533, 2.5111, 2.0810, 2.2958], device='cuda:0'), covar=tensor([0.0946, 0.3201, 0.2075, 0.0402, 0.3268, 0.2070, 0.3084, 0.2609], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0386, 0.0325, 0.0325, 0.0408, 0.0440, 0.0347, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 04:52:17,489 INFO [train.py:904] (0/8) Epoch 10, batch 2400, loss[loss=0.2188, simple_loss=0.2937, pruned_loss=0.07191, over 15444.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2742, pruned_loss=0.0545, over 3326134.06 frames. ], batch size: 191, lr: 6.95e-03, grad_scale: 8.0 2023-04-29 04:52:36,428 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8985, 3.2341, 2.7879, 5.0821, 4.3707, 4.6816, 1.6112, 3.4198], device='cuda:0'), covar=tensor([0.1250, 0.0571, 0.1074, 0.0118, 0.0238, 0.0328, 0.1425, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0156, 0.0178, 0.0137, 0.0199, 0.0212, 0.0176, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 04:52:40,865 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:52:42,621 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=93771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:53:26,817 INFO [train.py:904] (0/8) Epoch 10, batch 2450, loss[loss=0.2033, simple_loss=0.2754, pruned_loss=0.06555, over 16788.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2743, pruned_loss=0.05428, over 3327781.39 frames. ], batch size: 102, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:53:49,969 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=93819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:05,074 INFO [optim.py:368] (0/8) 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 04:54:05,472 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=93831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:54:34,543 INFO [train.py:904] (0/8) Epoch 10, batch 2500, loss[loss=0.2208, simple_loss=0.2927, pruned_loss=0.07447, over 16390.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2735, pruned_loss=0.05432, over 3315006.01 frames. ], batch size: 146, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:54:35,205 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 04:54:51,506 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.58 vs. limit=2.0 2023-04-29 04:55:00,638 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9713, 1.7484, 2.3289, 2.8052, 2.6097, 3.3930, 1.9186, 3.1750], device='cuda:0'), covar=tensor([0.0168, 0.0368, 0.0234, 0.0214, 0.0209, 0.0120, 0.0349, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0155, 0.0160, 0.0165, 0.0122, 0.0170, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 04:55:30,429 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1503, 1.9540, 2.4289, 2.9032, 2.7448, 3.4927, 2.2279, 3.3344], device='cuda:0'), covar=tensor([0.0134, 0.0330, 0.0217, 0.0203, 0.0188, 0.0098, 0.0279, 0.0097], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0155, 0.0159, 0.0165, 0.0121, 0.0170, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 04:55:43,659 INFO [train.py:904] (0/8) Epoch 10, batch 2550, loss[loss=0.1714, simple_loss=0.2639, pruned_loss=0.03947, over 17217.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2742, pruned_loss=0.0539, over 3317429.97 frames. ], batch size: 44, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:56:23,963 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.408e+02 2.927e+02 3.582e+02 7.180e+02, threshold=5.854e+02, percent-clipped=2.0 2023-04-29 04:56:52,859 INFO [train.py:904] (0/8) Epoch 10, batch 2600, loss[loss=0.1891, simple_loss=0.2858, pruned_loss=0.04626, over 17045.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2737, pruned_loss=0.05376, over 3323446.39 frames. ], batch size: 50, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:57:32,258 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=93981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:57:57,916 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-94000.pt 2023-04-29 04:58:03,853 INFO [train.py:904] (0/8) Epoch 10, batch 2650, loss[loss=0.2205, simple_loss=0.321, pruned_loss=0.06, over 17116.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2749, pruned_loss=0.0539, over 3325288.74 frames. ], batch size: 49, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 04:58:43,804 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.229e+02 2.744e+02 3.271e+02 8.724e+02, threshold=5.488e+02, percent-clipped=1.0 2023-04-29 04:58:50,782 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.08 vs. limit=5.0 2023-04-29 04:59:00,325 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94042.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 04:59:13,560 INFO [train.py:904] (0/8) Epoch 10, batch 2700, loss[loss=0.1681, simple_loss=0.2506, pruned_loss=0.04282, over 17011.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2743, pruned_loss=0.05276, over 3331564.83 frames. ], batch size: 41, lr: 6.94e-03, grad_scale: 8.0 2023-04-29 05:00:23,047 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-04-29 05:00:23,307 INFO [train.py:904] (0/8) Epoch 10, batch 2750, loss[loss=0.1724, simple_loss=0.2616, pruned_loss=0.04154, over 17136.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2742, pruned_loss=0.05254, over 3337255.23 frames. ], batch size: 48, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:00:25,287 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-29 05:00:55,388 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94126.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:01:01,090 INFO [optim.py:368] (0/8) 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 05:01:10,576 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7795, 3.8452, 2.9917, 2.2013, 2.7203, 2.2829, 3.9215, 3.5209], device='cuda:0'), covar=tensor([0.2095, 0.0544, 0.1258, 0.2251, 0.2024, 0.1641, 0.0411, 0.1059], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0256, 0.0281, 0.0271, 0.0279, 0.0218, 0.0265, 0.0295], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:01:29,699 INFO [train.py:904] (0/8) Epoch 10, batch 2800, loss[loss=0.1682, simple_loss=0.2576, pruned_loss=0.03943, over 16783.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2735, pruned_loss=0.05227, over 3335094.01 frames. ], batch size: 39, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:01:58,409 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7983, 2.1302, 2.1211, 4.4738, 1.9833, 2.6135, 2.2346, 2.2701], device='cuda:0'), covar=tensor([0.0817, 0.3309, 0.2247, 0.0344, 0.3987, 0.2311, 0.2913, 0.3424], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0385, 0.0324, 0.0322, 0.0406, 0.0438, 0.0344, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:02:24,209 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4632, 2.0984, 2.3127, 4.1977, 2.1843, 2.5828, 2.1902, 2.3670], device='cuda:0'), covar=tensor([0.0999, 0.3262, 0.2086, 0.0382, 0.3294, 0.2169, 0.3034, 0.2624], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0386, 0.0325, 0.0323, 0.0407, 0.0440, 0.0345, 0.0456], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:02:39,414 INFO [train.py:904] (0/8) Epoch 10, batch 2850, loss[loss=0.1641, simple_loss=0.249, pruned_loss=0.03955, over 16831.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2725, pruned_loss=0.05172, over 3335580.14 frames. ], batch size: 39, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:03:00,011 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0841, 5.9112, 6.0411, 5.7070, 5.8589, 6.2733, 5.7466, 5.4801], device='cuda:0'), covar=tensor([0.0907, 0.1487, 0.1439, 0.1622, 0.2076, 0.0780, 0.1313, 0.2280], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0494, 0.0518, 0.0429, 0.0569, 0.0548, 0.0418, 0.0576], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 05:03:08,994 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94223.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:03:20,116 INFO [optim.py:368] (0/8) 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 05:03:49,051 INFO [train.py:904] (0/8) Epoch 10, batch 2900, loss[loss=0.2096, simple_loss=0.2904, pruned_loss=0.06437, over 16813.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2724, pruned_loss=0.05264, over 3331767.37 frames. ], batch size: 96, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:04:33,819 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94284.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:04:58,059 INFO [train.py:904] (0/8) Epoch 10, batch 2950, loss[loss=0.1682, simple_loss=0.2541, pruned_loss=0.04116, over 16709.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2719, pruned_loss=0.05354, over 3323290.55 frames. ], batch size: 39, lr: 6.93e-03, grad_scale: 8.0 2023-04-29 05:05:39,546 INFO [optim.py:368] (0/8) 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 05:05:49,265 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94337.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:03,081 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=94348.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:06:08,007 INFO [train.py:904] (0/8) Epoch 10, batch 3000, loss[loss=0.1905, simple_loss=0.2662, pruned_loss=0.05738, over 16893.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2717, pruned_loss=0.05401, over 3317848.67 frames. ], batch size: 96, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:06:08,008 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 05:06:13,446 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7923, 4.3295, 3.4332, 2.2907, 2.7813, 2.5354, 4.5576, 3.6338], device='cuda:0'), covar=tensor([0.2365, 0.0411, 0.1122, 0.2283, 0.2399, 0.1608, 0.0237, 0.1000], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0257, 0.0280, 0.0272, 0.0279, 0.0219, 0.0267, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:06:17,144 INFO [train.py:938] (0/8) Epoch 10, validation: loss=0.1426, simple_loss=0.2488, pruned_loss=0.01818, over 944034.00 frames. 2023-04-29 05:06:17,145 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 05:07:03,372 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1932, 4.1811, 4.6399, 4.6350, 4.6504, 4.2860, 4.3635, 4.1353], device='cuda:0'), covar=tensor([0.0380, 0.0521, 0.0347, 0.0421, 0.0417, 0.0337, 0.0697, 0.0595], device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0356, 0.0360, 0.0341, 0.0400, 0.0374, 0.0484, 0.0299], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 05:07:26,697 INFO [train.py:904] (0/8) Epoch 10, batch 3050, loss[loss=0.1964, simple_loss=0.263, pruned_loss=0.06493, over 16849.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2725, pruned_loss=0.05473, over 3316471.74 frames. ], batch size: 116, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:07:36,840 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=94409.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:07:57,752 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94426.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:08:05,350 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.835e+02 2.667e+02 3.252e+02 4.083e+02 7.974e+02, threshold=6.505e+02, percent-clipped=3.0 2023-04-29 05:08:12,080 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5097, 2.9689, 2.5680, 4.8397, 4.0249, 4.4524, 1.6846, 3.1510], device='cuda:0'), covar=tensor([0.1401, 0.0652, 0.1204, 0.0186, 0.0286, 0.0363, 0.1388, 0.0778], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0157, 0.0179, 0.0139, 0.0200, 0.0214, 0.0176, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 05:08:33,257 INFO [train.py:904] (0/8) Epoch 10, batch 3100, loss[loss=0.1723, simple_loss=0.2691, pruned_loss=0.03779, over 17100.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2714, pruned_loss=0.05359, over 3319544.52 frames. ], batch size: 49, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:09:04,484 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:09:43,511 INFO [train.py:904] (0/8) Epoch 10, batch 3150, loss[loss=0.1977, simple_loss=0.2765, pruned_loss=0.05946, over 15494.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2711, pruned_loss=0.05318, over 3323529.78 frames. ], batch size: 190, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:09:51,101 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5022, 2.0585, 2.2505, 4.1770, 2.1652, 2.6111, 2.1606, 2.3145], device='cuda:0'), covar=tensor([0.0935, 0.3232, 0.2029, 0.0393, 0.3370, 0.2029, 0.2924, 0.2717], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0390, 0.0328, 0.0327, 0.0413, 0.0445, 0.0350, 0.0462], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:10:23,690 INFO [optim.py:368] (0/8) 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 05:10:26,497 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5995, 3.5744, 3.8483, 1.8638, 4.0707, 4.0172, 3.1965, 2.9098], device='cuda:0'), covar=tensor([0.0627, 0.0182, 0.0136, 0.1053, 0.0046, 0.0123, 0.0303, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0100, 0.0088, 0.0138, 0.0070, 0.0102, 0.0121, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 05:10:52,169 INFO [train.py:904] (0/8) Epoch 10, batch 3200, loss[loss=0.1737, simple_loss=0.2533, pruned_loss=0.04701, over 16843.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2694, pruned_loss=0.05244, over 3319964.99 frames. ], batch size: 96, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:11:32,217 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94579.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:11:55,496 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4363, 2.2358, 2.3089, 4.3258, 2.1685, 2.7781, 2.3065, 2.4759], device='cuda:0'), covar=tensor([0.0945, 0.3078, 0.2069, 0.0365, 0.3453, 0.1976, 0.2762, 0.2779], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0389, 0.0328, 0.0326, 0.0412, 0.0444, 0.0349, 0.0461], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:12:04,580 INFO [train.py:904] (0/8) Epoch 10, batch 3250, loss[loss=0.156, simple_loss=0.2412, pruned_loss=0.03537, over 16792.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2703, pruned_loss=0.05308, over 3319056.83 frames. ], batch size: 39, lr: 6.92e-03, grad_scale: 8.0 2023-04-29 05:12:17,758 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.36 vs. limit=5.0 2023-04-29 05:12:44,906 INFO [optim.py:368] (0/8) 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 05:12:53,079 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94637.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:13:14,625 INFO [train.py:904] (0/8) Epoch 10, batch 3300, loss[loss=0.1968, simple_loss=0.2805, pruned_loss=0.05653, over 16438.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2721, pruned_loss=0.05393, over 3319559.87 frames. ], batch size: 68, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:01,719 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:14:21,438 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9714, 3.9562, 3.8893, 3.0181, 3.9144, 1.8179, 3.6508, 3.3831], device='cuda:0'), covar=tensor([0.0142, 0.0110, 0.0162, 0.0470, 0.0112, 0.2747, 0.0158, 0.0319], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0120, 0.0169, 0.0163, 0.0139, 0.0181, 0.0158, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:14:24,569 INFO [train.py:904] (0/8) Epoch 10, batch 3350, loss[loss=0.1464, simple_loss=0.2341, pruned_loss=0.02937, over 16766.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2719, pruned_loss=0.05367, over 3307212.60 frames. ], batch size: 39, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:14:28,547 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=94704.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:15:00,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6529, 4.7058, 4.5640, 4.3841, 4.0220, 4.6832, 4.5116, 4.2343], device='cuda:0'), covar=tensor([0.0655, 0.0526, 0.0303, 0.0277, 0.1202, 0.0419, 0.0448, 0.0675], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0317, 0.0300, 0.0278, 0.0327, 0.0316, 0.0207, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 05:15:05,006 INFO [optim.py:368] (0/8) 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 05:15:35,795 INFO [train.py:904] (0/8) Epoch 10, batch 3400, loss[loss=0.1475, simple_loss=0.2317, pruned_loss=0.03162, over 16747.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2723, pruned_loss=0.05387, over 3310410.98 frames. ], batch size: 39, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:16:44,921 INFO [train.py:904] (0/8) Epoch 10, batch 3450, loss[loss=0.1975, simple_loss=0.2808, pruned_loss=0.05707, over 16525.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2715, pruned_loss=0.05376, over 3313164.62 frames. ], batch size: 68, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:16:57,659 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8324, 3.7262, 3.8293, 3.9893, 4.0697, 3.6088, 3.9159, 4.0595], device='cuda:0'), covar=tensor([0.1132, 0.0877, 0.1263, 0.0619, 0.0556, 0.2042, 0.1390, 0.0585], device='cuda:0'), in_proj_covar=tensor([0.0545, 0.0671, 0.0831, 0.0686, 0.0510, 0.0528, 0.0538, 0.0605], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:17:26,309 INFO [optim.py:368] (0/8) 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 05:17:53,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8342, 5.1008, 5.3028, 5.1164, 5.0544, 5.7265, 5.2651, 4.9641], device='cuda:0'), covar=tensor([0.1098, 0.1944, 0.1592, 0.1998, 0.3348, 0.1036, 0.1306, 0.2483], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0498, 0.0525, 0.0432, 0.0573, 0.0551, 0.0424, 0.0585], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 05:17:56,546 INFO [train.py:904] (0/8) Epoch 10, batch 3500, loss[loss=0.183, simple_loss=0.2635, pruned_loss=0.05118, over 16826.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2694, pruned_loss=0.05239, over 3311970.25 frames. ], batch size: 90, lr: 6.91e-03, grad_scale: 8.0 2023-04-29 05:18:07,773 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-29 05:18:35,794 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=94879.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:19:06,953 INFO [train.py:904] (0/8) Epoch 10, batch 3550, loss[loss=0.2443, simple_loss=0.3145, pruned_loss=0.087, over 11772.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2685, pruned_loss=0.05151, over 3321431.77 frames. ], batch size: 247, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:19:41,449 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=94927.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:19:47,560 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.230e+02 2.675e+02 3.251e+02 5.912e+02, threshold=5.350e+02, percent-clipped=0.0 2023-04-29 05:20:10,516 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 05:20:17,540 INFO [train.py:904] (0/8) Epoch 10, batch 3600, loss[loss=0.1906, simple_loss=0.2615, pruned_loss=0.0598, over 16675.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2674, pruned_loss=0.05145, over 3307389.18 frames. ], batch size: 76, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:02,135 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-04-29 05:21:28,701 INFO [train.py:904] (0/8) Epoch 10, batch 3650, loss[loss=0.2039, simple_loss=0.2618, pruned_loss=0.07305, over 16875.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2663, pruned_loss=0.05232, over 3304795.86 frames. ], batch size: 90, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:21:32,955 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95004.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:22:10,216 INFO [optim.py:368] (0/8) 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 05:22:41,715 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4358, 3.5515, 3.0574, 2.9062, 3.0614, 3.3637, 3.2170, 3.1229], device='cuda:0'), covar=tensor([0.0506, 0.0437, 0.0254, 0.0265, 0.0571, 0.0341, 0.1379, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0314, 0.0298, 0.0277, 0.0325, 0.0313, 0.0205, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 05:22:43,032 INFO [train.py:904] (0/8) Epoch 10, batch 3700, loss[loss=0.1998, simple_loss=0.2673, pruned_loss=0.06612, over 11188.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2655, pruned_loss=0.05402, over 3288649.33 frames. ], batch size: 246, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:22:43,315 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95052.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:23:56,105 INFO [train.py:904] (0/8) Epoch 10, batch 3750, loss[loss=0.2225, simple_loss=0.3101, pruned_loss=0.06744, over 17023.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2664, pruned_loss=0.05578, over 3284743.99 frames. ], batch size: 41, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:24:00,838 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95105.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:24:25,495 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7172, 2.8193, 2.3603, 4.0621, 3.3936, 4.0929, 1.5401, 2.7209], device='cuda:0'), covar=tensor([0.1294, 0.0563, 0.1112, 0.0156, 0.0183, 0.0311, 0.1375, 0.0825], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0159, 0.0180, 0.0143, 0.0205, 0.0214, 0.0178, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 05:24:38,276 INFO [optim.py:368] (0/8) 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 05:24:44,276 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0899, 2.6791, 2.1086, 2.3540, 3.0549, 2.7134, 3.1731, 3.1773], device='cuda:0'), covar=tensor([0.0117, 0.0186, 0.0321, 0.0294, 0.0123, 0.0225, 0.0138, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0200, 0.0197, 0.0197, 0.0200, 0.0203, 0.0211, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:25:07,900 INFO [train.py:904] (0/8) Epoch 10, batch 3800, loss[loss=0.1857, simple_loss=0.2592, pruned_loss=0.05609, over 16720.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2684, pruned_loss=0.05762, over 3285683.94 frames. ], batch size: 89, lr: 6.90e-03, grad_scale: 8.0 2023-04-29 05:25:28,674 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95166.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:25:52,504 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9548, 3.4645, 3.4774, 3.4531, 3.4559, 3.3254, 3.0027, 3.4136], device='cuda:0'), covar=tensor([0.0723, 0.0609, 0.0600, 0.0629, 0.0749, 0.0591, 0.1266, 0.0556], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0341, 0.0344, 0.0326, 0.0385, 0.0361, 0.0466, 0.0285], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 05:26:20,913 INFO [train.py:904] (0/8) Epoch 10, batch 3850, loss[loss=0.1988, simple_loss=0.2685, pruned_loss=0.06457, over 16470.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2679, pruned_loss=0.05773, over 3283193.53 frames. ], batch size: 146, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:27:00,985 INFO [optim.py:368] (0/8) 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 05:27:31,963 INFO [train.py:904] (0/8) Epoch 10, batch 3900, loss[loss=0.1746, simple_loss=0.2472, pruned_loss=0.05102, over 16511.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2674, pruned_loss=0.05772, over 3287421.21 frames. ], batch size: 75, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:27:43,917 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:27:56,173 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:28:07,349 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4662, 3.7610, 3.9572, 2.6580, 3.6208, 3.9611, 3.7875, 2.3248], device='cuda:0'), covar=tensor([0.0343, 0.0073, 0.0027, 0.0257, 0.0049, 0.0057, 0.0042, 0.0300], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0067, 0.0067, 0.0121, 0.0075, 0.0083, 0.0073, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 05:28:45,287 INFO [train.py:904] (0/8) Epoch 10, batch 3950, loss[loss=0.1813, simple_loss=0.2717, pruned_loss=0.04543, over 17133.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2665, pruned_loss=0.05811, over 3285754.11 frames. ], batch size: 49, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:28:50,474 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 05:29:12,367 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:23,129 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:29:25,665 INFO [optim.py:368] (0/8) 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 05:29:38,095 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 4000, loss[loss=0.177, simple_loss=0.2546, pruned_loss=0.04971, over 16443.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2655, pruned_loss=0.05803, over 3295764.81 frames. ], batch size: 68, lr: 6.89e-03, grad_scale: 8.0 2023-04-29 05:30:28,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0096, 4.2357, 2.5468, 4.8652, 3.2036, 4.9630, 2.6487, 3.2454], device='cuda:0'), covar=tensor([0.0193, 0.0263, 0.1613, 0.0085, 0.0763, 0.0160, 0.1380, 0.0694], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0164, 0.0186, 0.0127, 0.0166, 0.0207, 0.0190, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 05:30:28,556 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8743, 4.0160, 4.0905, 2.3649, 3.5044, 2.7156, 4.1293, 4.1789], device='cuda:0'), covar=tensor([0.0147, 0.0519, 0.0431, 0.1510, 0.0646, 0.0769, 0.0437, 0.0626], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0145, 0.0156, 0.0142, 0.0135, 0.0124, 0.0136, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 05:31:05,668 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95400.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:31:07,596 INFO [train.py:904] (0/8) Epoch 10, batch 4050, loss[loss=0.1755, simple_loss=0.2597, pruned_loss=0.04562, over 16647.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2654, pruned_loss=0.05666, over 3289801.16 frames. ], batch size: 134, lr: 6.89e-03, grad_scale: 16.0 2023-04-29 05:31:49,146 INFO [optim.py:368] (0/8) 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 05:32:20,033 INFO [train.py:904] (0/8) Epoch 10, batch 4100, loss[loss=0.178, simple_loss=0.2722, pruned_loss=0.04187, over 16722.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2664, pruned_loss=0.05586, over 3282849.83 frames. ], batch size: 89, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:32:34,259 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95461.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 05:32:37,060 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7317, 2.6767, 1.9900, 2.4681, 3.0227, 2.8441, 3.4335, 3.4369], device='cuda:0'), covar=tensor([0.0042, 0.0252, 0.0381, 0.0307, 0.0161, 0.0228, 0.0124, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0198, 0.0196, 0.0195, 0.0200, 0.0200, 0.0208, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:33:33,945 INFO [train.py:904] (0/8) Epoch 10, batch 4150, loss[loss=0.2318, simple_loss=0.3175, pruned_loss=0.07299, over 16368.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2736, pruned_loss=0.05843, over 3267528.27 frames. ], batch size: 146, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:34:17,110 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.709e+02 3.150e+02 3.936e+02 7.135e+02, threshold=6.300e+02, percent-clipped=10.0 2023-04-29 05:34:32,369 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4371, 3.2845, 2.6072, 2.0243, 2.4098, 2.0975, 3.4812, 3.1713], device='cuda:0'), covar=tensor([0.2533, 0.0859, 0.1688, 0.2350, 0.2088, 0.1883, 0.0620, 0.1055], device='cuda:0'), in_proj_covar=tensor([0.0298, 0.0258, 0.0282, 0.0275, 0.0288, 0.0219, 0.0269, 0.0299], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 05:34:49,636 INFO [train.py:904] (0/8) Epoch 10, batch 4200, loss[loss=0.2401, simple_loss=0.3314, pruned_loss=0.07442, over 15445.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2818, pruned_loss=0.06083, over 3243513.23 frames. ], batch size: 191, lr: 6.88e-03, grad_scale: 16.0 2023-04-29 05:35:30,549 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8336, 3.9272, 4.2363, 4.2215, 4.2392, 3.9379, 3.8963, 3.9362], device='cuda:0'), covar=tensor([0.0321, 0.0420, 0.0392, 0.0358, 0.0359, 0.0361, 0.0965, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0329, 0.0332, 0.0316, 0.0376, 0.0353, 0.0454, 0.0280], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 05:36:04,073 INFO [train.py:904] (0/8) Epoch 10, batch 4250, loss[loss=0.2161, simple_loss=0.3013, pruned_loss=0.06546, over 16628.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2846, pruned_loss=0.06056, over 3208762.72 frames. ], batch size: 57, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:36:24,747 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:26,059 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:31,692 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4623, 3.0212, 2.6950, 2.2657, 2.2517, 2.1824, 3.0264, 2.8389], device='cuda:0'), covar=tensor([0.2116, 0.0685, 0.1308, 0.1990, 0.2222, 0.1793, 0.0568, 0.1097], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0257, 0.0280, 0.0274, 0.0286, 0.0219, 0.0268, 0.0296], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:36:37,907 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:36:49,134 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.325e+02 2.809e+02 3.314e+02 5.860e+02, threshold=5.619e+02, percent-clipped=0.0 2023-04-29 05:37:19,477 INFO [train.py:904] (0/8) Epoch 10, batch 4300, loss[loss=0.2317, simple_loss=0.3202, pruned_loss=0.07157, over 16732.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2861, pruned_loss=0.05974, over 3208628.57 frames. ], batch size: 83, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:37:59,697 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:15,377 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8985, 4.1515, 3.8640, 3.9700, 3.6760, 3.7909, 3.7555, 4.1318], device='cuda:0'), covar=tensor([0.0841, 0.0842, 0.1012, 0.0658, 0.0740, 0.1426, 0.0839, 0.0923], device='cuda:0'), in_proj_covar=tensor([0.0505, 0.0645, 0.0533, 0.0443, 0.0404, 0.0416, 0.0533, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:38:23,548 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95695.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:38:33,671 INFO [train.py:904] (0/8) Epoch 10, batch 4350, loss[loss=0.1969, simple_loss=0.2847, pruned_loss=0.05452, over 16821.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2898, pruned_loss=0.06103, over 3206492.56 frames. ], batch size: 42, lr: 6.88e-03, grad_scale: 8.0 2023-04-29 05:38:39,234 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4720, 3.5392, 3.2730, 3.0453, 3.1557, 3.4120, 3.2602, 3.1609], device='cuda:0'), covar=tensor([0.0488, 0.0349, 0.0256, 0.0263, 0.0599, 0.0303, 0.1158, 0.0461], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0294, 0.0281, 0.0258, 0.0303, 0.0291, 0.0192, 0.0326], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:38:51,094 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 05:39:18,687 INFO [optim.py:368] (0/8) 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 05:39:49,468 INFO [train.py:904] (0/8) Epoch 10, batch 4400, loss[loss=0.2044, simple_loss=0.299, pruned_loss=0.05495, over 16893.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2925, pruned_loss=0.0626, over 3190304.62 frames. ], batch size: 96, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:40:02,188 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95761.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:41:01,605 INFO [train.py:904] (0/8) Epoch 10, batch 4450, loss[loss=0.2412, simple_loss=0.3258, pruned_loss=0.07824, over 16401.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2964, pruned_loss=0.06419, over 3184078.00 frames. ], batch size: 146, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:41:07,069 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0009, 4.0921, 3.8555, 3.6159, 3.6003, 3.9880, 3.6600, 3.6620], device='cuda:0'), covar=tensor([0.0456, 0.0256, 0.0241, 0.0245, 0.0659, 0.0286, 0.0736, 0.0569], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0292, 0.0280, 0.0256, 0.0302, 0.0289, 0.0192, 0.0326], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:41:08,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1726, 4.2567, 4.0177, 3.7757, 3.7344, 4.1559, 3.8362, 3.8166], device='cuda:0'), covar=tensor([0.0446, 0.0237, 0.0241, 0.0253, 0.0735, 0.0265, 0.0627, 0.0571], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0292, 0.0280, 0.0256, 0.0302, 0.0289, 0.0191, 0.0326], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:41:11,315 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2048, 3.5932, 3.3904, 1.7253, 2.9956, 2.3963, 3.4803, 3.7281], device='cuda:0'), covar=tensor([0.0193, 0.0543, 0.0526, 0.1913, 0.0689, 0.0864, 0.0561, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0143, 0.0154, 0.0141, 0.0134, 0.0123, 0.0134, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 05:41:12,950 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95809.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:17,751 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95812.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 05:41:40,067 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8032, 4.5863, 4.5187, 2.9190, 3.9644, 4.4228, 4.0349, 2.7189], device='cuda:0'), covar=tensor([0.0355, 0.0014, 0.0029, 0.0283, 0.0054, 0.0058, 0.0045, 0.0290], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0066, 0.0066, 0.0121, 0.0076, 0.0083, 0.0073, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 05:41:46,153 INFO [optim.py:368] (0/8) 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 05:42:08,793 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95847.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:42:15,008 INFO [train.py:904] (0/8) Epoch 10, batch 4500, loss[loss=0.1774, simple_loss=0.2675, pruned_loss=0.04363, over 16442.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2967, pruned_loss=0.06433, over 3192351.57 frames. ], batch size: 75, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:42:45,922 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95873.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:43:27,217 INFO [train.py:904] (0/8) Epoch 10, batch 4550, loss[loss=0.2202, simple_loss=0.3031, pruned_loss=0.06861, over 16874.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.2972, pruned_loss=0.06492, over 3203027.29 frames. ], batch size: 116, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:43:35,504 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=95908.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 05:43:40,723 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 05:43:47,832 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:43:59,534 INFO [zipformer.py:625] (0/8) 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] (0/8) 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] (0/8) Epoch 10, batch 4600, loss[loss=0.1992, simple_loss=0.2887, pruned_loss=0.05485, over 16718.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2983, pruned_loss=0.06527, over 3189277.66 frames. ], batch size: 124, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:44:57,860 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:00,673 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-29 05:45:09,542 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=95972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:11,442 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=95973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:19,751 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7431, 4.4379, 4.2767, 4.8194, 5.0465, 4.4917, 4.9078, 5.0561], device='cuda:0'), covar=tensor([0.1070, 0.0947, 0.2217, 0.0838, 0.0591, 0.0841, 0.0908, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0492, 0.0613, 0.0752, 0.0623, 0.0464, 0.0481, 0.0489, 0.0553], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:45:20,953 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=95980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:35,434 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4559, 2.9365, 2.6064, 2.2617, 2.2860, 2.1059, 2.9188, 2.8672], device='cuda:0'), covar=tensor([0.1964, 0.0664, 0.1342, 0.1853, 0.1947, 0.1791, 0.0473, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0257, 0.0281, 0.0276, 0.0285, 0.0219, 0.0268, 0.0294], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:45:43,248 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=95995.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:45:49,806 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-96000.pt 2023-04-29 05:45:55,190 INFO [train.py:904] (0/8) Epoch 10, batch 4650, loss[loss=0.2007, simple_loss=0.2817, pruned_loss=0.05982, over 17196.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.297, pruned_loss=0.06487, over 3183864.79 frames. ], batch size: 46, lr: 6.87e-03, grad_scale: 8.0 2023-04-29 05:45:55,721 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8342, 4.0402, 2.9612, 2.3862, 3.0148, 2.3666, 4.3549, 3.7547], device='cuda:0'), covar=tensor([0.2425, 0.0705, 0.1552, 0.1963, 0.2222, 0.1798, 0.0416, 0.0833], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0257, 0.0281, 0.0276, 0.0286, 0.0219, 0.0269, 0.0295], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:46:40,628 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.329e+02 2.110e+02 2.403e+02 2.925e+02 6.538e+02, threshold=4.805e+02, percent-clipped=1.0 2023-04-29 05:46:52,315 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4599, 2.4803, 1.9220, 2.3441, 2.8596, 2.5397, 3.2346, 3.1422], device='cuda:0'), covar=tensor([0.0051, 0.0279, 0.0398, 0.0312, 0.0158, 0.0281, 0.0108, 0.0148], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0195, 0.0193, 0.0190, 0.0194, 0.0197, 0.0198, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:46:55,319 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:46:58,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5029, 2.1123, 2.3101, 4.2240, 2.0706, 2.5769, 2.2047, 2.3638], device='cuda:0'), covar=tensor([0.0874, 0.3076, 0.2040, 0.0355, 0.3721, 0.2161, 0.2815, 0.2937], device='cuda:0'), in_proj_covar=tensor([0.0358, 0.0385, 0.0318, 0.0318, 0.0407, 0.0441, 0.0346, 0.0452], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:47:10,302 INFO [train.py:904] (0/8) Epoch 10, batch 4700, loss[loss=0.1858, simple_loss=0.2697, pruned_loss=0.05093, over 17050.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2936, pruned_loss=0.06303, over 3196047.92 frames. ], batch size: 55, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:47:13,900 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0300, 3.4317, 3.2793, 1.9759, 2.9108, 2.2447, 3.3904, 3.4648], device='cuda:0'), covar=tensor([0.0254, 0.0619, 0.0574, 0.1737, 0.0747, 0.0926, 0.0641, 0.0883], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0143, 0.0155, 0.0142, 0.0134, 0.0123, 0.0134, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 05:47:52,305 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:48:18,475 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1344, 5.0842, 4.9942, 4.6333, 4.4795, 4.9679, 4.9875, 4.6636], device='cuda:0'), covar=tensor([0.0481, 0.0343, 0.0218, 0.0214, 0.0986, 0.0402, 0.0217, 0.0491], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0287, 0.0274, 0.0252, 0.0298, 0.0285, 0.0187, 0.0319], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:48:24,003 INFO [train.py:904] (0/8) Epoch 10, batch 4750, loss[loss=0.1864, simple_loss=0.2795, pruned_loss=0.04668, over 16914.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2896, pruned_loss=0.06099, over 3197196.70 frames. ], batch size: 96, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:08,964 INFO [optim.py:368] (0/8) 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 05:49:22,624 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96141.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:49:38,024 INFO [train.py:904] (0/8) Epoch 10, batch 4800, loss[loss=0.1987, simple_loss=0.2909, pruned_loss=0.05327, over 16413.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2867, pruned_loss=0.0592, over 3210678.61 frames. ], batch size: 146, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:49:58,194 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96165.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:49:59,975 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3469, 2.0508, 1.6658, 1.8961, 2.4283, 2.1823, 2.3352, 2.5873], device='cuda:0'), covar=tensor([0.0098, 0.0286, 0.0358, 0.0309, 0.0149, 0.0238, 0.0109, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0197, 0.0195, 0.0193, 0.0196, 0.0199, 0.0200, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:50:03,021 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96168.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:50:54,671 INFO [train.py:904] (0/8) Epoch 10, batch 4850, loss[loss=0.189, simple_loss=0.2821, pruned_loss=0.04794, over 16851.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2875, pruned_loss=0.05845, over 3205318.65 frames. ], batch size: 96, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:50:56,289 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96203.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:51:15,491 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 05:51:32,250 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96226.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:51:40,173 INFO [optim.py:368] (0/8) 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 05:51:53,142 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1207, 2.7747, 2.6664, 1.9639, 2.6002, 2.1925, 2.7932, 2.8376], device='cuda:0'), covar=tensor([0.0304, 0.0631, 0.0580, 0.1624, 0.0737, 0.0875, 0.0553, 0.0586], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0141, 0.0155, 0.0142, 0.0133, 0.0122, 0.0133, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 05:52:10,088 INFO [train.py:904] (0/8) Epoch 10, batch 4900, loss[loss=0.1698, simple_loss=0.2586, pruned_loss=0.04047, over 16640.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2867, pruned_loss=0.05742, over 3197416.46 frames. ], batch size: 57, lr: 6.86e-03, grad_scale: 8.0 2023-04-29 05:52:32,317 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.4987, 2.4647, 2.2124, 3.9112, 2.5920, 3.8716, 1.3713, 2.7142], device='cuda:0'), covar=tensor([0.1478, 0.0828, 0.1366, 0.0163, 0.0265, 0.0391, 0.1716, 0.0914], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0153, 0.0175, 0.0134, 0.0196, 0.0202, 0.0175, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 05:52:42,494 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:24,560 INFO [train.py:904] (0/8) Epoch 10, batch 4950, loss[loss=0.2207, simple_loss=0.3052, pruned_loss=0.06813, over 11841.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2861, pruned_loss=0.05684, over 3195774.55 frames. ], batch size: 246, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:53:44,078 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3458, 5.0512, 5.2685, 5.4837, 5.6944, 5.0198, 5.6536, 5.6433], device='cuda:0'), covar=tensor([0.1220, 0.0956, 0.1301, 0.0555, 0.0355, 0.0497, 0.0369, 0.0429], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0619, 0.0755, 0.0630, 0.0469, 0.0486, 0.0492, 0.0559], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:53:51,590 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:53:56,366 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:03,957 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1233, 4.2911, 3.3916, 2.5848, 3.4390, 2.7992, 4.5250, 4.0688], device='cuda:0'), covar=tensor([0.2253, 0.0701, 0.1438, 0.1847, 0.2014, 0.1373, 0.0492, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0256, 0.0278, 0.0273, 0.0281, 0.0216, 0.0266, 0.0291], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:54:05,870 INFO [optim.py:368] (0/8) 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 05:54:12,462 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:54:33,267 INFO [train.py:904] (0/8) Epoch 10, batch 5000, loss[loss=0.2148, simple_loss=0.3163, pruned_loss=0.05662, over 16721.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2879, pruned_loss=0.05733, over 3194665.04 frames. ], batch size: 124, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:54:42,340 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5340, 3.4845, 3.4489, 2.8722, 3.3967, 2.0056, 3.1300, 2.8453], device='cuda:0'), covar=tensor([0.0114, 0.0102, 0.0113, 0.0233, 0.0080, 0.1919, 0.0124, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0109, 0.0154, 0.0150, 0.0126, 0.0168, 0.0144, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 05:55:20,995 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:28,019 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:55:44,803 INFO [train.py:904] (0/8) Epoch 10, batch 5050, loss[loss=0.1979, simple_loss=0.285, pruned_loss=0.05538, over 16935.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2886, pruned_loss=0.05735, over 3198322.53 frames. ], batch size: 96, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:56:02,250 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 05:56:27,894 INFO [optim.py:368] (0/8) 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 05:56:32,886 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96436.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:54,560 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96451.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:56:55,807 INFO [train.py:904] (0/8) Epoch 10, batch 5100, loss[loss=0.1708, simple_loss=0.2619, pruned_loss=0.0399, over 16478.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2867, pruned_loss=0.05674, over 3196633.58 frames. ], batch size: 146, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:57:20,215 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96468.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:57:58,163 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1138, 4.0834, 4.4676, 4.4490, 4.4593, 4.1801, 4.1818, 4.1385], device='cuda:0'), covar=tensor([0.0268, 0.0510, 0.0327, 0.0311, 0.0313, 0.0271, 0.0687, 0.0365], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0322, 0.0327, 0.0310, 0.0373, 0.0347, 0.0446, 0.0277], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 05:58:08,613 INFO [train.py:904] (0/8) Epoch 10, batch 5150, loss[loss=0.2189, simple_loss=0.3087, pruned_loss=0.06451, over 16685.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2869, pruned_loss=0.05608, over 3190533.33 frames. ], batch size: 134, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:58:10,631 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:29,580 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96516.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 05:58:36,676 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96521.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:58:52,022 INFO [optim.py:368] (0/8) 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 05:59:21,923 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 05:59:22,766 INFO [train.py:904] (0/8) Epoch 10, batch 5200, loss[loss=0.2123, simple_loss=0.2858, pruned_loss=0.06937, over 16549.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2855, pruned_loss=0.05585, over 3171794.49 frames. ], batch size: 62, lr: 6.85e-03, grad_scale: 8.0 2023-04-29 05:59:40,484 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1240, 3.5179, 3.6426, 1.4180, 3.7846, 3.8794, 2.8051, 2.5078], device='cuda:0'), covar=tensor([0.1129, 0.0153, 0.0111, 0.1490, 0.0070, 0.0072, 0.0413, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0099, 0.0085, 0.0138, 0.0069, 0.0096, 0.0120, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 05:59:55,692 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 06:00:35,350 INFO [train.py:904] (0/8) Epoch 10, batch 5250, loss[loss=0.1693, simple_loss=0.2456, pruned_loss=0.04657, over 16972.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2823, pruned_loss=0.055, over 3183835.36 frames. ], batch size: 41, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:00:38,872 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6228, 2.6390, 2.2457, 3.7625, 2.7205, 3.8934, 1.4267, 2.8315], device='cuda:0'), covar=tensor([0.1306, 0.0641, 0.1208, 0.0108, 0.0180, 0.0339, 0.1524, 0.0757], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0154, 0.0175, 0.0134, 0.0197, 0.0203, 0.0175, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 06:01:21,023 INFO [optim.py:368] (0/8) 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 06:01:25,805 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96636.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:01:48,995 INFO [train.py:904] (0/8) Epoch 10, batch 5300, loss[loss=0.1772, simple_loss=0.2634, pruned_loss=0.04548, over 16572.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2782, pruned_loss=0.05335, over 3202669.85 frames. ], batch size: 62, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:02:30,153 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=96680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:02:36,339 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:03:03,547 INFO [train.py:904] (0/8) Epoch 10, batch 5350, loss[loss=0.2316, simple_loss=0.302, pruned_loss=0.08056, over 12344.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2767, pruned_loss=0.05271, over 3203636.46 frames. ], batch size: 248, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:03:21,008 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 06:03:21,041 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 06:03:21,686 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2736, 4.1055, 4.2821, 4.5168, 4.6543, 4.2218, 4.5510, 4.6063], device='cuda:0'), covar=tensor([0.1332, 0.1017, 0.1472, 0.0570, 0.0420, 0.0937, 0.0566, 0.0558], device='cuda:0'), in_proj_covar=tensor([0.0503, 0.0625, 0.0771, 0.0639, 0.0477, 0.0489, 0.0497, 0.0567], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:03:48,675 INFO [optim.py:368] (0/8) 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 06:03:52,807 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:04:08,476 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 5400, loss[loss=0.1875, simple_loss=0.2769, pruned_loss=0.04906, over 17030.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2799, pruned_loss=0.05364, over 3197625.81 frames. ], batch size: 55, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:04:51,430 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=96776.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:02,545 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:05:31,411 INFO [train.py:904] (0/8) Epoch 10, batch 5450, loss[loss=0.2171, simple_loss=0.2967, pruned_loss=0.06873, over 17111.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.284, pruned_loss=0.0562, over 3201602.42 frames. ], batch size: 47, lr: 6.84e-03, grad_scale: 4.0 2023-04-29 06:05:41,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 06:05:50,152 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 06:06:02,209 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:20,034 INFO [optim.py:368] (0/8) 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 06:06:27,196 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=96837.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:06:49,559 INFO [train.py:904] (0/8) Epoch 10, batch 5500, loss[loss=0.2346, simple_loss=0.3143, pruned_loss=0.07748, over 16642.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2914, pruned_loss=0.06111, over 3175285.30 frames. ], batch size: 62, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:07:17,929 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=96869.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:08:08,516 INFO [train.py:904] (0/8) Epoch 10, batch 5550, loss[loss=0.3372, simple_loss=0.3774, pruned_loss=0.1486, over 11152.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2994, pruned_loss=0.0669, over 3142750.43 frames. ], batch size: 248, lr: 6.83e-03, grad_scale: 4.0 2023-04-29 06:09:01,356 INFO [optim.py:368] (0/8) 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 06:09:28,007 INFO [train.py:904] (0/8) Epoch 10, batch 5600, loss[loss=0.2187, simple_loss=0.2989, pruned_loss=0.06922, over 16869.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3051, pruned_loss=0.07161, over 3117263.01 frames. ], batch size: 116, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:09:48,159 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8205, 2.5045, 2.4298, 1.7419, 2.6364, 2.7061, 2.3272, 2.2458], device='cuda:0'), covar=tensor([0.0823, 0.0191, 0.0186, 0.0979, 0.0098, 0.0162, 0.0383, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0099, 0.0087, 0.0140, 0.0069, 0.0097, 0.0121, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 06:10:15,413 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=96980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:51,254 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97000.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:10:53,954 INFO [train.py:904] (0/8) Epoch 10, batch 5650, loss[loss=0.2253, simple_loss=0.3142, pruned_loss=0.06821, over 17224.00 frames. ], tot_loss[loss=0.2305, simple_loss=0.3102, pruned_loss=0.07536, over 3100614.90 frames. ], batch size: 44, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:11:33,822 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:11:43,934 INFO [optim.py:368] (0/8) 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 06:12:02,151 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:12:11,119 INFO [train.py:904] (0/8) Epoch 10, batch 5700, loss[loss=0.2354, simple_loss=0.3146, pruned_loss=0.07814, over 17149.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3126, pruned_loss=0.07785, over 3076616.91 frames. ], batch size: 46, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:12:25,354 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:16,340 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:13:24,227 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 06:13:29,505 INFO [train.py:904] (0/8) Epoch 10, batch 5750, loss[loss=0.2724, simple_loss=0.3269, pruned_loss=0.109, over 11221.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3147, pruned_loss=0.07878, over 3065057.08 frames. ], batch size: 247, lr: 6.83e-03, grad_scale: 2.0 2023-04-29 06:13:48,783 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7066, 3.7806, 2.0332, 4.1776, 2.8054, 4.1194, 2.2700, 2.8745], device='cuda:0'), covar=tensor([0.0199, 0.0296, 0.1663, 0.0124, 0.0748, 0.0415, 0.1477, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0162, 0.0186, 0.0120, 0.0166, 0.0201, 0.0189, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 06:14:05,240 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 06:14:17,787 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:14:22,109 INFO [optim.py:368] (0/8) 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 06:14:40,074 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0964, 4.8788, 4.7723, 3.5555, 4.3141, 4.8068, 4.3476, 2.9025], device='cuda:0'), covar=tensor([0.0314, 0.0021, 0.0026, 0.0216, 0.0043, 0.0060, 0.0033, 0.0289], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0066, 0.0067, 0.0123, 0.0076, 0.0086, 0.0074, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 06:14:47,174 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 06:14:49,734 INFO [train.py:904] (0/8) Epoch 10, batch 5800, loss[loss=0.2029, simple_loss=0.3, pruned_loss=0.05295, over 16816.00 frames. ], tot_loss[loss=0.2341, simple_loss=0.3143, pruned_loss=0.07697, over 3078316.28 frames. ], batch size: 42, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:15:28,797 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2787, 4.9817, 4.8564, 5.3667, 5.5516, 4.8346, 5.4726, 5.5364], device='cuda:0'), covar=tensor([0.1311, 0.1009, 0.2197, 0.0855, 0.0768, 0.0970, 0.0788, 0.0798], device='cuda:0'), in_proj_covar=tensor([0.0489, 0.0609, 0.0739, 0.0621, 0.0472, 0.0476, 0.0488, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:15:31,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8275, 3.8626, 4.2300, 4.2186, 4.2203, 3.9042, 3.9622, 3.8903], device='cuda:0'), covar=tensor([0.0314, 0.0499, 0.0403, 0.0403, 0.0448, 0.0415, 0.0920, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0318, 0.0322, 0.0307, 0.0369, 0.0345, 0.0441, 0.0276], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 06:16:07,875 INFO [train.py:904] (0/8) Epoch 10, batch 5850, loss[loss=0.2441, simple_loss=0.3039, pruned_loss=0.09211, over 11109.00 frames. ], tot_loss[loss=0.2307, simple_loss=0.3116, pruned_loss=0.07492, over 3073792.16 frames. ], batch size: 248, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:16:31,904 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 06:16:56,368 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 06:17:00,835 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.062e+02 2.965e+02 3.707e+02 4.622e+02 9.015e+02, threshold=7.415e+02, percent-clipped=1.0 2023-04-29 06:17:28,950 INFO [train.py:904] (0/8) Epoch 10, batch 5900, loss[loss=0.2053, simple_loss=0.2947, pruned_loss=0.05795, over 16546.00 frames. ], tot_loss[loss=0.229, simple_loss=0.3102, pruned_loss=0.07392, over 3090290.72 frames. ], batch size: 75, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:17:45,057 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97260.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:18:40,302 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9695, 4.0130, 4.4021, 4.3778, 4.3646, 4.1101, 4.1140, 4.0063], device='cuda:0'), covar=tensor([0.0321, 0.0532, 0.0360, 0.0358, 0.0432, 0.0396, 0.0853, 0.0474], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0319, 0.0322, 0.0309, 0.0373, 0.0347, 0.0442, 0.0278], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 06:18:49,544 INFO [train.py:904] (0/8) Epoch 10, batch 5950, loss[loss=0.1962, simple_loss=0.288, pruned_loss=0.05226, over 17217.00 frames. ], tot_loss[loss=0.2276, simple_loss=0.3103, pruned_loss=0.07249, over 3089027.91 frames. ], batch size: 45, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:19:00,792 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 06:19:11,277 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-04-29 06:19:19,817 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:19:41,585 INFO [optim.py:368] (0/8) 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 06:19:52,318 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:01,444 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97347.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:09,128 INFO [train.py:904] (0/8) Epoch 10, batch 6000, loss[loss=0.2191, simple_loss=0.3013, pruned_loss=0.06848, over 16153.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.309, pruned_loss=0.07185, over 3097468.34 frames. ], batch size: 165, lr: 6.82e-03, grad_scale: 4.0 2023-04-29 06:20:09,129 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 06:20:23,732 INFO [train.py:938] (0/8) Epoch 10, validation: loss=0.165, simple_loss=0.2783, pruned_loss=0.02583, over 944034.00 frames. 2023-04-29 06:20:23,733 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 06:20:30,443 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:20:57,535 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97374.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:21:19,416 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7944, 5.1275, 4.8634, 4.9043, 4.6270, 4.6262, 4.5484, 5.2408], device='cuda:0'), covar=tensor([0.0977, 0.0774, 0.0990, 0.0676, 0.0792, 0.0813, 0.0944, 0.0787], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0629, 0.0525, 0.0437, 0.0395, 0.0408, 0.0521, 0.0480], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:21:29,920 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-29 06:21:42,487 INFO [train.py:904] (0/8) Epoch 10, batch 6050, loss[loss=0.2096, simple_loss=0.2941, pruned_loss=0.06259, over 15525.00 frames. ], tot_loss[loss=0.2255, simple_loss=0.3079, pruned_loss=0.07159, over 3096612.01 frames. ], batch size: 191, lr: 6.82e-03, grad_scale: 2.0 2023-04-29 06:21:44,922 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97403.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:21:52,122 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:29,816 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:35,055 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:22:35,675 INFO [optim.py:368] (0/8) 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 06:23:02,115 INFO [train.py:904] (0/8) Epoch 10, batch 6100, loss[loss=0.2223, simple_loss=0.3058, pruned_loss=0.06942, over 16311.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3074, pruned_loss=0.07064, over 3100105.04 frames. ], batch size: 165, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:23:31,471 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.28 vs. limit=5.0 2023-04-29 06:23:48,677 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97480.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:24:23,713 INFO [train.py:904] (0/8) Epoch 10, batch 6150, loss[loss=0.2114, simple_loss=0.2921, pruned_loss=0.06537, over 16935.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3053, pruned_loss=0.06979, over 3107747.28 frames. ], batch size: 109, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:06,394 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6185, 2.3225, 2.4156, 4.2796, 2.1245, 2.7847, 2.3133, 2.5211], device='cuda:0'), covar=tensor([0.0840, 0.2864, 0.1870, 0.0332, 0.3523, 0.1830, 0.2643, 0.2580], device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0381, 0.0318, 0.0314, 0.0404, 0.0434, 0.0343, 0.0446], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:25:17,521 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.161e+02 3.229e+02 3.939e+02 5.020e+02 8.476e+02, threshold=7.879e+02, percent-clipped=2.0 2023-04-29 06:25:41,181 INFO [train.py:904] (0/8) Epoch 10, batch 6200, loss[loss=0.2425, simple_loss=0.3218, pruned_loss=0.08164, over 15245.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3033, pruned_loss=0.06911, over 3127354.89 frames. ], batch size: 190, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:25:46,710 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:26:22,326 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6122, 2.2903, 2.3732, 4.4438, 2.1369, 2.7294, 2.3115, 2.4493], device='cuda:0'), covar=tensor([0.0812, 0.2901, 0.1956, 0.0311, 0.3413, 0.1959, 0.2779, 0.2809], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0383, 0.0320, 0.0316, 0.0405, 0.0436, 0.0345, 0.0448], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:26:57,947 INFO [train.py:904] (0/8) Epoch 10, batch 6250, loss[loss=0.1972, simple_loss=0.2921, pruned_loss=0.05116, over 16637.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3036, pruned_loss=0.0693, over 3131606.97 frames. ], batch size: 76, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:27:06,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6256, 2.2898, 2.3871, 4.4221, 2.0792, 2.7477, 2.2943, 2.4854], device='cuda:0'), covar=tensor([0.0815, 0.2995, 0.2015, 0.0295, 0.3568, 0.2053, 0.2983, 0.2752], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0382, 0.0319, 0.0314, 0.0405, 0.0435, 0.0344, 0.0447], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:27:18,728 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:18,883 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97616.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:27:47,813 INFO [optim.py:368] (0/8) 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 06:28:11,806 INFO [train.py:904] (0/8) Epoch 10, batch 6300, loss[loss=0.205, simple_loss=0.2996, pruned_loss=0.0552, over 16857.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3029, pruned_loss=0.06825, over 3144581.27 frames. ], batch size: 102, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:28:18,582 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:28:48,472 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97674.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:29:24,494 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97698.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:29:30,243 INFO [train.py:904] (0/8) Epoch 10, batch 6350, loss[loss=0.2173, simple_loss=0.3014, pruned_loss=0.06655, over 16721.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3038, pruned_loss=0.06928, over 3148669.14 frames. ], batch size: 76, lr: 6.81e-03, grad_scale: 2.0 2023-04-29 06:29:31,925 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97703.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:33,049 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:29:47,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3819, 4.1203, 4.2665, 4.5956, 4.6461, 4.3657, 4.6904, 4.6713], device='cuda:0'), covar=tensor([0.1458, 0.1157, 0.1710, 0.0713, 0.0796, 0.0983, 0.0779, 0.0745], device='cuda:0'), in_proj_covar=tensor([0.0504, 0.0622, 0.0756, 0.0635, 0.0485, 0.0482, 0.0504, 0.0564], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:30:13,652 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97730.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:30:21,591 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97735.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 06:30:22,850 INFO [optim.py:368] (0/8) 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 06:30:46,257 INFO [train.py:904] (0/8) Epoch 10, batch 6400, loss[loss=0.219, simple_loss=0.2997, pruned_loss=0.06919, over 16898.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3042, pruned_loss=0.07074, over 3132573.89 frames. ], batch size: 96, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:00,796 INFO [train.py:904] (0/8) Epoch 10, batch 6450, loss[loss=0.2024, simple_loss=0.2822, pruned_loss=0.06136, over 17038.00 frames. ], tot_loss[loss=0.222, simple_loss=0.304, pruned_loss=0.06999, over 3131756.72 frames. ], batch size: 53, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:32:38,708 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 06:32:57,258 INFO [optim.py:368] (0/8) 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 06:33:21,812 INFO [train.py:904] (0/8) Epoch 10, batch 6500, loss[loss=0.2005, simple_loss=0.2911, pruned_loss=0.05489, over 16837.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3025, pruned_loss=0.06908, over 3138919.82 frames. ], batch size: 102, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:33:44,410 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 06:34:17,016 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=97888.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:34:19,715 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4205, 4.3997, 4.2772, 3.5472, 4.2794, 1.6825, 4.0030, 4.0718], device='cuda:0'), covar=tensor([0.0087, 0.0075, 0.0134, 0.0365, 0.0083, 0.2341, 0.0122, 0.0161], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0109, 0.0156, 0.0152, 0.0127, 0.0172, 0.0144, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:34:41,162 INFO [train.py:904] (0/8) Epoch 10, batch 6550, loss[loss=0.2472, simple_loss=0.3389, pruned_loss=0.07777, over 15307.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3055, pruned_loss=0.07031, over 3131981.76 frames. ], batch size: 190, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:34:55,682 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=97911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:05,036 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97916.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:35:10,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 06:35:35,381 INFO [optim.py:368] (0/8) 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 06:35:56,374 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=97949.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:36:00,715 INFO [train.py:904] (0/8) Epoch 10, batch 6600, loss[loss=0.1951, simple_loss=0.292, pruned_loss=0.04909, over 16707.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3078, pruned_loss=0.0708, over 3125140.89 frames. ], batch size: 57, lr: 6.80e-03, grad_scale: 4.0 2023-04-29 06:36:19,348 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=97964.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:14,630 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=97998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:37:17,155 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-98000.pt 2023-04-29 06:37:22,493 INFO [train.py:904] (0/8) Epoch 10, batch 6650, loss[loss=0.2096, simple_loss=0.291, pruned_loss=0.06413, over 15388.00 frames. ], tot_loss[loss=0.2261, simple_loss=0.3079, pruned_loss=0.07217, over 3102237.85 frames. ], batch size: 190, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:37:24,123 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:05,323 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98030.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 06:38:05,346 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 06:38:31,007 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98046.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:38,425 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:38:39,796 INFO [train.py:904] (0/8) Epoch 10, batch 6700, loss[loss=0.2287, simple_loss=0.3085, pruned_loss=0.07451, over 16271.00 frames. ], tot_loss[loss=0.2247, simple_loss=0.3061, pruned_loss=0.07169, over 3105213.82 frames. ], batch size: 165, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:39:14,304 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6758, 4.6870, 4.5089, 4.2797, 4.1365, 4.5925, 4.4761, 4.2583], device='cuda:0'), covar=tensor([0.0564, 0.0420, 0.0241, 0.0243, 0.0916, 0.0410, 0.0368, 0.0661], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0292, 0.0274, 0.0252, 0.0298, 0.0285, 0.0187, 0.0318], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:39:20,427 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:39:57,938 INFO [train.py:904] (0/8) Epoch 10, batch 6750, loss[loss=0.1921, simple_loss=0.2757, pruned_loss=0.0543, over 16684.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3057, pruned_loss=0.07222, over 3090131.49 frames. ], batch size: 62, lr: 6.79e-03, grad_scale: 4.0 2023-04-29 06:40:04,364 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:40:04,608 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 06:40:49,808 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.169e+02 3.584e+02 4.478e+02 5.468e+02 7.454e+02, threshold=8.956e+02, percent-clipped=0.0 2023-04-29 06:41:15,038 INFO [train.py:904] (0/8) Epoch 10, batch 6800, loss[loss=0.2045, simple_loss=0.2997, pruned_loss=0.05467, over 16681.00 frames. ], tot_loss[loss=0.2248, simple_loss=0.3057, pruned_loss=0.07198, over 3095266.31 frames. ], batch size: 76, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:41:38,626 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98167.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:42:33,974 INFO [train.py:904] (0/8) Epoch 10, batch 6850, loss[loss=0.2316, simple_loss=0.3011, pruned_loss=0.08103, over 12137.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3073, pruned_loss=0.07265, over 3101216.54 frames. ], batch size: 247, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:42:39,901 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2612, 3.3981, 3.7017, 1.6962, 3.9139, 3.9289, 3.0349, 2.8802], device='cuda:0'), covar=tensor([0.0782, 0.0184, 0.0119, 0.1235, 0.0045, 0.0100, 0.0320, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0099, 0.0087, 0.0139, 0.0068, 0.0097, 0.0121, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 06:42:47,565 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98211.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:42:48,192 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 06:43:13,602 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7069, 3.9197, 4.4195, 2.0375, 4.6020, 4.5541, 3.4185, 3.4571], device='cuda:0'), covar=tensor([0.0655, 0.0152, 0.0110, 0.1045, 0.0039, 0.0073, 0.0230, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0098, 0.0086, 0.0138, 0.0068, 0.0096, 0.0119, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 06:43:24,607 INFO [optim.py:368] (0/8) 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 06:43:37,277 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 6900, loss[loss=0.2452, simple_loss=0.3292, pruned_loss=0.08062, over 16271.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3094, pruned_loss=0.07159, over 3126471.35 frames. ], batch size: 165, lr: 6.79e-03, grad_scale: 8.0 2023-04-29 06:44:01,152 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:44:05,066 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 06:45:09,164 INFO [train.py:904] (0/8) Epoch 10, batch 6950, loss[loss=0.2368, simple_loss=0.3118, pruned_loss=0.08092, over 15334.00 frames. ], tot_loss[loss=0.2308, simple_loss=0.3124, pruned_loss=0.07466, over 3101629.43 frames. ], batch size: 191, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:45:54,259 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98330.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:46:01,761 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.874e+02 3.204e+02 3.949e+02 4.845e+02 8.099e+02, threshold=7.898e+02, percent-clipped=3.0 2023-04-29 06:46:04,469 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 06:46:27,413 INFO [train.py:904] (0/8) Epoch 10, batch 7000, loss[loss=0.231, simple_loss=0.3282, pruned_loss=0.06693, over 16685.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3119, pruned_loss=0.0733, over 3115667.30 frames. ], batch size: 57, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:07,456 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98378.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 06:47:43,144 INFO [train.py:904] (0/8) Epoch 10, batch 7050, loss[loss=0.2363, simple_loss=0.3107, pruned_loss=0.08101, over 15334.00 frames. ], tot_loss[loss=0.2286, simple_loss=0.3122, pruned_loss=0.07246, over 3122709.67 frames. ], batch size: 191, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:47:56,640 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98410.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:48:17,572 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 06:48:34,412 INFO [optim.py:368] (0/8) 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 06:48:38,679 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1724, 4.1819, 4.4243, 4.2264, 4.3119, 4.7940, 4.3804, 4.1559], device='cuda:0'), covar=tensor([0.1581, 0.1829, 0.1757, 0.1987, 0.2397, 0.1056, 0.1392, 0.2469], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0474, 0.0509, 0.0411, 0.0543, 0.0535, 0.0412, 0.0561], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 06:48:59,553 INFO [train.py:904] (0/8) Epoch 10, batch 7100, loss[loss=0.2099, simple_loss=0.2988, pruned_loss=0.06052, over 16482.00 frames. ], tot_loss[loss=0.2281, simple_loss=0.3107, pruned_loss=0.07274, over 3101662.39 frames. ], batch size: 146, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:49:14,099 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:49:27,370 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:50:00,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8582, 5.4311, 5.6992, 5.3976, 5.4729, 6.0329, 5.5085, 5.3238], device='cuda:0'), covar=tensor([0.0882, 0.1820, 0.1740, 0.1866, 0.2645, 0.1037, 0.1475, 0.2455], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0477, 0.0512, 0.0415, 0.0548, 0.0541, 0.0417, 0.0566], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 06:50:06,144 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8077, 2.4734, 2.5193, 1.6972, 2.6551, 2.8183, 2.3180, 2.2788], device='cuda:0'), covar=tensor([0.0855, 0.0182, 0.0176, 0.1077, 0.0099, 0.0156, 0.0437, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0099, 0.0086, 0.0138, 0.0068, 0.0096, 0.0120, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 06:50:11,986 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6795, 3.5587, 3.8134, 3.5744, 3.7347, 4.1524, 3.8355, 3.6146], device='cuda:0'), covar=tensor([0.2365, 0.2587, 0.2120, 0.2583, 0.3063, 0.1859, 0.1683, 0.2831], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0477, 0.0512, 0.0415, 0.0548, 0.0541, 0.0417, 0.0566], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 06:50:12,815 INFO [train.py:904] (0/8) Epoch 10, batch 7150, loss[loss=0.2768, simple_loss=0.3317, pruned_loss=0.111, over 11080.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3084, pruned_loss=0.07262, over 3085303.14 frames. ], batch size: 247, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:50:32,410 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 06:51:03,788 INFO [optim.py:368] (0/8) 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 06:51:15,745 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:16,726 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98544.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:51:29,375 INFO [train.py:904] (0/8) Epoch 10, batch 7200, loss[loss=0.184, simple_loss=0.2746, pruned_loss=0.04671, over 15432.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.3063, pruned_loss=0.07094, over 3069081.96 frames. ], batch size: 191, lr: 6.78e-03, grad_scale: 8.0 2023-04-29 06:52:32,653 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98592.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:52:49,223 INFO [train.py:904] (0/8) Epoch 10, batch 7250, loss[loss=0.21, simple_loss=0.2902, pruned_loss=0.06488, over 16696.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3034, pruned_loss=0.06932, over 3085872.08 frames. ], batch size: 76, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:52:53,591 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98604.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:53:45,149 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.674e+02 3.052e+02 3.513e+02 4.472e+02 1.147e+03, threshold=7.026e+02, percent-clipped=3.0 2023-04-29 06:54:05,881 INFO [train.py:904] (0/8) Epoch 10, batch 7300, loss[loss=0.243, simple_loss=0.307, pruned_loss=0.0895, over 11697.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3023, pruned_loss=0.06905, over 3087050.32 frames. ], batch size: 246, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:55:23,892 INFO [train.py:904] (0/8) Epoch 10, batch 7350, loss[loss=0.2136, simple_loss=0.3071, pruned_loss=0.06008, over 16774.00 frames. ], tot_loss[loss=0.2218, simple_loss=0.3034, pruned_loss=0.07005, over 3067218.42 frames. ], batch size: 83, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:19,127 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.271e+02 3.294e+02 3.832e+02 4.622e+02 6.263e+02, threshold=7.664e+02, percent-clipped=0.0 2023-04-29 06:56:41,003 INFO [train.py:904] (0/8) Epoch 10, batch 7400, loss[loss=0.1989, simple_loss=0.2883, pruned_loss=0.05477, over 16482.00 frames. ], tot_loss[loss=0.2221, simple_loss=0.3041, pruned_loss=0.07, over 3088084.24 frames. ], batch size: 68, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:56:57,050 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=98762.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:03,094 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:57:29,119 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4651, 3.4684, 2.8648, 2.0825, 2.3970, 2.2589, 3.5796, 3.2995], device='cuda:0'), covar=tensor([0.2652, 0.0591, 0.1415, 0.2227, 0.2155, 0.1684, 0.0443, 0.0954], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0256, 0.0282, 0.0276, 0.0280, 0.0217, 0.0265, 0.0288], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 06:57:59,098 INFO [train.py:904] (0/8) Epoch 10, batch 7450, loss[loss=0.2214, simple_loss=0.3059, pruned_loss=0.06844, over 16901.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3051, pruned_loss=0.07036, over 3116350.59 frames. ], batch size: 116, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:58:14,252 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=98810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 06:58:57,291 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.965e+02 3.303e+02 3.942e+02 4.934e+02 1.157e+03, threshold=7.883e+02, percent-clipped=3.0 2023-04-29 06:59:19,899 INFO [train.py:904] (0/8) Epoch 10, batch 7500, loss[loss=0.2008, simple_loss=0.2762, pruned_loss=0.06269, over 16384.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3058, pruned_loss=0.06999, over 3112472.79 frames. ], batch size: 35, lr: 6.77e-03, grad_scale: 2.0 2023-04-29 06:59:52,962 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5432, 2.6668, 2.4402, 3.9614, 3.0969, 3.9911, 1.4246, 2.6942], device='cuda:0'), covar=tensor([0.1491, 0.0699, 0.1260, 0.0156, 0.0350, 0.0392, 0.1646, 0.0940], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0156, 0.0179, 0.0135, 0.0200, 0.0205, 0.0178, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 07:00:19,953 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:22,953 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7152, 1.7188, 1.4602, 1.4269, 1.8069, 1.5375, 1.6610, 1.8870], device='cuda:0'), covar=tensor([0.0113, 0.0239, 0.0318, 0.0278, 0.0164, 0.0217, 0.0140, 0.0169], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0193, 0.0191, 0.0191, 0.0192, 0.0195, 0.0195, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:00:27,386 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98895.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:34,225 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=98899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:00:38,285 INFO [train.py:904] (0/8) Epoch 10, batch 7550, loss[loss=0.2816, simple_loss=0.334, pruned_loss=0.1146, over 11630.00 frames. ], tot_loss[loss=0.2229, simple_loss=0.3052, pruned_loss=0.07035, over 3107473.73 frames. ], batch size: 246, lr: 6.76e-03, grad_scale: 2.0 2023-04-29 07:01:32,286 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.425e+02 3.280e+02 3.927e+02 5.001e+02 7.830e+02, threshold=7.854e+02, percent-clipped=0.0 2023-04-29 07:01:53,317 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98951.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:01:53,982 INFO [train.py:904] (0/8) Epoch 10, batch 7600, loss[loss=0.212, simple_loss=0.2911, pruned_loss=0.06646, over 16984.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.3041, pruned_loss=0.07017, over 3128339.44 frames. ], batch size: 55, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:02:01,642 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=98956.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:02:23,750 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=98971.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:03:11,989 INFO [train.py:904] (0/8) Epoch 10, batch 7650, loss[loss=0.2224, simple_loss=0.3071, pruned_loss=0.06891, over 16879.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3052, pruned_loss=0.07135, over 3121010.39 frames. ], batch size: 109, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:03:12,350 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0973, 5.3950, 5.1048, 5.1511, 4.8088, 4.7384, 4.8733, 5.4642], device='cuda:0'), covar=tensor([0.0978, 0.0736, 0.0923, 0.0667, 0.0740, 0.0756, 0.0938, 0.0769], device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0638, 0.0535, 0.0443, 0.0399, 0.0423, 0.0534, 0.0485], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:03:26,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0641, 2.5519, 2.6215, 1.8947, 2.7611, 2.8449, 2.3430, 2.4233], device='cuda:0'), covar=tensor([0.0638, 0.0207, 0.0187, 0.0849, 0.0086, 0.0176, 0.0393, 0.0393], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0099, 0.0085, 0.0137, 0.0068, 0.0096, 0.0120, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 07:03:44,307 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7097, 2.4407, 2.2313, 3.3524, 2.4117, 3.6137, 1.3504, 2.6740], device='cuda:0'), covar=tensor([0.1278, 0.0667, 0.1112, 0.0138, 0.0171, 0.0362, 0.1570, 0.0761], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0136, 0.0202, 0.0207, 0.0179, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 07:03:59,249 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:08,573 INFO [optim.py:368] (0/8) 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 07:04:28,277 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:04:29,019 INFO [train.py:904] (0/8) Epoch 10, batch 7700, loss[loss=0.2327, simple_loss=0.3118, pruned_loss=0.07676, over 16617.00 frames. ], tot_loss[loss=0.225, simple_loss=0.3055, pruned_loss=0.07223, over 3108263.46 frames. ], batch size: 62, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:04:50,207 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:05:32,241 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9871, 3.9932, 3.9206, 3.2465, 3.9706, 1.6411, 3.7241, 3.5271], device='cuda:0'), covar=tensor([0.0098, 0.0079, 0.0135, 0.0319, 0.0075, 0.2486, 0.0112, 0.0211], device='cuda:0'), in_proj_covar=tensor([0.0122, 0.0107, 0.0155, 0.0150, 0.0126, 0.0173, 0.0142, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:05:43,707 INFO [train.py:904] (0/8) Epoch 10, batch 7750, loss[loss=0.2193, simple_loss=0.3033, pruned_loss=0.06772, over 16496.00 frames. ], tot_loss[loss=0.2266, simple_loss=0.3066, pruned_loss=0.07327, over 3071725.46 frames. ], batch size: 75, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:05:51,255 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8892, 2.0155, 2.3104, 3.1106, 2.0799, 2.1909, 2.1857, 2.1011], device='cuda:0'), covar=tensor([0.0984, 0.2785, 0.1841, 0.0575, 0.3612, 0.2150, 0.2670, 0.2745], device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0385, 0.0321, 0.0319, 0.0413, 0.0436, 0.0346, 0.0450], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:05:52,398 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6045, 2.2198, 2.2163, 4.2601, 2.1443, 2.6572, 2.2740, 2.4486], device='cuda:0'), covar=tensor([0.0822, 0.2939, 0.2138, 0.0369, 0.3443, 0.2008, 0.2747, 0.2700], device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0385, 0.0321, 0.0319, 0.0413, 0.0436, 0.0346, 0.0450], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:05:59,719 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99112.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:06:02,450 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99114.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:06:06,057 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 07:06:38,061 INFO [optim.py:368] (0/8) 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 07:06:59,339 INFO [train.py:904] (0/8) Epoch 10, batch 7800, loss[loss=0.1994, simple_loss=0.2889, pruned_loss=0.05492, over 16905.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3074, pruned_loss=0.07374, over 3070216.66 frames. ], batch size: 109, lr: 6.76e-03, grad_scale: 4.0 2023-04-29 07:07:05,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6258, 2.6006, 1.7753, 2.6751, 2.1127, 2.7672, 2.0042, 2.3621], device='cuda:0'), covar=tensor([0.0236, 0.0335, 0.1168, 0.0162, 0.0607, 0.0483, 0.1099, 0.0549], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0163, 0.0187, 0.0120, 0.0165, 0.0204, 0.0192, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 07:08:06,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4496, 3.4706, 1.7399, 3.7791, 2.4344, 3.8062, 1.8314, 2.6614], device='cuda:0'), covar=tensor([0.0210, 0.0354, 0.1922, 0.0128, 0.0883, 0.0558, 0.1702, 0.0769], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0162, 0.0187, 0.0120, 0.0165, 0.0204, 0.0192, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 07:08:12,311 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99199.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:08:16,580 INFO [train.py:904] (0/8) Epoch 10, batch 7850, loss[loss=0.2291, simple_loss=0.308, pruned_loss=0.07511, over 16396.00 frames. ], tot_loss[loss=0.2279, simple_loss=0.3079, pruned_loss=0.07394, over 3061258.90 frames. ], batch size: 146, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:08:52,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2900, 5.2229, 5.0768, 4.7522, 4.6159, 5.0549, 5.1441, 4.6961], device='cuda:0'), covar=tensor([0.0495, 0.0325, 0.0237, 0.0256, 0.1112, 0.0369, 0.0211, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0295, 0.0272, 0.0252, 0.0294, 0.0289, 0.0188, 0.0318], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:08:57,347 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99229.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:10,228 INFO [optim.py:368] (0/8) 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 07:09:22,402 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99246.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:23,682 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99247.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:29,668 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99251.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:09:30,443 INFO [train.py:904] (0/8) Epoch 10, batch 7900, loss[loss=0.2432, simple_loss=0.328, pruned_loss=0.07923, over 15502.00 frames. ], tot_loss[loss=0.2259, simple_loss=0.3065, pruned_loss=0.07267, over 3065139.95 frames. ], batch size: 191, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:09:55,532 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:10:31,512 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 7950, loss[loss=0.2116, simple_loss=0.2937, pruned_loss=0.06472, over 16674.00 frames. ], tot_loss[loss=0.2262, simple_loss=0.3068, pruned_loss=0.07275, over 3078788.97 frames. ], batch size: 57, lr: 6.75e-03, grad_scale: 4.0 2023-04-29 07:11:22,456 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99324.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:27,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99327.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:30,530 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:11:43,786 INFO [optim.py:368] (0/8) 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 07:12:06,035 INFO [train.py:904] (0/8) Epoch 10, batch 8000, loss[loss=0.1927, simple_loss=0.2887, pruned_loss=0.04841, over 16858.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3069, pruned_loss=0.07341, over 3071663.92 frames. ], batch size: 96, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:12:56,579 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1968, 4.1227, 4.0985, 3.4121, 4.1278, 1.6458, 3.9340, 3.8460], device='cuda:0'), covar=tensor([0.0100, 0.0080, 0.0136, 0.0323, 0.0080, 0.2442, 0.0112, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0123, 0.0108, 0.0156, 0.0151, 0.0127, 0.0174, 0.0142, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:12:56,611 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99385.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:21,642 INFO [train.py:904] (0/8) Epoch 10, batch 8050, loss[loss=0.2129, simple_loss=0.3001, pruned_loss=0.06285, over 16757.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3064, pruned_loss=0.07241, over 3100949.90 frames. ], batch size: 124, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:13:29,872 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:13:50,471 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:14:01,613 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5818, 2.6342, 1.7855, 2.6502, 2.1681, 2.7773, 2.0761, 2.3159], device='cuda:0'), covar=tensor([0.0205, 0.0311, 0.1106, 0.0169, 0.0602, 0.0490, 0.0987, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0162, 0.0187, 0.0120, 0.0166, 0.0205, 0.0193, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 07:14:18,157 INFO [optim.py:368] (0/8) 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 07:14:39,584 INFO [train.py:904] (0/8) Epoch 10, batch 8100, loss[loss=0.2096, simple_loss=0.2958, pruned_loss=0.06168, over 17045.00 frames. ], tot_loss[loss=0.2249, simple_loss=0.3061, pruned_loss=0.07181, over 3104774.68 frames. ], batch size: 55, lr: 6.75e-03, grad_scale: 8.0 2023-04-29 07:15:03,440 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99468.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:15:22,577 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 8150, loss[loss=0.1779, simple_loss=0.2648, pruned_loss=0.04548, over 17021.00 frames. ], tot_loss[loss=0.2219, simple_loss=0.3034, pruned_loss=0.0702, over 3121028.41 frames. ], batch size: 50, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:16:35,926 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99529.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:16:49,693 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.193e+02 2.958e+02 3.724e+02 4.411e+02 7.985e+02, threshold=7.447e+02, percent-clipped=3.0 2023-04-29 07:17:00,884 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99546.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:08,247 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:09,688 INFO [train.py:904] (0/8) Epoch 10, batch 8200, loss[loss=0.1965, simple_loss=0.2795, pruned_loss=0.05674, over 16413.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3006, pruned_loss=0.06924, over 3139409.24 frames. ], batch size: 68, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:17:35,547 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99568.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:17:51,240 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 07:18:04,670 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99585.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:11,204 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99589.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:20,131 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:28,355 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99599.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:18:32,974 INFO [train.py:904] (0/8) Epoch 10, batch 8250, loss[loss=0.1889, simple_loss=0.2837, pruned_loss=0.04708, over 15526.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.3, pruned_loss=0.06735, over 3110821.86 frames. ], batch size: 191, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:19:10,196 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:15,994 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99627.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:19,851 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99629.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:19:36,139 INFO [optim.py:368] (0/8) 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 07:19:55,547 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 8300, loss[loss=0.1838, simple_loss=0.2619, pruned_loss=0.05287, over 11986.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2973, pruned_loss=0.06457, over 3093796.71 frames. ], batch size: 247, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:20:25,224 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-29 07:20:37,236 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,563 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:20:44,721 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=99680.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:21:21,643 INFO [train.py:904] (0/8) Epoch 10, batch 8350, loss[loss=0.206, simple_loss=0.2984, pruned_loss=0.05676, over 16763.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.296, pruned_loss=0.06217, over 3094252.33 frames. ], batch size: 124, lr: 6.74e-03, grad_scale: 4.0 2023-04-29 07:21:29,891 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:18,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7795, 3.7441, 4.0995, 4.0852, 4.0727, 3.8618, 3.8321, 3.7987], device='cuda:0'), covar=tensor([0.0302, 0.0611, 0.0384, 0.0399, 0.0434, 0.0370, 0.0832, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0321, 0.0319, 0.0306, 0.0369, 0.0340, 0.0438, 0.0277], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 07:22:21,851 INFO [optim.py:368] (0/8) 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 07:22:26,029 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=99741.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:22:43,907 INFO [train.py:904] (0/8) Epoch 10, batch 8400, loss[loss=0.1859, simple_loss=0.2848, pruned_loss=0.0435, over 16557.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2934, pruned_loss=0.06033, over 3075117.45 frames. ], batch size: 62, lr: 6.74e-03, grad_scale: 8.0 2023-04-29 07:22:49,353 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99755.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:23:25,434 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 8450, loss[loss=0.1899, simple_loss=0.2792, pruned_loss=0.0503, over 15344.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2915, pruned_loss=0.05851, over 3081744.22 frames. ], batch size: 190, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:24:42,085 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99824.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:25:06,074 INFO [optim.py:368] (0/8) 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 07:25:25,950 INFO [train.py:904] (0/8) Epoch 10, batch 8500, loss[loss=0.1705, simple_loss=0.2518, pruned_loss=0.04462, over 11905.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2873, pruned_loss=0.05622, over 3058512.89 frames. ], batch size: 248, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:26:21,593 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:26:50,576 INFO [train.py:904] (0/8) Epoch 10, batch 8550, loss[loss=0.2033, simple_loss=0.2768, pruned_loss=0.06491, over 11815.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2843, pruned_loss=0.05499, over 3026109.54 frames. ], batch size: 247, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:27:19,944 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5788, 3.0991, 3.1614, 1.7435, 2.6475, 2.2598, 3.1238, 3.2513], device='cuda:0'), covar=tensor([0.0295, 0.0637, 0.0522, 0.1842, 0.0799, 0.0907, 0.0706, 0.0742], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0135, 0.0150, 0.0137, 0.0130, 0.0121, 0.0131, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 07:27:34,877 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:34,990 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:27:50,550 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:02,248 INFO [optim.py:368] (0/8) 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 07:28:16,661 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=99945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:28:22,283 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8318, 1.2373, 1.5052, 1.7485, 1.7784, 1.7933, 1.4771, 1.8055], device='cuda:0'), covar=tensor([0.0152, 0.0298, 0.0155, 0.0180, 0.0188, 0.0139, 0.0290, 0.0080], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0166, 0.0148, 0.0151, 0.0161, 0.0115, 0.0166, 0.0105], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 07:28:29,560 INFO [train.py:904] (0/8) Epoch 10, batch 8600, loss[loss=0.1756, simple_loss=0.2647, pruned_loss=0.04328, over 16611.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2844, pruned_loss=0.054, over 3001046.27 frames. ], batch size: 57, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:29:07,000 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6817, 3.2074, 3.3441, 1.9839, 2.8528, 2.2590, 3.1754, 3.3757], device='cuda:0'), covar=tensor([0.0262, 0.0642, 0.0435, 0.1623, 0.0666, 0.0894, 0.0662, 0.0757], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0135, 0.0150, 0.0136, 0.0130, 0.0121, 0.0131, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 07:29:10,993 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=99972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:29:26,008 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=99980.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:30:03,523 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-100000.pt 2023-04-29 07:30:11,027 INFO [train.py:904] (0/8) Epoch 10, batch 8650, loss[loss=0.1767, simple_loss=0.2704, pruned_loss=0.0415, over 15296.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2823, pruned_loss=0.05233, over 3016135.38 frames. ], batch size: 191, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:30:24,878 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100007.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 07:30:34,504 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9831, 4.0093, 3.9620, 3.4192, 3.9233, 1.6913, 3.7656, 3.6929], device='cuda:0'), covar=tensor([0.0087, 0.0080, 0.0109, 0.0235, 0.0081, 0.2316, 0.0110, 0.0189], device='cuda:0'), in_proj_covar=tensor([0.0120, 0.0106, 0.0150, 0.0143, 0.0123, 0.0170, 0.0138, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:31:10,432 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100028.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:31:27,307 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:31:31,868 INFO [optim.py:368] (0/8) 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 07:31:56,454 INFO [train.py:904] (0/8) Epoch 10, batch 8700, loss[loss=0.1677, simple_loss=0.2577, pruned_loss=0.03882, over 16780.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2795, pruned_loss=0.05049, over 3045496.89 frames. ], batch size: 83, lr: 6.73e-03, grad_scale: 8.0 2023-04-29 07:32:28,005 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100068.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:32:42,794 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100076.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:33:35,366 INFO [train.py:904] (0/8) Epoch 10, batch 8750, loss[loss=0.2082, simple_loss=0.2961, pruned_loss=0.0601, over 16223.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2792, pruned_loss=0.04967, over 3060507.38 frames. ], batch size: 165, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:34:32,198 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:34:32,247 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100124.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:35:02,704 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.530e+02 3.072e+02 4.161e+02 6.852e+02, threshold=6.144e+02, percent-clipped=8.0 2023-04-29 07:35:29,751 INFO [train.py:904] (0/8) Epoch 10, batch 8800, loss[loss=0.1864, simple_loss=0.2734, pruned_loss=0.04973, over 12393.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2783, pruned_loss=0.04903, over 3048266.40 frames. ], batch size: 247, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:36:11,842 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100172.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:37:15,889 INFO [train.py:904] (0/8) Epoch 10, batch 8850, loss[loss=0.1852, simple_loss=0.2863, pruned_loss=0.04208, over 15347.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2804, pruned_loss=0.0479, over 3050648.51 frames. ], batch size: 191, lr: 6.72e-03, grad_scale: 8.0 2023-04-29 07:38:03,471 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:34,790 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100238.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:38:36,936 INFO [optim.py:368] (0/8) 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 07:38:48,395 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100245.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:39:01,684 INFO [train.py:904] (0/8) Epoch 10, batch 8900, loss[loss=0.2017, simple_loss=0.2909, pruned_loss=0.05623, over 15320.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2813, pruned_loss=0.04761, over 3050323.06 frames. ], batch size: 191, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:39:41,046 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4651, 4.3862, 4.2300, 3.7900, 4.3564, 1.6135, 4.0989, 4.2242], device='cuda:0'), covar=tensor([0.0082, 0.0080, 0.0169, 0.0294, 0.0086, 0.2310, 0.0126, 0.0173], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0106, 0.0148, 0.0142, 0.0122, 0.0171, 0.0138, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:39:43,461 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:40:45,188 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100293.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:01,083 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:41:05,116 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-29 07:41:05,927 INFO [train.py:904] (0/8) Epoch 10, batch 8950, loss[loss=0.178, simple_loss=0.2688, pruned_loss=0.04364, over 16704.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.28, pruned_loss=0.0474, over 3064700.51 frames. ], batch size: 134, lr: 6.72e-03, grad_scale: 4.0 2023-04-29 07:42:21,922 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:42:29,138 INFO [optim.py:368] (0/8) 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 07:42:55,816 INFO [train.py:904] (0/8) Epoch 10, batch 9000, loss[loss=0.1846, simple_loss=0.2692, pruned_loss=0.05, over 12331.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2767, pruned_loss=0.04598, over 3051346.14 frames. ], batch size: 248, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:42:55,818 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 07:43:05,470 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 07:43:16,710 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4422, 3.4657, 2.7504, 2.0393, 2.2645, 2.1793, 3.6075, 3.2586], device='cuda:0'), covar=tensor([0.2508, 0.0638, 0.1475, 0.2337, 0.2359, 0.1813, 0.0424, 0.0951], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0246, 0.0271, 0.0265, 0.0261, 0.0212, 0.0255, 0.0276], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:43:29,848 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100363.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:43:43,201 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4503, 3.5658, 3.3261, 3.1322, 3.0200, 3.4846, 3.2590, 3.2372], device='cuda:0'), covar=tensor([0.0619, 0.0448, 0.0317, 0.0259, 0.0678, 0.0424, 0.1144, 0.0516], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0282, 0.0263, 0.0244, 0.0285, 0.0279, 0.0182, 0.0306], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:44:12,446 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:44:50,493 INFO [train.py:904] (0/8) Epoch 10, batch 9050, loss[loss=0.201, simple_loss=0.2954, pruned_loss=0.05331, over 16270.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2775, pruned_loss=0.04677, over 3054356.01 frames. ], batch size: 146, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:44:51,419 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:45:39,632 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 07:46:08,565 INFO [optim.py:368] (0/8) 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 07:46:21,171 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-29 07:46:37,394 INFO [train.py:904] (0/8) Epoch 10, batch 9100, loss[loss=0.1772, simple_loss=0.274, pruned_loss=0.04021, over 16539.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2765, pruned_loss=0.04693, over 3058063.33 frames. ], batch size: 68, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:46:58,870 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100463.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:48:34,518 INFO [train.py:904] (0/8) Epoch 10, batch 9150, loss[loss=0.1899, simple_loss=0.2775, pruned_loss=0.0511, over 16956.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2771, pruned_loss=0.0467, over 3051129.95 frames. ], batch size: 109, lr: 6.71e-03, grad_scale: 4.0 2023-04-29 07:49:54,594 INFO [optim.py:368] (0/8) 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 07:50:13,994 INFO [train.py:904] (0/8) Epoch 10, batch 9200, loss[loss=0.1652, simple_loss=0.2459, pruned_loss=0.04228, over 12314.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2731, pruned_loss=0.04588, over 3040184.60 frames. ], batch size: 248, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:51:35,130 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100594.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 07:51:50,968 INFO [train.py:904] (0/8) Epoch 10, batch 9250, loss[loss=0.1688, simple_loss=0.2622, pruned_loss=0.03771, over 16946.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2727, pruned_loss=0.04589, over 3043720.55 frames. ], batch size: 116, lr: 6.71e-03, grad_scale: 8.0 2023-04-29 07:53:14,008 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.494e+02 2.424e+02 2.858e+02 3.550e+02 5.723e+02, threshold=5.716e+02, percent-clipped=3.0 2023-04-29 07:53:42,543 INFO [train.py:904] (0/8) Epoch 10, batch 9300, loss[loss=0.1748, simple_loss=0.2585, pruned_loss=0.04552, over 16720.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2713, pruned_loss=0.04549, over 3038718.33 frames. ], batch size: 57, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:53:57,456 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7606, 3.8011, 3.8936, 3.8059, 3.8902, 4.2668, 3.9384, 3.7075], device='cuda:0'), covar=tensor([0.2096, 0.1886, 0.1965, 0.2272, 0.2613, 0.1550, 0.1398, 0.2594], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0442, 0.0477, 0.0385, 0.0504, 0.0506, 0.0387, 0.0520], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-29 07:54:09,135 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 10, batch 9350, loss[loss=0.1719, simple_loss=0.2671, pruned_loss=0.03836, over 16659.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2712, pruned_loss=0.04567, over 3032064.35 frames. ], batch size: 89, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:55:50,101 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100711.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:55:50,542 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-29 07:56:23,379 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9972, 2.3179, 2.3933, 2.9178, 2.1550, 3.3225, 1.6763, 2.8792], device='cuda:0'), covar=tensor([0.1142, 0.0486, 0.0916, 0.0109, 0.0081, 0.0355, 0.1272, 0.0565], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0153, 0.0177, 0.0132, 0.0185, 0.0201, 0.0178, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 07:56:28,460 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=100730.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 07:56:48,376 INFO [optim.py:368] (0/8) 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 07:57:12,457 INFO [train.py:904] (0/8) Epoch 10, batch 9400, loss[loss=0.1821, simple_loss=0.2891, pruned_loss=0.03753, over 16246.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2709, pruned_loss=0.04549, over 3015615.58 frames. ], batch size: 165, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 07:57:25,172 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=100758.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 07:58:32,476 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=100791.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 07:58:52,911 INFO [train.py:904] (0/8) Epoch 10, batch 9450, loss[loss=0.1621, simple_loss=0.2597, pruned_loss=0.03224, over 16778.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2727, pruned_loss=0.04558, over 3017200.54 frames. ], batch size: 124, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:00:08,496 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 08:00:10,875 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.411e+02 3.092e+02 3.645e+02 7.379e+02, threshold=6.183e+02, percent-clipped=4.0 2023-04-29 08:00:34,630 INFO [train.py:904] (0/8) Epoch 10, batch 9500, loss[loss=0.1661, simple_loss=0.2611, pruned_loss=0.0355, over 16252.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2724, pruned_loss=0.04527, over 3026146.14 frames. ], batch size: 166, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:02:03,014 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=100894.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:02:20,736 INFO [train.py:904] (0/8) Epoch 10, batch 9550, loss[loss=0.2009, simple_loss=0.2957, pruned_loss=0.053, over 15311.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2714, pruned_loss=0.04551, over 3018016.41 frames. ], batch size: 191, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:03:04,879 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0497, 2.2896, 1.8944, 2.1159, 2.6711, 2.3644, 2.9033, 2.8891], device='cuda:0'), covar=tensor([0.0081, 0.0313, 0.0372, 0.0360, 0.0219, 0.0296, 0.0121, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0190, 0.0187, 0.0185, 0.0187, 0.0191, 0.0183, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:03:40,252 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.820e+02 2.447e+02 2.728e+02 3.779e+02 7.191e+02, threshold=5.455e+02, percent-clipped=3.0 2023-04-29 08:03:45,436 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=100942.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:04:02,638 INFO [train.py:904] (0/8) Epoch 10, batch 9600, loss[loss=0.2014, simple_loss=0.3024, pruned_loss=0.05018, over 15364.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2734, pruned_loss=0.04631, over 3008134.50 frames. ], batch size: 191, lr: 6.70e-03, grad_scale: 8.0 2023-04-29 08:05:42,860 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8066, 2.1047, 1.7228, 1.9222, 2.4930, 2.1706, 2.6036, 2.6526], device='cuda:0'), covar=tensor([0.0092, 0.0294, 0.0369, 0.0334, 0.0193, 0.0277, 0.0132, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0193, 0.0191, 0.0188, 0.0190, 0.0193, 0.0186, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:05:52,890 INFO [train.py:904] (0/8) Epoch 10, batch 9650, loss[loss=0.1792, simple_loss=0.2767, pruned_loss=0.04087, over 16954.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2753, pruned_loss=0.04644, over 3004043.42 frames. ], batch size: 116, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:07:14,889 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9923, 3.2232, 2.9938, 5.1105, 4.0120, 4.5648, 1.6052, 3.5117], device='cuda:0'), covar=tensor([0.1222, 0.0603, 0.0904, 0.0092, 0.0180, 0.0272, 0.1458, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0151, 0.0175, 0.0130, 0.0180, 0.0197, 0.0176, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 08:07:15,511 INFO [optim.py:368] (0/8) 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 08:07:37,313 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7498, 3.6613, 3.9007, 3.7641, 3.7998, 4.2491, 3.9110, 3.6458], device='cuda:0'), covar=tensor([0.2114, 0.2188, 0.1962, 0.2166, 0.2971, 0.1537, 0.1393, 0.2517], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0442, 0.0474, 0.0380, 0.0502, 0.0504, 0.0383, 0.0516], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:07:41,464 INFO [train.py:904] (0/8) Epoch 10, batch 9700, loss[loss=0.1701, simple_loss=0.2596, pruned_loss=0.04032, over 17029.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2737, pruned_loss=0.04603, over 2998180.43 frames. ], batch size: 55, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:07:47,480 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4182, 4.6505, 4.3928, 4.4419, 4.1380, 4.2143, 4.1725, 4.6484], device='cuda:0'), covar=tensor([0.0851, 0.0880, 0.1078, 0.0655, 0.0798, 0.1233, 0.0922, 0.0942], device='cuda:0'), in_proj_covar=tensor([0.0483, 0.0605, 0.0499, 0.0420, 0.0382, 0.0400, 0.0507, 0.0463], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:07:52,553 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:08:53,878 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101086.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 08:09:25,203 INFO [train.py:904] (0/8) Epoch 10, batch 9750, loss[loss=0.1803, simple_loss=0.2766, pruned_loss=0.04197, over 16274.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.273, pruned_loss=0.04637, over 3001626.68 frames. ], batch size: 165, lr: 6.69e-03, grad_scale: 8.0 2023-04-29 08:09:32,434 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:09:34,757 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101107.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:10:45,038 INFO [optim.py:368] (0/8) 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 08:11:05,045 INFO [train.py:904] (0/8) Epoch 10, batch 9800, loss[loss=0.1904, simple_loss=0.2917, pruned_loss=0.04454, over 16394.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2731, pruned_loss=0.04527, over 3025053.03 frames. ], batch size: 146, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:11:34,515 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9169, 4.2098, 4.0485, 4.0462, 3.7330, 3.8084, 3.8295, 4.1941], device='cuda:0'), covar=tensor([0.0924, 0.0814, 0.0781, 0.0562, 0.0705, 0.1400, 0.0871, 0.0890], device='cuda:0'), in_proj_covar=tensor([0.0485, 0.0607, 0.0499, 0.0423, 0.0384, 0.0401, 0.0508, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:11:36,081 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:12:49,201 INFO [train.py:904] (0/8) Epoch 10, batch 9850, loss[loss=0.1999, simple_loss=0.2791, pruned_loss=0.06032, over 12499.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.274, pruned_loss=0.04513, over 3017487.04 frames. ], batch size: 248, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:14:17,930 INFO [optim.py:368] (0/8) 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 08:14:41,632 INFO [train.py:904] (0/8) Epoch 10, batch 9900, loss[loss=0.1807, simple_loss=0.2645, pruned_loss=0.04846, over 12624.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2742, pruned_loss=0.04497, over 3021170.62 frames. ], batch size: 246, lr: 6.69e-03, grad_scale: 4.0 2023-04-29 08:14:50,716 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2986, 2.0022, 2.0423, 3.8317, 1.9557, 2.3351, 2.1311, 2.1178], device='cuda:0'), covar=tensor([0.0813, 0.3340, 0.2343, 0.0366, 0.3832, 0.2334, 0.3053, 0.3478], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0369, 0.0312, 0.0304, 0.0398, 0.0412, 0.0335, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:15:44,273 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0144, 4.0070, 3.9284, 3.4647, 3.9038, 1.6583, 3.7641, 3.6334], device='cuda:0'), covar=tensor([0.0078, 0.0076, 0.0114, 0.0204, 0.0085, 0.2268, 0.0098, 0.0200], device='cuda:0'), in_proj_covar=tensor([0.0119, 0.0105, 0.0148, 0.0136, 0.0122, 0.0170, 0.0135, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-29 08:15:51,893 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6150, 4.4352, 4.6649, 4.8129, 4.9741, 4.4881, 4.9783, 4.9820], device='cuda:0'), covar=tensor([0.1352, 0.0956, 0.1301, 0.0576, 0.0492, 0.0749, 0.0392, 0.0513], device='cuda:0'), in_proj_covar=tensor([0.0464, 0.0580, 0.0699, 0.0596, 0.0454, 0.0451, 0.0467, 0.0530], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:16:40,052 INFO [train.py:904] (0/8) Epoch 10, batch 9950, loss[loss=0.2001, simple_loss=0.3024, pruned_loss=0.04887, over 16782.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2765, pruned_loss=0.04489, over 3039421.70 frames. ], batch size: 134, lr: 6.68e-03, grad_scale: 4.0 2023-04-29 08:18:13,670 INFO [optim.py:368] (0/8) 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 08:18:42,236 INFO [train.py:904] (0/8) Epoch 10, batch 10000, loss[loss=0.1608, simple_loss=0.2528, pruned_loss=0.03447, over 16620.00 frames. ], tot_loss[loss=0.182, simple_loss=0.275, pruned_loss=0.04451, over 3051105.15 frames. ], batch size: 62, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:19:30,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 08:19:54,106 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101386.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:20:09,659 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9646, 3.9691, 4.3715, 4.3576, 4.3695, 4.1168, 4.0859, 3.9859], device='cuda:0'), covar=tensor([0.0287, 0.0481, 0.0403, 0.0425, 0.0406, 0.0330, 0.0741, 0.0400], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0304, 0.0307, 0.0293, 0.0351, 0.0323, 0.0411, 0.0264], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0002], device='cuda:0') 2023-04-29 08:20:19,637 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1965, 1.9301, 1.9909, 3.7006, 1.9530, 2.3487, 2.1201, 2.1282], device='cuda:0'), covar=tensor([0.0873, 0.3281, 0.2184, 0.0376, 0.3716, 0.2209, 0.2839, 0.3207], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0368, 0.0312, 0.0303, 0.0396, 0.0411, 0.0333, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:20:23,607 INFO [train.py:904] (0/8) Epoch 10, batch 10050, loss[loss=0.2021, simple_loss=0.2951, pruned_loss=0.05454, over 16226.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2759, pruned_loss=0.04457, over 3067947.83 frames. ], batch size: 165, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:21:23,828 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-29 08:21:25,330 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101434.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:21:36,391 INFO [optim.py:368] (0/8) 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 08:21:56,150 INFO [train.py:904] (0/8) Epoch 10, batch 10100, loss[loss=0.1816, simple_loss=0.262, pruned_loss=0.05058, over 12433.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2765, pruned_loss=0.04512, over 3060045.28 frames. ], batch size: 246, lr: 6.68e-03, grad_scale: 8.0 2023-04-29 08:22:15,353 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=101463.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:23:15,344 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-10.pt 2023-04-29 08:23:38,430 INFO [train.py:904] (0/8) Epoch 11, batch 0, loss[loss=0.1996, simple_loss=0.2899, pruned_loss=0.05468, over 17264.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2899, pruned_loss=0.05468, over 17264.00 frames. ], batch size: 52, lr: 6.37e-03, grad_scale: 8.0 2023-04-29 08:23:38,431 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 08:23:45,829 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 08:24:43,130 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.588e+02 2.573e+02 2.991e+02 3.932e+02 8.479e+02, threshold=5.981e+02, percent-clipped=4.0 2023-04-29 08:24:53,232 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4883, 4.2764, 4.4363, 4.6775, 4.8213, 4.3697, 4.6799, 4.7607], device='cuda:0'), covar=tensor([0.2003, 0.1610, 0.1976, 0.0936, 0.0795, 0.0980, 0.1349, 0.1488], device='cuda:0'), in_proj_covar=tensor([0.0481, 0.0599, 0.0724, 0.0617, 0.0467, 0.0465, 0.0483, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:24:55,132 INFO [train.py:904] (0/8) Epoch 11, batch 50, loss[loss=0.217, simple_loss=0.29, pruned_loss=0.07197, over 16275.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2903, pruned_loss=0.06661, over 750261.34 frames. ], batch size: 165, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:26:05,598 INFO [train.py:904] (0/8) Epoch 11, batch 100, loss[loss=0.2051, simple_loss=0.2832, pruned_loss=0.06351, over 15541.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2831, pruned_loss=0.06159, over 1316113.01 frames. ], batch size: 190, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:26:35,050 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3045, 3.4144, 1.8874, 3.5339, 2.4851, 3.5441, 2.0594, 2.7390], device='cuda:0'), covar=tensor([0.0233, 0.0337, 0.1546, 0.0216, 0.0804, 0.0592, 0.1362, 0.0674], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0157, 0.0185, 0.0120, 0.0164, 0.0196, 0.0193, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 08:27:03,338 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.860e+02 2.523e+02 3.127e+02 3.796e+02 9.987e+02, threshold=6.254e+02, percent-clipped=2.0 2023-04-29 08:27:12,720 INFO [train.py:904] (0/8) Epoch 11, batch 150, loss[loss=0.2033, simple_loss=0.2797, pruned_loss=0.06342, over 16716.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2798, pruned_loss=0.0611, over 1765562.42 frames. ], batch size: 134, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:27:58,587 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101684.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:28:23,286 INFO [train.py:904] (0/8) Epoch 11, batch 200, loss[loss=0.1868, simple_loss=0.2749, pruned_loss=0.04934, over 17065.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2791, pruned_loss=0.05982, over 2118034.17 frames. ], batch size: 55, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:21,763 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.429e+02 2.923e+02 3.406e+02 5.474e+02, threshold=5.846e+02, percent-clipped=1.0 2023-04-29 08:29:23,377 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:29:31,794 INFO [train.py:904] (0/8) Epoch 11, batch 250, loss[loss=0.1672, simple_loss=0.2581, pruned_loss=0.03821, over 17225.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2776, pruned_loss=0.0592, over 2386878.63 frames. ], batch size: 45, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:29:46,125 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=101763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:30:17,882 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0693, 5.6529, 5.8599, 5.6263, 5.7428, 6.2219, 5.7908, 5.6219], device='cuda:0'), covar=tensor([0.0814, 0.1919, 0.1690, 0.1948, 0.2453, 0.0897, 0.1236, 0.2032], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0474, 0.0509, 0.0409, 0.0546, 0.0540, 0.0407, 0.0553], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 08:30:37,971 INFO [train.py:904] (0/8) Epoch 11, batch 300, loss[loss=0.1838, simple_loss=0.2726, pruned_loss=0.04749, over 16617.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2742, pruned_loss=0.0563, over 2601376.59 frames. ], batch size: 62, lr: 6.37e-03, grad_scale: 1.0 2023-04-29 08:30:51,060 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=101811.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:31:35,543 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.476e+02 3.003e+02 3.668e+02 8.355e+02, threshold=6.006e+02, percent-clipped=2.0 2023-04-29 08:31:47,613 INFO [train.py:904] (0/8) Epoch 11, batch 350, loss[loss=0.1855, simple_loss=0.2629, pruned_loss=0.05407, over 16505.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2724, pruned_loss=0.05516, over 2754653.29 frames. ], batch size: 68, lr: 6.36e-03, grad_scale: 1.0 2023-04-29 08:32:56,333 INFO [train.py:904] (0/8) Epoch 11, batch 400, loss[loss=0.1538, simple_loss=0.2365, pruned_loss=0.03559, over 16797.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.27, pruned_loss=0.05478, over 2874779.83 frames. ], batch size: 39, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:33:22,122 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101920.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:33:54,607 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.298e+02 2.716e+02 3.213e+02 6.833e+02, threshold=5.433e+02, percent-clipped=1.0 2023-04-29 08:33:56,123 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5403, 3.4259, 2.4172, 2.1875, 2.4206, 2.0047, 3.4352, 3.1346], device='cuda:0'), covar=tensor([0.2518, 0.0650, 0.1828, 0.2220, 0.2385, 0.2130, 0.0643, 0.1257], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0254, 0.0282, 0.0273, 0.0271, 0.0220, 0.0265, 0.0290], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:34:05,272 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 08:34:06,160 INFO [train.py:904] (0/8) Epoch 11, batch 450, loss[loss=0.1833, simple_loss=0.2713, pruned_loss=0.04765, over 17063.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2681, pruned_loss=0.05387, over 2965904.90 frames. ], batch size: 53, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:34:46,993 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6086, 3.6733, 1.8376, 3.9424, 2.6522, 4.0492, 1.9796, 2.8377], device='cuda:0'), covar=tensor([0.0229, 0.0314, 0.1651, 0.0283, 0.0732, 0.0423, 0.1552, 0.0625], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0163, 0.0188, 0.0127, 0.0168, 0.0204, 0.0197, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 08:34:47,011 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=101981.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 08:34:54,622 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=101987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:35:13,835 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-102000.pt 2023-04-29 08:35:19,216 INFO [train.py:904] (0/8) Epoch 11, batch 500, loss[loss=0.1762, simple_loss=0.2638, pruned_loss=0.04437, over 16786.00 frames. ], tot_loss[loss=0.1867, simple_loss=0.267, pruned_loss=0.05325, over 3044697.67 frames. ], batch size: 57, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:35:36,869 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7230, 3.7055, 4.0334, 3.1255, 3.6374, 3.9996, 3.7917, 2.4329], device='cuda:0'), covar=tensor([0.0317, 0.0171, 0.0039, 0.0217, 0.0072, 0.0068, 0.0057, 0.0327], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0068, 0.0069, 0.0125, 0.0077, 0.0086, 0.0076, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 08:36:10,899 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 08:36:13,475 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:19,098 INFO [optim.py:368] (0/8) 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 08:36:24,780 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102048.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:36:29,133 INFO [train.py:904] (0/8) Epoch 11, batch 550, loss[loss=0.1871, simple_loss=0.2783, pruned_loss=0.04793, over 17051.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2659, pruned_loss=0.05199, over 3108011.31 frames. ], batch size: 53, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:37:40,181 INFO [train.py:904] (0/8) Epoch 11, batch 600, loss[loss=0.196, simple_loss=0.2806, pruned_loss=0.05572, over 17125.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2659, pruned_loss=0.0523, over 3156603.45 frames. ], batch size: 49, lr: 6.36e-03, grad_scale: 2.0 2023-04-29 08:38:14,554 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1395, 2.2613, 1.7676, 2.0778, 2.6058, 2.3833, 2.6406, 2.7301], device='cuda:0'), covar=tensor([0.0173, 0.0279, 0.0386, 0.0332, 0.0147, 0.0249, 0.0166, 0.0172], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0203, 0.0200, 0.0198, 0.0202, 0.0203, 0.0204, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:38:23,781 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6065, 5.9261, 5.6862, 5.7241, 5.3024, 5.3699, 5.4018, 6.0109], device='cuda:0'), covar=tensor([0.1081, 0.0868, 0.1029, 0.0691, 0.0831, 0.0633, 0.1050, 0.0859], device='cuda:0'), in_proj_covar=tensor([0.0541, 0.0675, 0.0560, 0.0470, 0.0427, 0.0441, 0.0570, 0.0516], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:38:38,911 INFO [optim.py:368] (0/8) 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 08:38:46,623 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6826, 4.6234, 4.5246, 3.9939, 4.5926, 1.6954, 4.3880, 4.2891], device='cuda:0'), covar=tensor([0.0103, 0.0080, 0.0133, 0.0283, 0.0075, 0.2445, 0.0105, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0112, 0.0159, 0.0147, 0.0129, 0.0177, 0.0145, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:38:49,718 INFO [train.py:904] (0/8) Epoch 11, batch 650, loss[loss=0.1901, simple_loss=0.2663, pruned_loss=0.05691, over 16713.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.265, pruned_loss=0.05283, over 3177448.52 frames. ], batch size: 134, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:38:52,497 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1331, 1.9681, 2.5518, 3.0736, 2.8950, 3.3629, 2.1760, 3.2983], device='cuda:0'), covar=tensor([0.0151, 0.0328, 0.0233, 0.0185, 0.0200, 0.0136, 0.0334, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0156, 0.0158, 0.0169, 0.0122, 0.0172, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 08:39:15,265 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-29 08:39:31,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2902, 3.6467, 3.7807, 2.1913, 3.0156, 2.4577, 3.6610, 3.7744], device='cuda:0'), covar=tensor([0.0301, 0.0760, 0.0486, 0.1773, 0.0813, 0.0951, 0.0614, 0.1072], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0141, 0.0157, 0.0142, 0.0136, 0.0125, 0.0136, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 08:39:58,904 INFO [train.py:904] (0/8) Epoch 11, batch 700, loss[loss=0.1768, simple_loss=0.2554, pruned_loss=0.04914, over 16832.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2646, pruned_loss=0.05238, over 3210483.69 frames. ], batch size: 102, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:40:22,949 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 08:40:57,197 INFO [optim.py:368] (0/8) 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 08:41:09,249 INFO [train.py:904] (0/8) Epoch 11, batch 750, loss[loss=0.1498, simple_loss=0.2334, pruned_loss=0.03313, over 16981.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2647, pruned_loss=0.05193, over 3232722.17 frames. ], batch size: 41, lr: 6.35e-03, grad_scale: 2.0 2023-04-29 08:41:42,364 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102276.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 08:41:53,335 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-04-29 08:42:11,119 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5222, 4.4970, 4.7114, 4.5727, 4.4235, 5.1745, 4.7861, 4.4649], device='cuda:0'), covar=tensor([0.1457, 0.1916, 0.1847, 0.2217, 0.3495, 0.1206, 0.1499, 0.2676], device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0486, 0.0526, 0.0419, 0.0561, 0.0553, 0.0417, 0.0568], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 08:42:18,055 INFO [train.py:904] (0/8) Epoch 11, batch 800, loss[loss=0.1578, simple_loss=0.236, pruned_loss=0.03976, over 17005.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.265, pruned_loss=0.05195, over 3255054.06 frames. ], batch size: 41, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:43:11,924 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102340.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:15,332 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=102343.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:43:16,116 INFO [optim.py:368] (0/8) 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 08:43:18,221 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-04-29 08:43:27,557 INFO [train.py:904] (0/8) Epoch 11, batch 850, loss[loss=0.1756, simple_loss=0.2532, pruned_loss=0.04903, over 16813.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2645, pruned_loss=0.05118, over 3272624.53 frames. ], batch size: 102, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:44:16,960 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:44:37,486 INFO [train.py:904] (0/8) Epoch 11, batch 900, loss[loss=0.1933, simple_loss=0.2868, pruned_loss=0.0499, over 17127.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2636, pruned_loss=0.05027, over 3279887.32 frames. ], batch size: 49, lr: 6.35e-03, grad_scale: 4.0 2023-04-29 08:45:35,220 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.314e+02 2.859e+02 3.467e+02 6.346e+02, threshold=5.718e+02, percent-clipped=1.0 2023-04-29 08:45:45,386 INFO [train.py:904] (0/8) Epoch 11, batch 950, loss[loss=0.1761, simple_loss=0.2513, pruned_loss=0.05044, over 16209.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2636, pruned_loss=0.0499, over 3294548.70 frames. ], batch size: 165, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:46:35,319 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3603, 4.1548, 4.3826, 4.5658, 4.6817, 4.2217, 4.4371, 4.6368], device='cuda:0'), covar=tensor([0.1379, 0.1027, 0.1348, 0.0693, 0.0525, 0.1186, 0.2089, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0529, 0.0663, 0.0816, 0.0682, 0.0514, 0.0514, 0.0529, 0.0602], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:46:45,958 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-29 08:46:54,283 INFO [train.py:904] (0/8) Epoch 11, batch 1000, loss[loss=0.1472, simple_loss=0.2423, pruned_loss=0.02605, over 17115.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2624, pruned_loss=0.05008, over 3284656.86 frames. ], batch size: 47, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:47:40,544 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4093, 1.5396, 2.0554, 2.2375, 2.4102, 2.3348, 1.5661, 2.4224], device='cuda:0'), covar=tensor([0.0152, 0.0351, 0.0202, 0.0206, 0.0192, 0.0170, 0.0371, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0156, 0.0157, 0.0168, 0.0122, 0.0172, 0.0112], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 08:47:52,041 INFO [optim.py:368] (0/8) 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 08:48:02,256 INFO [train.py:904] (0/8) Epoch 11, batch 1050, loss[loss=0.1955, simple_loss=0.2799, pruned_loss=0.05561, over 17067.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.262, pruned_loss=0.05016, over 3296640.93 frames. ], batch size: 53, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:48:17,210 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0483, 4.0257, 3.9028, 3.6176, 3.6105, 3.9818, 3.7121, 3.8061], device='cuda:0'), covar=tensor([0.0539, 0.0518, 0.0271, 0.0274, 0.0712, 0.0408, 0.1003, 0.0556], device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0316, 0.0296, 0.0275, 0.0318, 0.0314, 0.0201, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:48:28,316 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6450, 2.6240, 1.7225, 2.6979, 2.1155, 2.8138, 2.0921, 2.3509], device='cuda:0'), covar=tensor([0.0253, 0.0363, 0.1378, 0.0216, 0.0754, 0.0496, 0.1151, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0165, 0.0187, 0.0130, 0.0168, 0.0208, 0.0196, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 08:48:35,783 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102576.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:49:12,613 INFO [train.py:904] (0/8) Epoch 11, batch 1100, loss[loss=0.179, simple_loss=0.2533, pruned_loss=0.0523, over 16714.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2615, pruned_loss=0.05053, over 3288001.49 frames. ], batch size: 124, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:49:43,775 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=102624.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 08:50:06,741 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-04-29 08:50:07,618 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=102643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:50:09,067 INFO [optim.py:368] (0/8) 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 08:50:20,440 INFO [train.py:904] (0/8) Epoch 11, batch 1150, loss[loss=0.2002, simple_loss=0.2705, pruned_loss=0.06497, over 16687.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2612, pruned_loss=0.05023, over 3293881.93 frames. ], batch size: 134, lr: 6.34e-03, grad_scale: 4.0 2023-04-29 08:50:42,340 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7325, 2.9147, 2.5541, 2.7251, 3.1086, 2.9155, 3.6689, 3.3466], device='cuda:0'), covar=tensor([0.0085, 0.0229, 0.0311, 0.0296, 0.0184, 0.0249, 0.0154, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0207, 0.0203, 0.0203, 0.0206, 0.0204, 0.0210, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:51:14,264 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 11, batch 1200, loss[loss=0.1453, simple_loss=0.2301, pruned_loss=0.03028, over 15974.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.26, pruned_loss=0.04954, over 3285390.33 frames. ], batch size: 35, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:52:12,533 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6299, 6.0208, 5.7513, 5.8352, 5.3732, 5.2889, 5.4921, 6.1054], device='cuda:0'), covar=tensor([0.1294, 0.0853, 0.1081, 0.0771, 0.0850, 0.0645, 0.1079, 0.0953], device='cuda:0'), in_proj_covar=tensor([0.0554, 0.0691, 0.0575, 0.0484, 0.0434, 0.0449, 0.0581, 0.0529], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:52:27,634 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 11, batch 1250, loss[loss=0.2043, simple_loss=0.2737, pruned_loss=0.06747, over 16662.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2601, pruned_loss=0.04966, over 3289727.79 frames. ], batch size: 134, lr: 6.34e-03, grad_scale: 8.0 2023-04-29 08:52:47,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3318, 3.3990, 1.8541, 3.5275, 2.5627, 3.5453, 2.0196, 2.7020], device='cuda:0'), covar=tensor([0.0228, 0.0386, 0.1552, 0.0261, 0.0767, 0.0666, 0.1429, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0167, 0.0188, 0.0132, 0.0170, 0.0209, 0.0197, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 08:53:02,364 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3279, 3.8701, 3.9713, 1.9858, 3.0211, 2.5207, 3.7667, 3.8476], device='cuda:0'), covar=tensor([0.0262, 0.0685, 0.0496, 0.1838, 0.0816, 0.0966, 0.0673, 0.1085], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0141, 0.0156, 0.0142, 0.0135, 0.0124, 0.0136, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 08:53:49,409 INFO [train.py:904] (0/8) Epoch 11, batch 1300, loss[loss=0.1851, simple_loss=0.2615, pruned_loss=0.05432, over 16819.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2598, pruned_loss=0.04924, over 3291018.04 frames. ], batch size: 102, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:54:06,410 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 08:54:07,428 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3156, 5.1952, 5.0801, 4.6749, 4.6405, 5.1284, 5.1977, 4.7973], device='cuda:0'), covar=tensor([0.0514, 0.0384, 0.0266, 0.0282, 0.1081, 0.0383, 0.0248, 0.0644], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0318, 0.0298, 0.0277, 0.0321, 0.0316, 0.0204, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 08:54:46,489 INFO [optim.py:368] (0/8) 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 08:54:57,216 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:54:58,765 INFO [train.py:904] (0/8) Epoch 11, batch 1350, loss[loss=0.1536, simple_loss=0.2416, pruned_loss=0.03278, over 16798.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2599, pruned_loss=0.04874, over 3305896.89 frames. ], batch size: 42, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:55:36,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6142, 4.4249, 4.6307, 4.8465, 5.0155, 4.4716, 4.9041, 4.9507], device='cuda:0'), covar=tensor([0.1580, 0.1035, 0.1482, 0.0634, 0.0502, 0.0953, 0.0979, 0.0553], device='cuda:0'), in_proj_covar=tensor([0.0545, 0.0678, 0.0838, 0.0701, 0.0527, 0.0528, 0.0541, 0.0616], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:55:45,129 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=102887.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:56:05,629 INFO [train.py:904] (0/8) Epoch 11, batch 1400, loss[loss=0.146, simple_loss=0.23, pruned_loss=0.03097, over 16967.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2592, pruned_loss=0.04831, over 3311716.06 frames. ], batch size: 41, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:56:20,002 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102912.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:56:24,059 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 08:57:02,912 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 08:57:05,104 INFO [optim.py:368] (0/8) 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 08:57:10,121 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=102948.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 08:57:15,719 INFO [train.py:904] (0/8) Epoch 11, batch 1450, loss[loss=0.1804, simple_loss=0.2747, pruned_loss=0.04307, over 17041.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2589, pruned_loss=0.04798, over 3320087.53 frames. ], batch size: 55, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:02,098 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1627, 5.1587, 4.8715, 4.1617, 4.9623, 1.7356, 4.7202, 4.7930], device='cuda:0'), covar=tensor([0.0070, 0.0060, 0.0155, 0.0399, 0.0074, 0.2594, 0.0112, 0.0196], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0119, 0.0169, 0.0157, 0.0137, 0.0182, 0.0155, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 08:58:16,947 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 08:58:25,008 INFO [train.py:904] (0/8) Epoch 11, batch 1500, loss[loss=0.2071, simple_loss=0.2814, pruned_loss=0.06641, over 16925.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2599, pruned_loss=0.04863, over 3327900.27 frames. ], batch size: 109, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 08:58:52,460 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6409, 3.8048, 4.0365, 2.0018, 4.1699, 4.1595, 3.2994, 3.0225], device='cuda:0'), covar=tensor([0.0654, 0.0124, 0.0116, 0.1019, 0.0056, 0.0100, 0.0311, 0.0366], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0098, 0.0089, 0.0138, 0.0070, 0.0103, 0.0122, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 08:59:24,571 INFO [optim.py:368] (0/8) 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 08:59:35,065 INFO [train.py:904] (0/8) Epoch 11, batch 1550, loss[loss=0.1528, simple_loss=0.2403, pruned_loss=0.03266, over 17235.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2621, pruned_loss=0.05061, over 3323673.35 frames. ], batch size: 44, lr: 6.33e-03, grad_scale: 8.0 2023-04-29 09:00:07,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1519, 3.7739, 3.4112, 2.0244, 3.0195, 2.6375, 3.5610, 3.7702], device='cuda:0'), covar=tensor([0.0334, 0.0678, 0.0660, 0.1765, 0.0843, 0.0873, 0.0738, 0.0977], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0143, 0.0157, 0.0142, 0.0135, 0.0124, 0.0136, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 09:00:27,170 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6260, 5.9805, 5.7852, 5.8237, 5.3692, 5.3696, 5.4670, 6.1522], device='cuda:0'), covar=tensor([0.1228, 0.0896, 0.0917, 0.0642, 0.0782, 0.0548, 0.0872, 0.0790], device='cuda:0'), in_proj_covar=tensor([0.0547, 0.0685, 0.0565, 0.0478, 0.0429, 0.0441, 0.0573, 0.0523], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:00:39,951 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103098.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:00:43,366 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-29 09:00:45,001 INFO [train.py:904] (0/8) Epoch 11, batch 1600, loss[loss=0.1726, simple_loss=0.263, pruned_loss=0.0411, over 17033.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2642, pruned_loss=0.05103, over 3325038.88 frames. ], batch size: 50, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:01:00,278 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 09:01:26,012 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:01:42,626 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 2.417e+02 3.013e+02 3.629e+02 9.117e+02, threshold=6.027e+02, percent-clipped=4.0 2023-04-29 09:01:53,531 INFO [train.py:904] (0/8) Epoch 11, batch 1650, loss[loss=0.1804, simple_loss=0.2738, pruned_loss=0.04348, over 17153.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2653, pruned_loss=0.05182, over 3319829.66 frames. ], batch size: 46, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:02:03,325 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103159.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:02:49,447 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103193.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:02:52,374 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0096, 2.3707, 2.3757, 4.8593, 2.2271, 2.9372, 2.4034, 2.6326], device='cuda:0'), covar=tensor([0.0826, 0.3217, 0.2290, 0.0292, 0.3774, 0.2136, 0.3022, 0.3206], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0392, 0.0331, 0.0325, 0.0412, 0.0447, 0.0355, 0.0463], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:03:02,617 INFO [train.py:904] (0/8) Epoch 11, batch 1700, loss[loss=0.184, simple_loss=0.2758, pruned_loss=0.04607, over 17036.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2669, pruned_loss=0.05177, over 3329302.96 frames. ], batch size: 55, lr: 6.32e-03, grad_scale: 8.0 2023-04-29 09:03:10,661 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103207.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:03:59,060 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-29 09:04:01,657 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103243.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:04:02,561 INFO [optim.py:368] (0/8) 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 09:04:13,838 INFO [train.py:904] (0/8) Epoch 11, batch 1750, loss[loss=0.2015, simple_loss=0.2689, pruned_loss=0.06701, over 16463.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2683, pruned_loss=0.05266, over 3317630.88 frames. ], batch size: 146, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:04:33,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7721, 3.8696, 2.1340, 4.2277, 2.7437, 4.1962, 2.3053, 2.9794], device='cuda:0'), covar=tensor([0.0252, 0.0312, 0.1498, 0.0256, 0.0731, 0.0513, 0.1340, 0.0666], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0167, 0.0187, 0.0133, 0.0169, 0.0210, 0.0195, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 09:04:50,415 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-29 09:05:22,491 INFO [train.py:904] (0/8) Epoch 11, batch 1800, loss[loss=0.1946, simple_loss=0.2875, pruned_loss=0.05083, over 16647.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2694, pruned_loss=0.05246, over 3318913.79 frames. ], batch size: 57, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:21,373 INFO [optim.py:368] (0/8) 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 09:06:31,969 INFO [train.py:904] (0/8) Epoch 11, batch 1850, loss[loss=0.2008, simple_loss=0.2815, pruned_loss=0.06006, over 16462.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2706, pruned_loss=0.05267, over 3320989.43 frames. ], batch size: 75, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:06:40,880 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 09:06:50,802 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103366.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:07:15,130 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8465, 5.2459, 5.4205, 5.2166, 5.1674, 5.8550, 5.3848, 5.1157], device='cuda:0'), covar=tensor([0.0977, 0.1625, 0.1815, 0.1825, 0.2747, 0.0935, 0.1254, 0.2266], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0500, 0.0540, 0.0434, 0.0576, 0.0565, 0.0425, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 09:07:39,141 INFO [train.py:904] (0/8) Epoch 11, batch 1900, loss[loss=0.1757, simple_loss=0.2547, pruned_loss=0.04835, over 16722.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2698, pruned_loss=0.05226, over 3306608.28 frames. ], batch size: 134, lr: 6.32e-03, grad_scale: 4.0 2023-04-29 09:08:16,097 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103427.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 09:08:40,925 INFO [optim.py:368] (0/8) 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 09:08:51,898 INFO [train.py:904] (0/8) Epoch 11, batch 1950, loss[loss=0.1902, simple_loss=0.2783, pruned_loss=0.05109, over 17071.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2691, pruned_loss=0.05174, over 3311731.46 frames. ], batch size: 53, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:08:55,102 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:08:55,396 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7663, 2.3206, 2.3900, 4.6175, 2.1654, 2.8506, 2.3683, 2.4735], device='cuda:0'), covar=tensor([0.0898, 0.3280, 0.2227, 0.0330, 0.3849, 0.2185, 0.2854, 0.3339], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0391, 0.0330, 0.0323, 0.0411, 0.0446, 0.0354, 0.0461], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:09:01,901 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0707, 3.9299, 4.1353, 4.2993, 4.3929, 3.9551, 4.2019, 4.3522], device='cuda:0'), covar=tensor([0.1267, 0.0899, 0.1160, 0.0533, 0.0486, 0.1365, 0.1294, 0.0572], device='cuda:0'), in_proj_covar=tensor([0.0545, 0.0678, 0.0837, 0.0697, 0.0526, 0.0526, 0.0539, 0.0615], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:09:41,335 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103488.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:00,150 INFO [train.py:904] (0/8) Epoch 11, batch 2000, loss[loss=0.2015, simple_loss=0.2841, pruned_loss=0.05945, over 11875.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2686, pruned_loss=0.05134, over 3303691.84 frames. ], batch size: 246, lr: 6.31e-03, grad_scale: 8.0 2023-04-29 09:10:07,042 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103507.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:10:49,541 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8455, 4.1415, 3.9294, 4.0310, 3.5863, 3.7653, 3.8257, 4.0866], device='cuda:0'), covar=tensor([0.1213, 0.0969, 0.0915, 0.0620, 0.0864, 0.1474, 0.0892, 0.1155], device='cuda:0'), in_proj_covar=tensor([0.0549, 0.0691, 0.0572, 0.0482, 0.0433, 0.0446, 0.0577, 0.0528], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:10:58,936 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103543.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:11:00,879 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.679e+02 2.374e+02 2.796e+02 3.607e+02 7.066e+02, threshold=5.592e+02, percent-clipped=4.0 2023-04-29 09:11:11,330 INFO [train.py:904] (0/8) Epoch 11, batch 2050, loss[loss=0.2022, simple_loss=0.2923, pruned_loss=0.05611, over 16742.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.269, pruned_loss=0.05201, over 3298727.70 frames. ], batch size: 62, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:11:16,438 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103555.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:11:51,272 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4561, 4.2959, 4.5928, 2.5420, 4.9169, 4.8751, 3.5089, 4.0591], device='cuda:0'), covar=tensor([0.0513, 0.0150, 0.0189, 0.0931, 0.0041, 0.0085, 0.0340, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0099, 0.0090, 0.0138, 0.0069, 0.0104, 0.0121, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 09:12:05,788 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103591.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:12:15,829 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9103, 4.2130, 3.9823, 4.0489, 3.6644, 3.7761, 3.8776, 4.1786], device='cuda:0'), covar=tensor([0.1251, 0.0955, 0.0996, 0.0745, 0.0811, 0.1378, 0.0890, 0.1011], device='cuda:0'), in_proj_covar=tensor([0.0547, 0.0689, 0.0570, 0.0481, 0.0431, 0.0444, 0.0575, 0.0524], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:12:21,601 INFO [train.py:904] (0/8) Epoch 11, batch 2100, loss[loss=0.1782, simple_loss=0.2614, pruned_loss=0.04745, over 16547.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2703, pruned_loss=0.05276, over 3288551.97 frames. ], batch size: 68, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:12:22,201 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.30 vs. limit=5.0 2023-04-29 09:13:22,847 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.732e+02 2.751e+02 3.128e+02 3.657e+02 9.004e+02, threshold=6.256e+02, percent-clipped=1.0 2023-04-29 09:13:31,355 INFO [train.py:904] (0/8) Epoch 11, batch 2150, loss[loss=0.1921, simple_loss=0.2888, pruned_loss=0.04768, over 17154.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2705, pruned_loss=0.05223, over 3303626.47 frames. ], batch size: 49, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:14:09,515 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 09:14:42,136 INFO [train.py:904] (0/8) Epoch 11, batch 2200, loss[loss=0.1713, simple_loss=0.2606, pruned_loss=0.04099, over 17207.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2711, pruned_loss=0.05256, over 3310659.71 frames. ], batch size: 46, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:10,364 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=103722.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:15:15,294 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 09:15:35,193 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 09:15:44,603 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2966, 3.3297, 3.5043, 1.9430, 3.6587, 3.6386, 2.8293, 2.8272], device='cuda:0'), covar=tensor([0.0745, 0.0177, 0.0178, 0.0968, 0.0075, 0.0158, 0.0491, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0100, 0.0090, 0.0139, 0.0070, 0.0105, 0.0122, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 09:15:51,102 INFO [train.py:904] (0/8) Epoch 11, batch 2250, loss[loss=0.2058, simple_loss=0.2754, pruned_loss=0.06807, over 16473.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2718, pruned_loss=0.05334, over 3312303.07 frames. ], batch size: 146, lr: 6.31e-03, grad_scale: 4.0 2023-04-29 09:15:54,431 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103754.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:12,353 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:44,208 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=103788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:16:52,332 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2257, 5.1873, 5.0080, 4.2572, 5.0507, 1.6284, 4.8182, 4.8932], device='cuda:0'), covar=tensor([0.0081, 0.0068, 0.0153, 0.0402, 0.0081, 0.2613, 0.0110, 0.0194], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0119, 0.0168, 0.0157, 0.0137, 0.0179, 0.0154, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:16:54,353 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 09:17:01,701 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 11, batch 2300, loss[loss=0.1599, simple_loss=0.2449, pruned_loss=0.03749, over 17243.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2725, pruned_loss=0.05371, over 3320264.35 frames. ], batch size: 45, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:17:02,662 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:35,741 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=103826.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:17:46,893 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-29 09:17:48,571 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=103836.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:18:02,533 INFO [optim.py:368] (0/8) 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 09:18:11,660 INFO [train.py:904] (0/8) Epoch 11, batch 2350, loss[loss=0.2082, simple_loss=0.2818, pruned_loss=0.06732, over 16814.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2734, pruned_loss=0.05433, over 3328700.71 frames. ], batch size: 116, lr: 6.30e-03, grad_scale: 4.0 2023-04-29 09:18:45,248 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103876.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:19:19,658 INFO [train.py:904] (0/8) Epoch 11, batch 2400, loss[loss=0.1808, simple_loss=0.2769, pruned_loss=0.04232, over 17139.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2735, pruned_loss=0.05384, over 3337977.57 frames. ], batch size: 48, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:20:09,863 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 09:20:23,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9236, 1.7567, 2.3366, 2.8949, 2.5900, 3.4376, 2.0568, 3.2785], device='cuda:0'), covar=tensor([0.0148, 0.0377, 0.0236, 0.0218, 0.0233, 0.0112, 0.0351, 0.0090], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0175, 0.0159, 0.0161, 0.0170, 0.0126, 0.0173, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 09:20:28,626 INFO [train.py:904] (0/8) Epoch 11, batch 2450, loss[loss=0.193, simple_loss=0.2646, pruned_loss=0.06067, over 16874.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2737, pruned_loss=0.05308, over 3334957.37 frames. ], batch size: 96, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:21:21,538 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103990.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:31,562 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=103997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:21:37,324 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-104000.pt 2023-04-29 09:21:42,821 INFO [train.py:904] (0/8) Epoch 11, batch 2500, loss[loss=0.1869, simple_loss=0.2628, pruned_loss=0.05544, over 16694.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2734, pruned_loss=0.05249, over 3329929.61 frames. ], batch size: 89, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:11,547 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104022.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 09:22:43,747 INFO [optim.py:368] (0/8) 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 09:22:47,560 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=104049.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:49,908 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104051.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:22:50,610 INFO [train.py:904] (0/8) Epoch 11, batch 2550, loss[loss=0.1448, simple_loss=0.23, pruned_loss=0.02977, over 17001.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2725, pruned_loss=0.05244, over 3331240.92 frames. ], batch size: 41, lr: 6.30e-03, grad_scale: 8.0 2023-04-29 09:22:59,893 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=104058.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:12,645 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4180, 5.3547, 5.1789, 4.5776, 5.2281, 2.2384, 4.9577, 5.2006], device='cuda:0'), covar=tensor([0.0055, 0.0052, 0.0124, 0.0315, 0.0062, 0.1893, 0.0092, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0120, 0.0170, 0.0159, 0.0139, 0.0180, 0.0157, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:23:16,013 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104070.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 09:23:40,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7042, 2.8691, 2.6166, 4.5946, 3.6153, 4.2313, 1.5392, 3.1403], device='cuda:0'), covar=tensor([0.1383, 0.0739, 0.1157, 0.0208, 0.0349, 0.0409, 0.1548, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0156, 0.0178, 0.0144, 0.0196, 0.0209, 0.0177, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 09:23:46,035 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6888, 2.4317, 2.3135, 3.3990, 2.8091, 3.6139, 1.4110, 2.6944], device='cuda:0'), covar=tensor([0.1268, 0.0658, 0.1058, 0.0187, 0.0163, 0.0378, 0.1447, 0.0767], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0157, 0.0178, 0.0144, 0.0197, 0.0209, 0.0177, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 09:23:51,230 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104096.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:23:59,585 INFO [train.py:904] (0/8) Epoch 11, batch 2600, loss[loss=0.192, simple_loss=0.2704, pruned_loss=0.05681, over 16771.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2732, pruned_loss=0.05221, over 3324301.25 frames. ], batch size: 83, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:24:11,977 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104121.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:25:01,157 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 11, batch 2650, loss[loss=0.2159, simple_loss=0.3022, pruned_loss=0.06477, over 16625.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2737, pruned_loss=0.05206, over 3328217.22 frames. ], batch size: 62, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:25:27,724 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0258, 3.9764, 3.9486, 3.3885, 3.9440, 1.8575, 3.7612, 3.5138], device='cuda:0'), covar=tensor([0.0106, 0.0091, 0.0146, 0.0237, 0.0088, 0.2291, 0.0123, 0.0220], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0119, 0.0169, 0.0158, 0.0137, 0.0179, 0.0155, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:26:18,848 INFO [train.py:904] (0/8) Epoch 11, batch 2700, loss[loss=0.1851, simple_loss=0.2796, pruned_loss=0.04525, over 16774.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2736, pruned_loss=0.05179, over 3330721.83 frames. ], batch size: 62, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:26:44,064 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 09:27:00,698 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104232.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:27:19,008 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 11, batch 2750, loss[loss=0.1865, simple_loss=0.2686, pruned_loss=0.05219, over 16422.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2737, pruned_loss=0.05163, over 3334356.02 frames. ], batch size: 146, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:27:52,095 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-04-29 09:28:09,755 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 09:28:36,580 INFO [train.py:904] (0/8) Epoch 11, batch 2800, loss[loss=0.1858, simple_loss=0.2822, pruned_loss=0.04469, over 17092.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2737, pruned_loss=0.05176, over 3332009.24 frames. ], batch size: 53, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:37,341 INFO [optim.py:368] (0/8) 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 09:29:37,605 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104346.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:29:44,362 INFO [train.py:904] (0/8) Epoch 11, batch 2850, loss[loss=0.1727, simple_loss=0.2668, pruned_loss=0.03931, over 17148.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2728, pruned_loss=0.05144, over 3317492.89 frames. ], batch size: 48, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:29:45,782 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:10,900 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8680, 4.6278, 4.8742, 5.1280, 5.2737, 4.6390, 5.2189, 5.2678], device='cuda:0'), covar=tensor([0.1540, 0.1090, 0.1722, 0.0651, 0.0512, 0.0926, 0.0535, 0.0507], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0692, 0.0855, 0.0707, 0.0533, 0.0541, 0.0547, 0.0628], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:30:44,160 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:30:52,874 INFO [train.py:904] (0/8) Epoch 11, batch 2900, loss[loss=0.1806, simple_loss=0.2703, pruned_loss=0.04543, over 17129.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2714, pruned_loss=0.05207, over 3321031.68 frames. ], batch size: 48, lr: 6.29e-03, grad_scale: 8.0 2023-04-29 09:30:57,328 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=104405.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:31:19,972 INFO [zipformer.py:625] (0/8) 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] (0/8) 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] (0/8) 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 09:32:04,107 INFO [train.py:904] (0/8) Epoch 11, batch 2950, loss[loss=0.178, simple_loss=0.2794, pruned_loss=0.03834, over 17131.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2708, pruned_loss=0.05275, over 3321578.67 frames. ], batch size: 49, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:32:12,422 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 09:32:27,836 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104469.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:33:12,390 INFO [train.py:904] (0/8) Epoch 11, batch 3000, loss[loss=0.194, simple_loss=0.2745, pruned_loss=0.05677, over 16660.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2713, pruned_loss=0.05337, over 3325496.20 frames. ], batch size: 89, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:33:12,391 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 09:33:22,065 INFO [train.py:938] (0/8) Epoch 11, validation: loss=0.1413, simple_loss=0.2475, pruned_loss=0.01754, over 944034.00 frames. 2023-04-29 09:33:22,066 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 09:34:04,213 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104532.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:34:20,860 INFO [optim.py:368] (0/8) 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 09:34:30,333 INFO [train.py:904] (0/8) Epoch 11, batch 3050, loss[loss=0.1666, simple_loss=0.2543, pruned_loss=0.0395, over 17212.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2709, pruned_loss=0.05325, over 3323499.73 frames. ], batch size: 46, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:34:39,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0760, 1.8385, 2.4560, 2.8690, 2.8106, 3.0407, 1.9410, 3.1437], device='cuda:0'), covar=tensor([0.0126, 0.0361, 0.0253, 0.0209, 0.0205, 0.0151, 0.0360, 0.0095], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0160, 0.0162, 0.0172, 0.0127, 0.0173, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 09:35:07,353 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104580.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:35:18,351 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 09:35:23,177 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3205, 2.0544, 2.2867, 3.9495, 2.0434, 2.5230, 2.1948, 2.2923], device='cuda:0'), covar=tensor([0.0948, 0.3015, 0.2041, 0.0439, 0.3388, 0.1960, 0.2763, 0.2658], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0394, 0.0331, 0.0326, 0.0413, 0.0452, 0.0356, 0.0465], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:35:37,112 INFO [train.py:904] (0/8) Epoch 11, batch 3100, loss[loss=0.2028, simple_loss=0.2893, pruned_loss=0.05817, over 16692.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2706, pruned_loss=0.0535, over 3327559.66 frames. ], batch size: 62, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:29,825 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-04-29 09:36:39,249 INFO [optim.py:368] (0/8) 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 09:36:39,641 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:36:47,525 INFO [train.py:904] (0/8) Epoch 11, batch 3150, loss[loss=0.1797, simple_loss=0.2579, pruned_loss=0.05076, over 16731.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2697, pruned_loss=0.05297, over 3336483.03 frames. ], batch size: 83, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:36:49,191 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:46,270 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:37:57,456 INFO [train.py:904] (0/8) Epoch 11, batch 3200, loss[loss=0.1721, simple_loss=0.2494, pruned_loss=0.04735, over 15958.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2685, pruned_loss=0.05223, over 3327473.93 frames. ], batch size: 35, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:38:01,348 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=104705.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:38:57,888 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.739e+02 2.472e+02 2.965e+02 3.891e+02 6.936e+02, threshold=5.930e+02, percent-clipped=3.0 2023-04-29 09:39:06,607 INFO [train.py:904] (0/8) Epoch 11, batch 3250, loss[loss=0.1875, simple_loss=0.2735, pruned_loss=0.05071, over 16471.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.269, pruned_loss=0.052, over 3326769.98 frames. ], batch size: 68, lr: 6.28e-03, grad_scale: 8.0 2023-04-29 09:39:08,057 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=104753.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:39:54,772 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1784, 2.0296, 2.5531, 2.9248, 2.8706, 3.5394, 2.2630, 3.4342], device='cuda:0'), covar=tensor([0.0138, 0.0318, 0.0209, 0.0214, 0.0185, 0.0113, 0.0304, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0172, 0.0158, 0.0161, 0.0170, 0.0125, 0.0170, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 09:40:15,738 INFO [train.py:904] (0/8) Epoch 11, batch 3300, loss[loss=0.1592, simple_loss=0.2474, pruned_loss=0.03544, over 17238.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2702, pruned_loss=0.05232, over 3323336.80 frames. ], batch size: 45, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:40:17,662 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-29 09:40:40,135 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 09:40:43,003 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1420, 5.1527, 5.6928, 5.6735, 5.6532, 5.3111, 5.2862, 4.9888], device='cuda:0'), covar=tensor([0.0304, 0.0477, 0.0329, 0.0421, 0.0435, 0.0323, 0.0757, 0.0363], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0355, 0.0354, 0.0333, 0.0402, 0.0370, 0.0477, 0.0300], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 09:41:06,482 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-29 09:41:16,325 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.335e+02 2.858e+02 3.500e+02 6.085e+02, threshold=5.716e+02, percent-clipped=1.0 2023-04-29 09:41:24,650 INFO [train.py:904] (0/8) Epoch 11, batch 3350, loss[loss=0.1634, simple_loss=0.2533, pruned_loss=0.03675, over 17169.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2704, pruned_loss=0.05177, over 3326587.42 frames. ], batch size: 46, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:42:33,972 INFO [train.py:904] (0/8) Epoch 11, batch 3400, loss[loss=0.2071, simple_loss=0.2774, pruned_loss=0.06837, over 16898.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2702, pruned_loss=0.05171, over 3327926.21 frames. ], batch size: 109, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:43:33,852 INFO [optim.py:368] (0/8) 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 09:43:41,675 INFO [train.py:904] (0/8) Epoch 11, batch 3450, loss[loss=0.1496, simple_loss=0.2378, pruned_loss=0.03074, over 16764.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2686, pruned_loss=0.05154, over 3330563.59 frames. ], batch size: 39, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:44:16,236 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 09:44:50,201 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4060, 3.3230, 3.6726, 2.4845, 3.2858, 3.7188, 3.3991, 2.1572], device='cuda:0'), covar=tensor([0.0383, 0.0119, 0.0045, 0.0304, 0.0077, 0.0075, 0.0064, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0071, 0.0071, 0.0125, 0.0078, 0.0090, 0.0077, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 09:44:52,766 INFO [train.py:904] (0/8) Epoch 11, batch 3500, loss[loss=0.1389, simple_loss=0.2217, pruned_loss=0.02808, over 16860.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2678, pruned_loss=0.05138, over 3321708.36 frames. ], batch size: 39, lr: 6.27e-03, grad_scale: 8.0 2023-04-29 09:45:55,146 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.324e+02 2.297e+02 2.659e+02 3.238e+02 6.959e+02, threshold=5.317e+02, percent-clipped=2.0 2023-04-29 09:46:03,259 INFO [train.py:904] (0/8) Epoch 11, batch 3550, loss[loss=0.1569, simple_loss=0.2368, pruned_loss=0.03851, over 16969.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2664, pruned_loss=0.05076, over 3319167.84 frames. ], batch size: 41, lr: 6.27e-03, grad_scale: 4.0 2023-04-29 09:46:34,521 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0497, 1.9591, 2.4940, 2.9784, 2.7782, 3.3578, 2.1404, 3.3613], device='cuda:0'), covar=tensor([0.0154, 0.0366, 0.0225, 0.0197, 0.0225, 0.0123, 0.0350, 0.0098], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0173, 0.0158, 0.0161, 0.0170, 0.0125, 0.0169, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 09:46:53,972 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6611, 6.0603, 5.7596, 5.8349, 5.4102, 5.4259, 5.4859, 6.1876], device='cuda:0'), covar=tensor([0.1346, 0.0846, 0.0943, 0.0672, 0.0868, 0.0572, 0.0978, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0566, 0.0705, 0.0583, 0.0492, 0.0443, 0.0454, 0.0588, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:47:02,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0876, 3.9654, 4.1602, 4.3227, 4.3785, 3.9662, 4.1042, 4.3591], device='cuda:0'), covar=tensor([0.1250, 0.0770, 0.1090, 0.0538, 0.0529, 0.1320, 0.1661, 0.0545], device='cuda:0'), in_proj_covar=tensor([0.0560, 0.0698, 0.0862, 0.0716, 0.0539, 0.0550, 0.0552, 0.0634], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:47:12,593 INFO [train.py:904] (0/8) Epoch 11, batch 3600, loss[loss=0.1816, simple_loss=0.2581, pruned_loss=0.05257, over 16647.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2661, pruned_loss=0.05064, over 3315629.28 frames. ], batch size: 134, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:48:17,998 INFO [optim.py:368] (0/8) 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 09:48:24,362 INFO [train.py:904] (0/8) Epoch 11, batch 3650, loss[loss=0.1869, simple_loss=0.2625, pruned_loss=0.0556, over 11476.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2649, pruned_loss=0.05045, over 3296797.41 frames. ], batch size: 247, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:49:37,402 INFO [train.py:904] (0/8) Epoch 11, batch 3700, loss[loss=0.2196, simple_loss=0.2899, pruned_loss=0.07462, over 16616.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.263, pruned_loss=0.05197, over 3291546.87 frames. ], batch size: 89, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:50:41,016 INFO [optim.py:368] (0/8) 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 09:50:48,708 INFO [train.py:904] (0/8) Epoch 11, batch 3750, loss[loss=0.2033, simple_loss=0.2735, pruned_loss=0.06656, over 16767.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2647, pruned_loss=0.05418, over 3284810.72 frames. ], batch size: 83, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:50:56,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9054, 5.2189, 4.9679, 4.9587, 4.6832, 4.6081, 4.6707, 5.2677], device='cuda:0'), covar=tensor([0.1081, 0.0753, 0.0971, 0.0654, 0.0823, 0.0946, 0.0933, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0558, 0.0691, 0.0571, 0.0484, 0.0436, 0.0446, 0.0578, 0.0528], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:51:57,083 INFO [train.py:904] (0/8) Epoch 11, batch 3800, loss[loss=0.2033, simple_loss=0.2725, pruned_loss=0.06703, over 16766.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.266, pruned_loss=0.05559, over 3287827.87 frames. ], batch size: 124, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:52:46,279 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0806, 5.0749, 4.9280, 4.6424, 4.5446, 5.0510, 4.8466, 4.6975], device='cuda:0'), covar=tensor([0.0501, 0.0335, 0.0223, 0.0244, 0.0905, 0.0277, 0.0340, 0.0508], device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0329, 0.0307, 0.0285, 0.0330, 0.0328, 0.0209, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 09:53:00,960 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7853, 6.0699, 5.8187, 5.9046, 5.4797, 5.1821, 5.5213, 6.2514], device='cuda:0'), covar=tensor([0.1069, 0.0752, 0.0829, 0.0640, 0.0719, 0.0596, 0.0817, 0.0657], device='cuda:0'), in_proj_covar=tensor([0.0563, 0.0696, 0.0575, 0.0488, 0.0439, 0.0451, 0.0581, 0.0531], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:53:02,335 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 11, batch 3850, loss[loss=0.186, simple_loss=0.2602, pruned_loss=0.0559, over 16744.00 frames. ], tot_loss[loss=0.19, simple_loss=0.2666, pruned_loss=0.05673, over 3284307.75 frames. ], batch size: 124, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:53:16,097 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 09:53:59,983 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105388.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:54:19,877 INFO [train.py:904] (0/8) Epoch 11, batch 3900, loss[loss=0.1773, simple_loss=0.2536, pruned_loss=0.05047, over 16736.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2655, pruned_loss=0.05677, over 3281887.89 frames. ], batch size: 102, lr: 6.26e-03, grad_scale: 8.0 2023-04-29 09:55:02,355 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4121, 4.3918, 4.5544, 4.4504, 4.4797, 4.9958, 4.5328, 4.2919], device='cuda:0'), covar=tensor([0.1415, 0.1904, 0.1801, 0.1760, 0.2502, 0.1002, 0.1385, 0.2301], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0496, 0.0533, 0.0426, 0.0563, 0.0557, 0.0423, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 09:55:25,286 INFO [optim.py:368] (0/8) 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 09:55:28,921 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105449.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 09:55:31,854 INFO [train.py:904] (0/8) Epoch 11, batch 3950, loss[loss=0.1928, simple_loss=0.2589, pruned_loss=0.06333, over 16866.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2646, pruned_loss=0.05689, over 3279443.58 frames. ], batch size: 116, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:55:33,547 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5803, 2.3729, 1.8154, 2.1373, 2.8152, 2.6263, 2.9181, 2.8788], device='cuda:0'), covar=tensor([0.0139, 0.0236, 0.0347, 0.0311, 0.0128, 0.0192, 0.0140, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0204, 0.0199, 0.0199, 0.0203, 0.0201, 0.0212, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:55:49,317 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7623, 4.0542, 3.1040, 2.4195, 2.8372, 2.4983, 4.2598, 3.6665], device='cuda:0'), covar=tensor([0.2429, 0.0534, 0.1415, 0.2133, 0.2272, 0.1644, 0.0432, 0.0892], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0258, 0.0286, 0.0279, 0.0291, 0.0224, 0.0270, 0.0303], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 09:55:53,780 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5645, 2.3697, 2.3947, 4.4656, 2.2250, 2.7374, 2.3937, 2.5941], device='cuda:0'), covar=tensor([0.0875, 0.2904, 0.2088, 0.0307, 0.3430, 0.2007, 0.2691, 0.2692], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0396, 0.0332, 0.0326, 0.0414, 0.0457, 0.0361, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 09:56:21,186 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105486.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:56:44,048 INFO [train.py:904] (0/8) Epoch 11, batch 4000, loss[loss=0.1824, simple_loss=0.258, pruned_loss=0.0534, over 16701.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2645, pruned_loss=0.05699, over 3280938.77 frames. ], batch size: 57, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:56:44,566 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105502.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:00,462 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7588, 3.7102, 4.2035, 1.8755, 4.4081, 4.4603, 3.1891, 3.2543], device='cuda:0'), covar=tensor([0.0715, 0.0237, 0.0143, 0.1242, 0.0043, 0.0067, 0.0360, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0101, 0.0090, 0.0140, 0.0071, 0.0106, 0.0122, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 09:57:29,888 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9826, 3.5690, 3.5711, 2.2240, 3.2322, 3.5328, 3.3034, 1.9669], device='cuda:0'), covar=tensor([0.0431, 0.0046, 0.0032, 0.0328, 0.0074, 0.0081, 0.0066, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0124, 0.0078, 0.0089, 0.0077, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 09:57:37,961 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2706, 4.0051, 4.0941, 2.5592, 3.5854, 3.9885, 3.6393, 2.2855], device='cuda:0'), covar=tensor([0.0431, 0.0038, 0.0029, 0.0323, 0.0060, 0.0085, 0.0058, 0.0340], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0070, 0.0070, 0.0124, 0.0078, 0.0089, 0.0077, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 09:57:48,231 INFO [optim.py:368] (0/8) 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 09:57:48,858 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105547.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:57:55,347 INFO [train.py:904] (0/8) Epoch 11, batch 4050, loss[loss=0.164, simple_loss=0.2532, pruned_loss=0.03747, over 16868.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2648, pruned_loss=0.05591, over 3274401.53 frames. ], batch size: 102, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:58:11,104 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105563.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 09:58:21,183 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 09:59:08,745 INFO [train.py:904] (0/8) Epoch 11, batch 4100, loss[loss=0.182, simple_loss=0.2622, pruned_loss=0.05088, over 17115.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.266, pruned_loss=0.05521, over 3257967.80 frames. ], batch size: 47, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 09:59:14,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6379, 4.4739, 4.6932, 4.8681, 4.9697, 4.4829, 4.9654, 4.9796], device='cuda:0'), covar=tensor([0.1218, 0.0854, 0.1167, 0.0504, 0.0424, 0.0793, 0.0435, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0537, 0.0672, 0.0819, 0.0692, 0.0522, 0.0529, 0.0529, 0.0612], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:00:18,930 INFO [optim.py:368] (0/8) 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 10:00:26,726 INFO [train.py:904] (0/8) Epoch 11, batch 4150, loss[loss=0.2639, simple_loss=0.326, pruned_loss=0.1009, over 11579.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2735, pruned_loss=0.05814, over 3223791.91 frames. ], batch size: 246, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:00:28,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3519, 3.2576, 3.3357, 3.4484, 3.4818, 3.2493, 3.4820, 3.5545], device='cuda:0'), covar=tensor([0.0920, 0.0785, 0.0945, 0.0559, 0.0504, 0.2157, 0.0752, 0.0628], device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0662, 0.0805, 0.0680, 0.0513, 0.0521, 0.0520, 0.0602], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:01:44,708 INFO [train.py:904] (0/8) Epoch 11, batch 4200, loss[loss=0.2091, simple_loss=0.3011, pruned_loss=0.05859, over 16215.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2806, pruned_loss=0.05974, over 3197385.16 frames. ], batch size: 165, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:02:45,164 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7429, 5.0211, 4.7656, 4.8068, 4.4869, 4.4434, 4.4079, 5.0692], device='cuda:0'), covar=tensor([0.0899, 0.0725, 0.0868, 0.0642, 0.0767, 0.1003, 0.0971, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0544, 0.0672, 0.0560, 0.0471, 0.0428, 0.0439, 0.0562, 0.0517], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:02:49,768 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105744.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 10:02:53,623 INFO [optim.py:368] (0/8) 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 10:02:59,820 INFO [train.py:904] (0/8) Epoch 11, batch 4250, loss[loss=0.188, simple_loss=0.2886, pruned_loss=0.04373, over 16755.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2843, pruned_loss=0.05954, over 3183773.89 frames. ], batch size: 89, lr: 6.25e-03, grad_scale: 8.0 2023-04-29 10:03:59,178 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105792.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:08,468 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=105799.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:04:12,708 INFO [train.py:904] (0/8) Epoch 11, batch 4300, loss[loss=0.1897, simple_loss=0.2884, pruned_loss=0.04553, over 16725.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2855, pruned_loss=0.05869, over 3178388.77 frames. ], batch size: 76, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:04:13,506 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 10:05:06,340 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6540, 1.6522, 2.2611, 2.7222, 2.6120, 3.1880, 1.6809, 2.9086], device='cuda:0'), covar=tensor([0.0158, 0.0401, 0.0224, 0.0225, 0.0195, 0.0087, 0.0426, 0.0103], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0157, 0.0160, 0.0167, 0.0123, 0.0169, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 10:05:11,573 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:17,312 INFO [optim.py:368] (0/8) 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 10:05:25,483 INFO [train.py:904] (0/8) Epoch 11, batch 4350, loss[loss=0.2283, simple_loss=0.3027, pruned_loss=0.07697, over 11955.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2889, pruned_loss=0.05981, over 3184091.61 frames. ], batch size: 246, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:05:27,861 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105853.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:34,708 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=105858.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:05:37,332 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=105860.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:06:38,334 INFO [train.py:904] (0/8) Epoch 11, batch 4400, loss[loss=0.2065, simple_loss=0.2916, pruned_loss=0.06069, over 16234.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2911, pruned_loss=0.06082, over 3189807.64 frames. ], batch size: 165, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:07:21,255 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6337, 4.7488, 4.4233, 4.1558, 4.1216, 4.5947, 4.3588, 4.2310], device='cuda:0'), covar=tensor([0.0494, 0.0191, 0.0247, 0.0240, 0.0744, 0.0254, 0.0394, 0.0547], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0304, 0.0287, 0.0266, 0.0306, 0.0302, 0.0195, 0.0331], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:07:23,540 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7284, 4.8451, 5.0991, 4.8226, 4.9474, 5.5066, 5.1454, 4.7417], device='cuda:0'), covar=tensor([0.0870, 0.1719, 0.1523, 0.1796, 0.2322, 0.0874, 0.0997, 0.2101], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0477, 0.0517, 0.0416, 0.0549, 0.0544, 0.0409, 0.0561], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 10:07:33,253 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 10:07:40,711 INFO [optim.py:368] (0/8) 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 10:07:48,899 INFO [train.py:904] (0/8) Epoch 11, batch 4450, loss[loss=0.2233, simple_loss=0.3129, pruned_loss=0.06681, over 16839.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2943, pruned_loss=0.06191, over 3201998.71 frames. ], batch size: 116, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:08:13,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1699, 4.9223, 4.9064, 3.4647, 3.8614, 4.6314, 4.1577, 3.1248], device='cuda:0'), covar=tensor([0.0325, 0.0013, 0.0020, 0.0243, 0.0070, 0.0066, 0.0048, 0.0273], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0069, 0.0069, 0.0124, 0.0077, 0.0089, 0.0076, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 10:08:48,889 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4886, 2.2278, 2.3791, 4.2985, 2.0965, 2.7458, 2.3784, 2.4198], device='cuda:0'), covar=tensor([0.0868, 0.2922, 0.2008, 0.0340, 0.3540, 0.1975, 0.2554, 0.3060], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0394, 0.0329, 0.0323, 0.0414, 0.0455, 0.0357, 0.0466], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:08:59,108 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-106000.pt 2023-04-29 10:09:04,920 INFO [train.py:904] (0/8) Epoch 11, batch 4500, loss[loss=0.2142, simple_loss=0.3028, pruned_loss=0.06273, over 16733.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2943, pruned_loss=0.06231, over 3216980.60 frames. ], batch size: 62, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:09:05,475 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2995, 2.5522, 2.0632, 2.4382, 2.8794, 2.4775, 3.1292, 3.0840], device='cuda:0'), covar=tensor([0.0057, 0.0245, 0.0350, 0.0262, 0.0142, 0.0248, 0.0106, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0201, 0.0197, 0.0196, 0.0201, 0.0202, 0.0205, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:09:27,887 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7132, 3.7720, 2.0191, 4.3141, 2.8192, 4.2362, 2.3700, 2.9174], device='cuda:0'), covar=tensor([0.0183, 0.0263, 0.1673, 0.0072, 0.0786, 0.0302, 0.1396, 0.0648], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0165, 0.0186, 0.0128, 0.0166, 0.0208, 0.0193, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 10:09:48,465 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-29 10:10:04,871 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:10:09,394 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 1.945e+02 2.292e+02 2.586e+02 4.524e+02, threshold=4.584e+02, percent-clipped=0.0 2023-04-29 10:10:17,384 INFO [train.py:904] (0/8) Epoch 11, batch 4550, loss[loss=0.2107, simple_loss=0.3039, pruned_loss=0.05871, over 16753.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.2946, pruned_loss=0.06246, over 3227912.17 frames. ], batch size: 89, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:10:39,137 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1164, 1.7870, 2.5232, 3.0404, 2.8611, 3.5309, 1.9588, 3.3980], device='cuda:0'), covar=tensor([0.0127, 0.0348, 0.0215, 0.0176, 0.0171, 0.0095, 0.0340, 0.0077], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0156, 0.0160, 0.0167, 0.0122, 0.0169, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 10:10:53,561 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:14,749 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:11:29,162 INFO [train.py:904] (0/8) Epoch 11, batch 4600, loss[loss=0.1751, simple_loss=0.2735, pruned_loss=0.03835, over 16486.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2956, pruned_loss=0.06264, over 3230252.63 frames. ], batch size: 146, lr: 6.24e-03, grad_scale: 8.0 2023-04-29 10:11:51,936 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1498, 4.1591, 2.4431, 5.0634, 3.1767, 4.7999, 2.7639, 3.3592], device='cuda:0'), covar=tensor([0.0173, 0.0249, 0.1459, 0.0060, 0.0695, 0.0247, 0.1314, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0163, 0.0186, 0.0128, 0.0166, 0.0206, 0.0193, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 10:12:01,520 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2807, 2.0815, 2.1859, 3.9858, 1.9869, 2.4412, 2.1930, 2.2350], device='cuda:0'), covar=tensor([0.0968, 0.3084, 0.2211, 0.0411, 0.3875, 0.2169, 0.2782, 0.3253], device='cuda:0'), in_proj_covar=tensor([0.0362, 0.0390, 0.0326, 0.0319, 0.0412, 0.0452, 0.0354, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:12:11,129 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-04-29 10:12:22,232 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106138.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:22,367 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3627, 2.1714, 2.2771, 4.0912, 2.1039, 2.4685, 2.2786, 2.3094], device='cuda:0'), covar=tensor([0.0920, 0.2856, 0.2123, 0.0367, 0.3594, 0.2040, 0.2573, 0.2978], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0389, 0.0325, 0.0318, 0.0411, 0.0450, 0.0353, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:12:28,058 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:29,386 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0274, 1.8270, 2.4495, 3.0163, 2.8969, 3.5218, 1.8514, 3.4211], device='cuda:0'), covar=tensor([0.0126, 0.0355, 0.0234, 0.0153, 0.0166, 0.0090, 0.0365, 0.0085], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0156, 0.0160, 0.0167, 0.0123, 0.0169, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 10:12:35,554 INFO [optim.py:368] (0/8) 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 10:12:37,657 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:41,945 INFO [train.py:904] (0/8) Epoch 11, batch 4650, loss[loss=0.199, simple_loss=0.2727, pruned_loss=0.06267, over 16654.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2942, pruned_loss=0.06229, over 3224682.66 frames. ], batch size: 57, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:12:47,347 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106155.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:48,716 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7215, 1.7469, 1.5060, 1.5709, 1.8640, 1.6347, 1.7328, 1.9412], device='cuda:0'), covar=tensor([0.0112, 0.0214, 0.0297, 0.0243, 0.0134, 0.0192, 0.0129, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0203, 0.0199, 0.0198, 0.0202, 0.0203, 0.0207, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:12:51,766 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:12:57,387 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1543, 1.9646, 1.6105, 1.7575, 2.2322, 1.9288, 2.1286, 2.3631], device='cuda:0'), covar=tensor([0.0119, 0.0295, 0.0386, 0.0357, 0.0163, 0.0268, 0.0137, 0.0171], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0203, 0.0199, 0.0198, 0.0202, 0.0202, 0.0208, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:13:13,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9777, 4.0696, 3.8223, 3.6139, 3.5500, 3.9560, 3.6235, 3.6438], device='cuda:0'), covar=tensor([0.0481, 0.0285, 0.0221, 0.0223, 0.0639, 0.0302, 0.0809, 0.0477], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0292, 0.0276, 0.0256, 0.0297, 0.0292, 0.0190, 0.0319], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:13:29,696 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6776, 2.6318, 1.7182, 2.7780, 2.1726, 2.7748, 2.0011, 2.3512], device='cuda:0'), covar=tensor([0.0283, 0.0455, 0.1472, 0.0193, 0.0801, 0.0478, 0.1369, 0.0645], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0163, 0.0185, 0.0127, 0.0165, 0.0205, 0.0192, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 10:13:37,561 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106190.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:13:42,312 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4998, 3.3649, 2.7247, 2.1113, 2.2753, 2.1119, 3.5281, 3.1178], device='cuda:0'), covar=tensor([0.2730, 0.0820, 0.1726, 0.2310, 0.2223, 0.1961, 0.0553, 0.1175], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0253, 0.0281, 0.0277, 0.0284, 0.0221, 0.0265, 0.0296], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:13:55,739 INFO [train.py:904] (0/8) Epoch 11, batch 4700, loss[loss=0.1917, simple_loss=0.2799, pruned_loss=0.05178, over 17196.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2912, pruned_loss=0.06078, over 3229892.88 frames. ], batch size: 46, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:14:01,837 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:14:25,906 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 10:15:01,680 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.393e+02 2.142e+02 2.451e+02 2.986e+02 5.737e+02, threshold=4.902e+02, percent-clipped=3.0 2023-04-29 10:15:09,056 INFO [train.py:904] (0/8) Epoch 11, batch 4750, loss[loss=0.1855, simple_loss=0.274, pruned_loss=0.04845, over 16225.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2875, pruned_loss=0.0588, over 3235291.99 frames. ], batch size: 165, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:15:32,333 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:15:42,193 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7472, 4.7829, 4.6204, 4.3178, 4.1216, 4.7054, 4.5164, 4.3409], device='cuda:0'), covar=tensor([0.0558, 0.0415, 0.0281, 0.0260, 0.1132, 0.0421, 0.0369, 0.0586], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0299, 0.0280, 0.0261, 0.0302, 0.0297, 0.0193, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:16:18,770 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 10:16:22,066 INFO [train.py:904] (0/8) Epoch 11, batch 4800, loss[loss=0.1876, simple_loss=0.2656, pruned_loss=0.05483, over 16996.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2837, pruned_loss=0.05688, over 3231818.98 frames. ], batch size: 53, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:16:49,764 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-04-29 10:17:02,272 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:12,267 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106336.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:17:27,379 INFO [optim.py:368] (0/8) 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 10:17:35,188 INFO [train.py:904] (0/8) Epoch 11, batch 4850, loss[loss=0.2054, simple_loss=0.2986, pruned_loss=0.05607, over 16816.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2849, pruned_loss=0.0564, over 3220748.27 frames. ], batch size: 116, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:17:51,101 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 10:17:59,237 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 10:18:00,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5156, 3.6225, 1.7905, 3.9967, 2.5782, 3.9550, 2.1482, 2.8612], device='cuda:0'), covar=tensor([0.0215, 0.0330, 0.1828, 0.0103, 0.0882, 0.0385, 0.1630, 0.0700], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0162, 0.0185, 0.0126, 0.0165, 0.0204, 0.0192, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 10:18:38,705 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106397.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:18:45,489 INFO [train.py:904] (0/8) Epoch 11, batch 4900, loss[loss=0.1898, simple_loss=0.2689, pruned_loss=0.05529, over 16648.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2845, pruned_loss=0.05588, over 3200828.13 frames. ], batch size: 57, lr: 6.23e-03, grad_scale: 8.0 2023-04-29 10:19:29,916 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106433.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:19:40,552 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 10:19:50,199 INFO [optim.py:368] (0/8) 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 10:19:51,816 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106448.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:19:56,921 INFO [train.py:904] (0/8) Epoch 11, batch 4950, loss[loss=0.1985, simple_loss=0.2841, pruned_loss=0.05645, over 11904.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2836, pruned_loss=0.05488, over 3200040.05 frames. ], batch size: 246, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:20:01,902 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=106454.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:20:02,969 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106455.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:00,655 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106496.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:08,423 INFO [train.py:904] (0/8) Epoch 11, batch 5000, loss[loss=0.183, simple_loss=0.2776, pruned_loss=0.04419, over 16248.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2853, pruned_loss=0.0552, over 3201940.49 frames. ], batch size: 165, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:21:10,915 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106503.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:21:28,726 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=106515.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:22:14,244 INFO [optim.py:368] (0/8) 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 10:22:21,434 INFO [train.py:904] (0/8) Epoch 11, batch 5050, loss[loss=0.2034, simple_loss=0.2939, pruned_loss=0.05644, over 16138.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2861, pruned_loss=0.05564, over 3189118.71 frames. ], batch size: 165, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:22:52,164 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.03 vs. limit=5.0 2023-04-29 10:23:32,377 INFO [train.py:904] (0/8) Epoch 11, batch 5100, loss[loss=0.1821, simple_loss=0.2714, pruned_loss=0.04641, over 16774.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2842, pruned_loss=0.05464, over 3202544.90 frames. ], batch size: 83, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:23:54,441 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-29 10:24:02,605 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3800, 4.2944, 4.2659, 2.9039, 3.5306, 4.1667, 3.7662, 2.4000], device='cuda:0'), covar=tensor([0.0449, 0.0017, 0.0020, 0.0286, 0.0068, 0.0069, 0.0064, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0069, 0.0070, 0.0127, 0.0079, 0.0091, 0.0078, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 10:24:03,614 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106624.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:24:38,783 INFO [optim.py:368] (0/8) 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 10:24:44,368 INFO [train.py:904] (0/8) Epoch 11, batch 5150, loss[loss=0.202, simple_loss=0.296, pruned_loss=0.05399, over 16838.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2844, pruned_loss=0.05416, over 3193404.97 frames. ], batch size: 96, lr: 6.22e-03, grad_scale: 4.0 2023-04-29 10:25:43,749 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:25:56,063 INFO [train.py:904] (0/8) Epoch 11, batch 5200, loss[loss=0.2087, simple_loss=0.2815, pruned_loss=0.06793, over 16428.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2827, pruned_loss=0.05359, over 3200513.41 frames. ], batch size: 35, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:26:42,271 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106733.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:27:00,560 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 10:27:04,341 INFO [optim.py:368] (0/8) 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 10:27:11,367 INFO [train.py:904] (0/8) Epoch 11, batch 5250, loss[loss=0.2157, simple_loss=0.2885, pruned_loss=0.07147, over 12245.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2804, pruned_loss=0.05333, over 3195876.81 frames. ], batch size: 247, lr: 6.22e-03, grad_scale: 8.0 2023-04-29 10:27:18,132 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7421, 4.9101, 5.1150, 4.8832, 4.9087, 5.4915, 4.9625, 4.6751], device='cuda:0'), covar=tensor([0.0895, 0.1614, 0.1361, 0.1470, 0.2202, 0.0822, 0.1150, 0.1989], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0467, 0.0506, 0.0409, 0.0541, 0.0536, 0.0404, 0.0559], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 10:27:23,555 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-29 10:27:52,546 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:28:22,267 INFO [train.py:904] (0/8) Epoch 11, batch 5300, loss[loss=0.15, simple_loss=0.2444, pruned_loss=0.0278, over 16887.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2777, pruned_loss=0.05237, over 3194590.66 frames. ], batch size: 102, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:28:23,454 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2938, 2.1435, 1.7709, 1.9507, 2.3425, 2.1452, 2.3094, 2.5681], device='cuda:0'), covar=tensor([0.0122, 0.0279, 0.0363, 0.0329, 0.0156, 0.0295, 0.0121, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0200, 0.0196, 0.0194, 0.0198, 0.0200, 0.0201, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:28:32,919 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=106810.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:28:52,631 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6865, 2.6246, 2.3639, 4.2232, 2.9539, 3.9909, 1.4546, 2.9357], device='cuda:0'), covar=tensor([0.1318, 0.0698, 0.1188, 0.0119, 0.0220, 0.0327, 0.1575, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0157, 0.0179, 0.0143, 0.0197, 0.0207, 0.0179, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 10:29:25,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6570, 4.5123, 4.6277, 4.8737, 5.0309, 4.4726, 5.0095, 5.0041], device='cuda:0'), covar=tensor([0.1422, 0.1033, 0.1629, 0.0653, 0.0472, 0.0925, 0.0487, 0.0520], device='cuda:0'), in_proj_covar=tensor([0.0517, 0.0649, 0.0793, 0.0665, 0.0505, 0.0511, 0.0513, 0.0589], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:29:27,202 INFO [optim.py:368] (0/8) 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 10:29:33,916 INFO [train.py:904] (0/8) Epoch 11, batch 5350, loss[loss=0.2009, simple_loss=0.2932, pruned_loss=0.05437, over 15485.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2759, pruned_loss=0.05168, over 3198476.67 frames. ], batch size: 191, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:30:45,868 INFO [train.py:904] (0/8) Epoch 11, batch 5400, loss[loss=0.2211, simple_loss=0.3078, pruned_loss=0.06718, over 16844.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2787, pruned_loss=0.05257, over 3199766.24 frames. ], batch size: 116, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:31:18,226 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106924.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:31:54,546 INFO [optim.py:368] (0/8) 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 10:32:02,037 INFO [train.py:904] (0/8) Epoch 11, batch 5450, loss[loss=0.2014, simple_loss=0.2867, pruned_loss=0.05803, over 16636.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2816, pruned_loss=0.05413, over 3198948.37 frames. ], batch size: 62, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:32:34,435 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=106972.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:03,883 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=106992.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:33:17,307 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9408, 4.7657, 4.8914, 5.1439, 5.3228, 4.6800, 5.3299, 5.3102], device='cuda:0'), covar=tensor([0.1504, 0.1051, 0.1636, 0.0634, 0.0528, 0.0804, 0.0438, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0515, 0.0645, 0.0786, 0.0659, 0.0503, 0.0508, 0.0509, 0.0586], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:33:19,271 INFO [train.py:904] (0/8) Epoch 11, batch 5500, loss[loss=0.2624, simple_loss=0.3309, pruned_loss=0.0969, over 15205.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2895, pruned_loss=0.05972, over 3164052.91 frames. ], batch size: 190, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:34:18,909 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:34:31,622 INFO [optim.py:368] (0/8) 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 10:34:37,911 INFO [train.py:904] (0/8) Epoch 11, batch 5550, loss[loss=0.2163, simple_loss=0.3029, pruned_loss=0.06488, over 16677.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2974, pruned_loss=0.06541, over 3140728.76 frames. ], batch size: 83, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:35:42,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0691, 2.2977, 2.2985, 2.8772, 2.0440, 3.1878, 1.8635, 2.6756], device='cuda:0'), covar=tensor([0.1085, 0.0513, 0.0954, 0.0186, 0.0183, 0.0371, 0.1278, 0.0652], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0158, 0.0180, 0.0143, 0.0198, 0.0207, 0.0180, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 10:35:52,463 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1616, 3.1990, 1.7418, 3.4463, 2.4606, 3.4628, 2.0102, 2.6285], device='cuda:0'), covar=tensor([0.0239, 0.0382, 0.1670, 0.0174, 0.0801, 0.0578, 0.1413, 0.0684], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0165, 0.0190, 0.0127, 0.0169, 0.0207, 0.0196, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 10:35:55,930 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5139, 4.1451, 4.0446, 2.6850, 3.5726, 4.0892, 3.7646, 2.3409], device='cuda:0'), covar=tensor([0.0391, 0.0022, 0.0035, 0.0308, 0.0080, 0.0075, 0.0050, 0.0338], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0069, 0.0070, 0.0127, 0.0077, 0.0091, 0.0077, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 10:35:57,902 INFO [train.py:904] (0/8) Epoch 11, batch 5600, loss[loss=0.3124, simple_loss=0.359, pruned_loss=0.1329, over 10772.00 frames. ], tot_loss[loss=0.2222, simple_loss=0.303, pruned_loss=0.0707, over 3092019.82 frames. ], batch size: 247, lr: 6.21e-03, grad_scale: 8.0 2023-04-29 10:36:12,210 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:02,638 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107140.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:37:17,302 INFO [optim.py:368] (0/8) 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 10:37:21,733 INFO [train.py:904] (0/8) Epoch 11, batch 5650, loss[loss=0.2398, simple_loss=0.3179, pruned_loss=0.08081, over 16860.00 frames. ], tot_loss[loss=0.2293, simple_loss=0.3081, pruned_loss=0.07531, over 3051265.61 frames. ], batch size: 109, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:37:22,219 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107152.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:37:23,933 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 10:37:32,231 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:38:10,178 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4782, 4.3073, 4.5302, 4.7141, 4.8501, 4.3520, 4.8504, 4.8335], device='cuda:0'), covar=tensor([0.1559, 0.1130, 0.1438, 0.0616, 0.0500, 0.0963, 0.0484, 0.0543], device='cuda:0'), in_proj_covar=tensor([0.0507, 0.0638, 0.0773, 0.0648, 0.0496, 0.0502, 0.0508, 0.0578], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:38:42,747 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107201.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:38:43,539 INFO [train.py:904] (0/8) Epoch 11, batch 5700, loss[loss=0.2739, simple_loss=0.3254, pruned_loss=0.1112, over 11171.00 frames. ], tot_loss[loss=0.2331, simple_loss=0.3107, pruned_loss=0.07774, over 3027217.80 frames. ], batch size: 248, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:39:02,198 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 10:39:02,907 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107213.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:39:59,322 INFO [optim.py:368] (0/8) 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 10:40:04,437 INFO [train.py:904] (0/8) Epoch 11, batch 5750, loss[loss=0.2001, simple_loss=0.2926, pruned_loss=0.05378, over 16832.00 frames. ], tot_loss[loss=0.2361, simple_loss=0.3137, pruned_loss=0.07923, over 3003690.90 frames. ], batch size: 102, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:40:34,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6223, 4.5785, 4.4275, 3.7746, 4.4352, 1.6623, 4.1975, 4.1927], device='cuda:0'), covar=tensor([0.0077, 0.0073, 0.0132, 0.0306, 0.0081, 0.2319, 0.0127, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0114, 0.0161, 0.0155, 0.0133, 0.0175, 0.0148, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:41:25,733 INFO [train.py:904] (0/8) Epoch 11, batch 5800, loss[loss=0.2404, simple_loss=0.3317, pruned_loss=0.0746, over 16479.00 frames. ], tot_loss[loss=0.2347, simple_loss=0.3132, pruned_loss=0.07816, over 3002875.07 frames. ], batch size: 146, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:42:27,315 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3316, 2.9102, 2.6282, 2.2470, 2.2654, 2.1917, 2.8569, 2.9041], device='cuda:0'), covar=tensor([0.2306, 0.0750, 0.1400, 0.2015, 0.1865, 0.1804, 0.0494, 0.0985], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0257, 0.0282, 0.0276, 0.0282, 0.0221, 0.0267, 0.0292], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:42:39,031 INFO [optim.py:368] (0/8) 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 10:42:43,696 INFO [train.py:904] (0/8) Epoch 11, batch 5850, loss[loss=0.239, simple_loss=0.3208, pruned_loss=0.07861, over 16932.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3103, pruned_loss=0.0757, over 3013290.20 frames. ], batch size: 109, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:43:26,774 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-29 10:43:51,599 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4452, 4.4421, 4.8487, 4.8366, 4.8620, 4.4827, 4.4881, 4.2530], device='cuda:0'), covar=tensor([0.0270, 0.0420, 0.0378, 0.0401, 0.0423, 0.0325, 0.0873, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0331, 0.0335, 0.0317, 0.0378, 0.0351, 0.0449, 0.0286], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 10:44:05,221 INFO [train.py:904] (0/8) Epoch 11, batch 5900, loss[loss=0.2267, simple_loss=0.3027, pruned_loss=0.07538, over 17042.00 frames. ], tot_loss[loss=0.2309, simple_loss=0.3105, pruned_loss=0.07569, over 3017042.35 frames. ], batch size: 53, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:45:21,997 INFO [optim.py:368] (0/8) 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 10:45:26,044 INFO [train.py:904] (0/8) Epoch 11, batch 5950, loss[loss=0.2284, simple_loss=0.3106, pruned_loss=0.07307, over 16729.00 frames. ], tot_loss[loss=0.2292, simple_loss=0.3107, pruned_loss=0.07386, over 3036812.35 frames. ], batch size: 134, lr: 6.20e-03, grad_scale: 4.0 2023-04-29 10:46:40,159 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107496.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:46:48,948 INFO [train.py:904] (0/8) Epoch 11, batch 6000, loss[loss=0.227, simple_loss=0.2964, pruned_loss=0.07886, over 11465.00 frames. ], tot_loss[loss=0.2283, simple_loss=0.31, pruned_loss=0.07327, over 3039552.98 frames. ], batch size: 246, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:46:48,949 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 10:46:59,887 INFO [train.py:938] (0/8) Epoch 11, validation: loss=0.163, simple_loss=0.2761, pruned_loss=0.02492, over 944034.00 frames. 2023-04-29 10:46:59,888 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 10:47:09,683 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107508.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 10:47:13,599 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:47:34,374 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6176, 4.0154, 4.0323, 2.2690, 3.3851, 2.6132, 4.0049, 4.2190], device='cuda:0'), covar=tensor([0.0203, 0.0552, 0.0498, 0.1731, 0.0645, 0.0869, 0.0558, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0144, 0.0158, 0.0144, 0.0135, 0.0125, 0.0136, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 10:47:47,906 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7835, 3.7025, 3.8495, 3.9926, 4.0592, 3.6793, 3.9795, 4.0560], device='cuda:0'), covar=tensor([0.1264, 0.0921, 0.1121, 0.0526, 0.0501, 0.1621, 0.0627, 0.0592], device='cuda:0'), in_proj_covar=tensor([0.0514, 0.0641, 0.0778, 0.0652, 0.0502, 0.0505, 0.0516, 0.0582], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:48:12,722 INFO [optim.py:368] (0/8) 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 10:48:18,428 INFO [train.py:904] (0/8) Epoch 11, batch 6050, loss[loss=0.2091, simple_loss=0.2949, pruned_loss=0.06165, over 15501.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3077, pruned_loss=0.07175, over 3070457.81 frames. ], batch size: 190, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:48:48,419 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107571.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:49:11,039 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9194, 5.2595, 5.0072, 4.9721, 4.7392, 4.6862, 4.6194, 5.3468], device='cuda:0'), covar=tensor([0.1071, 0.0806, 0.1034, 0.0785, 0.0825, 0.0832, 0.1113, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0543, 0.0667, 0.0557, 0.0462, 0.0422, 0.0435, 0.0559, 0.0511], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:49:11,073 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2957, 4.3460, 4.7420, 4.7090, 4.7122, 4.3645, 4.3268, 4.2171], device='cuda:0'), covar=tensor([0.0303, 0.0534, 0.0356, 0.0387, 0.0446, 0.0385, 0.1041, 0.0481], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0339, 0.0342, 0.0324, 0.0386, 0.0360, 0.0461, 0.0292], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 10:49:34,942 INFO [train.py:904] (0/8) Epoch 11, batch 6100, loss[loss=0.2539, simple_loss=0.343, pruned_loss=0.08239, over 16699.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3071, pruned_loss=0.07041, over 3091694.50 frames. ], batch size: 124, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:50:51,421 INFO [optim.py:368] (0/8) 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 10:50:55,589 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-29 10:50:56,538 INFO [train.py:904] (0/8) Epoch 11, batch 6150, loss[loss=0.2056, simple_loss=0.2912, pruned_loss=0.05995, over 16499.00 frames. ], tot_loss[loss=0.2237, simple_loss=0.306, pruned_loss=0.07072, over 3082120.05 frames. ], batch size: 75, lr: 6.19e-03, grad_scale: 8.0 2023-04-29 10:51:07,622 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:52:04,919 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.03 vs. limit=5.0 2023-04-29 10:52:14,175 INFO [train.py:904] (0/8) Epoch 11, batch 6200, loss[loss=0.2361, simple_loss=0.3145, pruned_loss=0.07881, over 15357.00 frames. ], tot_loss[loss=0.222, simple_loss=0.304, pruned_loss=0.06998, over 3081868.23 frames. ], batch size: 191, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:52:24,552 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3850, 2.9617, 2.6423, 2.2529, 2.2265, 2.1950, 2.9534, 2.9325], device='cuda:0'), covar=tensor([0.2117, 0.0672, 0.1295, 0.1799, 0.1826, 0.1628, 0.0456, 0.0823], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0258, 0.0283, 0.0278, 0.0282, 0.0221, 0.0268, 0.0293], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 10:52:42,553 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:07,945 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=107737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:53:27,430 INFO [optim.py:368] (0/8) 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 10:53:30,004 INFO [train.py:904] (0/8) Epoch 11, batch 6250, loss[loss=0.1999, simple_loss=0.2982, pruned_loss=0.05076, over 16804.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.303, pruned_loss=0.06916, over 3102543.44 frames. ], batch size: 83, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:09,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8674, 3.8835, 4.2857, 4.2561, 4.2568, 3.9429, 3.9511, 3.8816], device='cuda:0'), covar=tensor([0.0346, 0.0618, 0.0369, 0.0415, 0.0460, 0.0404, 0.0982, 0.0539], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0344, 0.0347, 0.0328, 0.0391, 0.0364, 0.0468, 0.0297], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 10:54:36,281 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107796.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:54:38,996 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=107798.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 10:54:45,111 INFO [train.py:904] (0/8) Epoch 11, batch 6300, loss[loss=0.1987, simple_loss=0.2851, pruned_loss=0.05616, over 16484.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3021, pruned_loss=0.06833, over 3090610.40 frames. ], batch size: 68, lr: 6.19e-03, grad_scale: 4.0 2023-04-29 10:54:54,106 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=107808.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:55:52,344 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107844.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 10:56:00,519 INFO [optim.py:368] (0/8) 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 10:56:03,049 INFO [train.py:904] (0/8) Epoch 11, batch 6350, loss[loss=0.3031, simple_loss=0.3422, pruned_loss=0.132, over 11347.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3029, pruned_loss=0.06971, over 3082684.66 frames. ], batch size: 248, lr: 6.18e-03, grad_scale: 4.0 2023-04-29 10:56:10,040 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=107856.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:56:25,878 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=107866.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 10:57:18,102 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-29 10:57:20,964 INFO [train.py:904] (0/8) Epoch 11, batch 6400, loss[loss=0.2113, simple_loss=0.2853, pruned_loss=0.06866, over 17248.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3033, pruned_loss=0.07113, over 3070089.78 frames. ], batch size: 44, lr: 6.18e-03, grad_scale: 8.0 2023-04-29 10:58:35,869 INFO [optim.py:368] (0/8) 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 10:58:35,885 INFO [train.py:904] (0/8) Epoch 11, batch 6450, loss[loss=0.208, simple_loss=0.2984, pruned_loss=0.05884, over 16870.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3024, pruned_loss=0.06961, over 3091928.88 frames. ], batch size: 96, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 10:59:49,158 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-108000.pt 2023-04-29 10:59:54,943 INFO [train.py:904] (0/8) Epoch 11, batch 6500, loss[loss=0.2294, simple_loss=0.3155, pruned_loss=0.07167, over 15327.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3002, pruned_loss=0.06856, over 3113577.22 frames. ], batch size: 190, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:00:14,843 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:00:26,685 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.05 vs. limit=5.0 2023-04-29 11:01:12,926 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.008e+02 3.175e+02 3.633e+02 4.590e+02 1.124e+03, threshold=7.266e+02, percent-clipped=3.0 2023-04-29 11:01:12,947 INFO [train.py:904] (0/8) Epoch 11, batch 6550, loss[loss=0.2324, simple_loss=0.3222, pruned_loss=0.07127, over 16459.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3031, pruned_loss=0.06934, over 3123210.96 frames. ], batch size: 35, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:01:22,154 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6186, 2.5742, 2.3450, 3.5232, 2.2874, 3.7625, 1.2900, 2.7318], device='cuda:0'), covar=tensor([0.1447, 0.0667, 0.1221, 0.0171, 0.0180, 0.0440, 0.1728, 0.0791], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0158, 0.0182, 0.0144, 0.0199, 0.0209, 0.0181, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 11:01:31,999 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 11:02:13,593 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108093.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:02:25,864 INFO [train.py:904] (0/8) Epoch 11, batch 6600, loss[loss=0.2357, simple_loss=0.3142, pruned_loss=0.07858, over 16608.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3053, pruned_loss=0.06993, over 3122998.92 frames. ], batch size: 62, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:02:50,221 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 11:03:16,852 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9529, 5.2909, 5.0052, 4.9829, 4.7240, 4.6022, 4.7342, 5.3882], device='cuda:0'), covar=tensor([0.1068, 0.0745, 0.0971, 0.0768, 0.0820, 0.0850, 0.0939, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0544, 0.0667, 0.0558, 0.0462, 0.0421, 0.0435, 0.0557, 0.0513], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:03:27,124 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9086, 2.7698, 2.3014, 2.8173, 3.2814, 2.8632, 3.7496, 3.5463], device='cuda:0'), covar=tensor([0.0042, 0.0267, 0.0383, 0.0272, 0.0158, 0.0265, 0.0116, 0.0144], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0199, 0.0196, 0.0196, 0.0199, 0.0201, 0.0204, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:03:41,612 INFO [optim.py:368] (0/8) 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 11:03:41,633 INFO [train.py:904] (0/8) Epoch 11, batch 6650, loss[loss=0.1966, simple_loss=0.288, pruned_loss=0.05259, over 16778.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.3053, pruned_loss=0.0701, over 3129176.48 frames. ], batch size: 83, lr: 6.18e-03, grad_scale: 2.0 2023-04-29 11:04:03,259 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:04:11,068 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.84 vs. limit=5.0 2023-04-29 11:04:56,965 INFO [train.py:904] (0/8) Epoch 11, batch 6700, loss[loss=0.2633, simple_loss=0.3383, pruned_loss=0.09411, over 15338.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.303, pruned_loss=0.06917, over 3134548.77 frames. ], batch size: 190, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:05:14,879 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108214.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:05:22,033 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1539, 2.0690, 2.2288, 3.8685, 1.9944, 2.5018, 2.1842, 2.2520], device='cuda:0'), covar=tensor([0.1017, 0.3113, 0.2140, 0.0424, 0.3654, 0.2010, 0.3028, 0.2819], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0388, 0.0326, 0.0319, 0.0410, 0.0446, 0.0354, 0.0456], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:05:41,549 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.58 vs. limit=5.0 2023-04-29 11:06:13,522 INFO [optim.py:368] (0/8) 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 11:06:13,543 INFO [train.py:904] (0/8) Epoch 11, batch 6750, loss[loss=0.1882, simple_loss=0.2749, pruned_loss=0.05074, over 16944.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3028, pruned_loss=0.06969, over 3120407.30 frames. ], batch size: 90, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:06:32,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3066, 2.0210, 2.0529, 4.0405, 2.0248, 2.4923, 2.1485, 2.2245], device='cuda:0'), covar=tensor([0.0947, 0.3258, 0.2353, 0.0389, 0.3771, 0.2128, 0.3187, 0.2998], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0387, 0.0326, 0.0319, 0.0410, 0.0446, 0.0354, 0.0456], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:07:28,512 INFO [train.py:904] (0/8) Epoch 11, batch 6800, loss[loss=0.2182, simple_loss=0.3008, pruned_loss=0.06779, over 16642.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3033, pruned_loss=0.06996, over 3117913.03 frames. ], batch size: 134, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:07:48,571 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:05,631 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108325.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:08:23,023 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-29 11:08:45,535 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.125e+02 2.977e+02 3.381e+02 4.053e+02 7.016e+02, threshold=6.762e+02, percent-clipped=0.0 2023-04-29 11:08:45,557 INFO [train.py:904] (0/8) Epoch 11, batch 6850, loss[loss=0.2258, simple_loss=0.3177, pruned_loss=0.0669, over 16291.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3046, pruned_loss=0.07032, over 3115889.00 frames. ], batch size: 165, lr: 6.17e-03, grad_scale: 4.0 2023-04-29 11:08:56,189 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 11:09:01,700 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:35,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5209, 4.4518, 4.8765, 4.8628, 4.8356, 4.5529, 4.5160, 4.3773], device='cuda:0'), covar=tensor([0.0302, 0.0598, 0.0461, 0.0404, 0.0448, 0.0431, 0.0954, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0340, 0.0344, 0.0321, 0.0391, 0.0357, 0.0460, 0.0292], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 11:09:35,695 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108386.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:45,985 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108393.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:09:59,724 INFO [train.py:904] (0/8) Epoch 11, batch 6900, loss[loss=0.2204, simple_loss=0.3042, pruned_loss=0.06831, over 16704.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3069, pruned_loss=0.07011, over 3107215.39 frames. ], batch size: 62, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:10:15,131 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3264, 3.7090, 3.6800, 1.9201, 2.9025, 2.4840, 3.8033, 3.7728], device='cuda:0'), covar=tensor([0.0236, 0.0600, 0.0571, 0.1857, 0.0825, 0.0875, 0.0494, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0145, 0.0159, 0.0145, 0.0137, 0.0126, 0.0136, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 11:10:45,864 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:10:59,098 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=108441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:11:15,804 INFO [train.py:904] (0/8) Epoch 11, batch 6950, loss[loss=0.2335, simple_loss=0.3131, pruned_loss=0.07697, over 16689.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3091, pruned_loss=0.07221, over 3090590.43 frames. ], batch size: 57, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:11:17,504 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-04-29 11:11:17,888 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 2.933e+02 3.744e+02 4.621e+02 9.342e+02, threshold=7.489e+02, percent-clipped=9.0 2023-04-29 11:12:20,828 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108493.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:12:28,209 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.23 vs. limit=5.0 2023-04-29 11:12:33,554 INFO [train.py:904] (0/8) Epoch 11, batch 7000, loss[loss=0.2226, simple_loss=0.3066, pruned_loss=0.06932, over 15402.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3093, pruned_loss=0.07219, over 3062868.97 frames. ], batch size: 190, lr: 6.17e-03, grad_scale: 2.0 2023-04-29 11:13:52,266 INFO [train.py:904] (0/8) Epoch 11, batch 7050, loss[loss=0.3133, simple_loss=0.3577, pruned_loss=0.1345, over 11483.00 frames. ], tot_loss[loss=0.227, simple_loss=0.3102, pruned_loss=0.07188, over 3068509.81 frames. ], batch size: 247, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:13:53,484 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.056e+02 3.044e+02 3.927e+02 4.822e+02 1.171e+03, threshold=7.854e+02, percent-clipped=4.0 2023-04-29 11:14:32,971 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:14:36,372 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7593, 1.7559, 1.5424, 1.5395, 1.8793, 1.5635, 1.6679, 1.9316], device='cuda:0'), covar=tensor([0.0099, 0.0168, 0.0261, 0.0236, 0.0130, 0.0187, 0.0125, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0199, 0.0197, 0.0197, 0.0201, 0.0201, 0.0204, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:15:11,209 INFO [train.py:904] (0/8) Epoch 11, batch 7100, loss[loss=0.2123, simple_loss=0.3023, pruned_loss=0.06122, over 16458.00 frames. ], tot_loss[loss=0.2256, simple_loss=0.3087, pruned_loss=0.07131, over 3071816.78 frames. ], batch size: 75, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:15:47,790 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 11:16:07,300 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:16:21,933 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-29 11:16:27,134 INFO [train.py:904] (0/8) Epoch 11, batch 7150, loss[loss=0.232, simple_loss=0.3125, pruned_loss=0.07573, over 16486.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3064, pruned_loss=0.07079, over 3080179.27 frames. ], batch size: 75, lr: 6.16e-03, grad_scale: 2.0 2023-04-29 11:16:28,931 INFO [optim.py:368] (0/8) 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 11:17:10,549 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108681.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:17:41,816 INFO [train.py:904] (0/8) Epoch 11, batch 7200, loss[loss=0.1844, simple_loss=0.279, pruned_loss=0.04483, over 16459.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3036, pruned_loss=0.06902, over 3075832.61 frames. ], batch size: 75, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:17:48,152 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3214, 3.2485, 3.3737, 3.4720, 3.5212, 3.2592, 3.4552, 3.5378], device='cuda:0'), covar=tensor([0.1223, 0.0935, 0.1060, 0.0600, 0.0629, 0.2326, 0.0884, 0.0688], device='cuda:0'), in_proj_covar=tensor([0.0509, 0.0631, 0.0763, 0.0639, 0.0492, 0.0496, 0.0509, 0.0575], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:19:02,072 INFO [train.py:904] (0/8) Epoch 11, batch 7250, loss[loss=0.197, simple_loss=0.2755, pruned_loss=0.05925, over 16628.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3013, pruned_loss=0.06791, over 3073246.40 frames. ], batch size: 57, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:19:03,150 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.903e+02 3.497e+02 4.380e+02 7.660e+02, threshold=6.994e+02, percent-clipped=1.0 2023-04-29 11:19:55,675 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108788.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:20:00,940 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=108791.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:20:16,100 INFO [train.py:904] (0/8) Epoch 11, batch 7300, loss[loss=0.2183, simple_loss=0.312, pruned_loss=0.06229, over 16780.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3003, pruned_loss=0.06727, over 3080236.79 frames. ], batch size: 83, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:34,181 INFO [train.py:904] (0/8) Epoch 11, batch 7350, loss[loss=0.1917, simple_loss=0.2737, pruned_loss=0.05486, over 16353.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.3001, pruned_loss=0.06736, over 3076389.67 frames. ], batch size: 35, lr: 6.16e-03, grad_scale: 4.0 2023-04-29 11:21:34,660 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=108852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:21:35,290 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.782e+02 3.104e+02 3.844e+02 4.749e+02 1.066e+03, threshold=7.688e+02, percent-clipped=6.0 2023-04-29 11:22:18,167 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 11:22:36,091 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1168, 4.1147, 2.3330, 4.8709, 3.0487, 4.8186, 2.5481, 3.1471], device='cuda:0'), covar=tensor([0.0196, 0.0314, 0.1693, 0.0187, 0.0790, 0.0425, 0.1590, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0163, 0.0188, 0.0125, 0.0167, 0.0203, 0.0195, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 11:22:38,995 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 11:22:54,330 INFO [train.py:904] (0/8) Epoch 11, batch 7400, loss[loss=0.2332, simple_loss=0.3059, pruned_loss=0.0803, over 15244.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3011, pruned_loss=0.0681, over 3071599.06 frames. ], batch size: 191, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:23:42,504 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=108933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:24:11,405 INFO [train.py:904] (0/8) Epoch 11, batch 7450, loss[loss=0.2129, simple_loss=0.3081, pruned_loss=0.05888, over 16415.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3025, pruned_loss=0.0695, over 3061824.91 frames. ], batch size: 146, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:24:13,641 INFO [optim.py:368] (0/8) 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 11:24:57,590 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 11:24:59,227 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=108981.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:25:11,428 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9059, 2.3260, 2.2747, 2.9133, 1.8302, 3.2692, 1.6480, 2.6907], device='cuda:0'), covar=tensor([0.1211, 0.0589, 0.1059, 0.0169, 0.0143, 0.0424, 0.1423, 0.0691], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0159, 0.0182, 0.0144, 0.0201, 0.0209, 0.0182, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 11:25:31,632 INFO [train.py:904] (0/8) Epoch 11, batch 7500, loss[loss=0.1923, simple_loss=0.275, pruned_loss=0.05479, over 16801.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3022, pruned_loss=0.06868, over 3061176.80 frames. ], batch size: 39, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:10,794 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-29 11:26:15,761 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109029.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:26:27,937 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109037.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 11:26:31,549 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4709, 3.2541, 2.7548, 2.0956, 2.2730, 2.1490, 3.3115, 3.0876], device='cuda:0'), covar=tensor([0.2504, 0.0784, 0.1423, 0.2276, 0.2251, 0.1907, 0.0541, 0.1134], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0257, 0.0283, 0.0276, 0.0281, 0.0221, 0.0266, 0.0293], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:26:43,177 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9186, 2.6999, 2.7032, 1.9318, 2.5720, 2.7149, 2.6523, 1.8578], device='cuda:0'), covar=tensor([0.0343, 0.0052, 0.0050, 0.0290, 0.0086, 0.0082, 0.0073, 0.0333], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0068, 0.0069, 0.0126, 0.0078, 0.0090, 0.0077, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 11:26:51,092 INFO [train.py:904] (0/8) Epoch 11, batch 7550, loss[loss=0.1869, simple_loss=0.2733, pruned_loss=0.05026, over 16715.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3018, pruned_loss=0.06894, over 3055844.79 frames. ], batch size: 76, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:26:52,324 INFO [optim.py:368] (0/8) 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 11:27:22,188 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0050, 3.8892, 4.0657, 4.2339, 4.3380, 3.9340, 4.2454, 4.3325], device='cuda:0'), covar=tensor([0.1575, 0.1111, 0.1467, 0.0674, 0.0572, 0.1528, 0.0754, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0513, 0.0634, 0.0769, 0.0648, 0.0501, 0.0500, 0.0510, 0.0580], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:27:46,053 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:27:53,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2624, 3.3698, 3.5694, 1.6998, 3.7253, 3.7384, 2.8306, 2.6866], device='cuda:0'), covar=tensor([0.0801, 0.0192, 0.0162, 0.1226, 0.0054, 0.0137, 0.0419, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0097, 0.0087, 0.0137, 0.0068, 0.0100, 0.0117, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 11:28:02,010 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109098.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:28:06,709 INFO [train.py:904] (0/8) Epoch 11, batch 7600, loss[loss=0.2025, simple_loss=0.2849, pruned_loss=0.05999, over 16225.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3016, pruned_loss=0.06927, over 3064423.76 frames. ], batch size: 165, lr: 6.15e-03, grad_scale: 8.0 2023-04-29 11:28:56,880 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:12,239 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:29:20,703 INFO [train.py:904] (0/8) Epoch 11, batch 7650, loss[loss=0.2175, simple_loss=0.3031, pruned_loss=0.06592, over 16762.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3028, pruned_loss=0.07058, over 3046738.84 frames. ], batch size: 89, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:29:23,630 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.975e+02 2.787e+02 3.621e+02 4.390e+02 9.045e+02, threshold=7.242e+02, percent-clipped=3.0 2023-04-29 11:30:25,867 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109194.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:30:36,529 INFO [train.py:904] (0/8) Epoch 11, batch 7700, loss[loss=0.2282, simple_loss=0.3054, pruned_loss=0.07545, over 15360.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3032, pruned_loss=0.07125, over 3049877.95 frames. ], batch size: 191, lr: 6.15e-03, grad_scale: 4.0 2023-04-29 11:31:24,714 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:31:53,782 INFO [train.py:904] (0/8) Epoch 11, batch 7750, loss[loss=0.2486, simple_loss=0.3307, pruned_loss=0.08326, over 15355.00 frames. ], tot_loss[loss=0.223, simple_loss=0.3037, pruned_loss=0.07119, over 3057817.00 frames. ], batch size: 190, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:31:54,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6473, 3.8332, 2.9688, 2.2358, 2.7119, 2.3448, 4.0786, 3.5636], device='cuda:0'), covar=tensor([0.2773, 0.0730, 0.1605, 0.2333, 0.2253, 0.1770, 0.0428, 0.0953], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0258, 0.0284, 0.0278, 0.0283, 0.0223, 0.0268, 0.0295], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:31:56,717 INFO [optim.py:368] (0/8) 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 11:31:59,160 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109255.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:39,019 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:32:41,172 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3957, 2.0294, 1.5840, 1.8753, 2.3518, 2.0628, 2.3209, 2.5480], device='cuda:0'), covar=tensor([0.0114, 0.0263, 0.0371, 0.0317, 0.0155, 0.0268, 0.0153, 0.0169], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0195, 0.0193, 0.0192, 0.0195, 0.0197, 0.0198, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:32:58,051 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:33:10,723 INFO [train.py:904] (0/8) Epoch 11, batch 7800, loss[loss=0.2444, simple_loss=0.3193, pruned_loss=0.08479, over 15245.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.3041, pruned_loss=0.07156, over 3061685.00 frames. ], batch size: 190, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:33:35,186 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:34:23,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0562, 3.1039, 3.2818, 1.6350, 3.4609, 3.4972, 2.7896, 2.5722], device='cuda:0'), covar=tensor([0.0862, 0.0241, 0.0222, 0.1221, 0.0073, 0.0146, 0.0402, 0.0459], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0098, 0.0087, 0.0137, 0.0069, 0.0101, 0.0118, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 11:34:25,573 INFO [train.py:904] (0/8) Epoch 11, batch 7850, loss[loss=0.2176, simple_loss=0.3071, pruned_loss=0.06409, over 16845.00 frames. ], tot_loss[loss=0.2236, simple_loss=0.305, pruned_loss=0.07109, over 3065600.58 frames. ], batch size: 102, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:34:30,498 INFO [optim.py:368] (0/8) 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 11:34:30,931 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109355.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:35:07,106 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109379.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 11:35:22,519 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 11:35:27,278 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109393.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 11:35:41,217 INFO [train.py:904] (0/8) Epoch 11, batch 7900, loss[loss=0.2104, simple_loss=0.2873, pruned_loss=0.06677, over 16625.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3041, pruned_loss=0.07035, over 3077062.63 frames. ], batch size: 57, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:35:58,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4384, 4.4588, 4.2841, 4.0335, 3.9191, 4.3762, 4.1155, 4.0406], device='cuda:0'), covar=tensor([0.0639, 0.0498, 0.0282, 0.0306, 0.1031, 0.0480, 0.0531, 0.0683], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0307, 0.0279, 0.0257, 0.0299, 0.0297, 0.0193, 0.0325], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:36:52,137 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109447.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:37:00,096 INFO [train.py:904] (0/8) Epoch 11, batch 7950, loss[loss=0.2057, simple_loss=0.294, pruned_loss=0.0587, over 16801.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3046, pruned_loss=0.07104, over 3062106.94 frames. ], batch size: 124, lr: 6.14e-03, grad_scale: 2.0 2023-04-29 11:37:04,716 INFO [optim.py:368] (0/8) 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 11:37:38,579 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 11:38:04,155 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:38:14,153 INFO [train.py:904] (0/8) Epoch 11, batch 8000, loss[loss=0.2132, simple_loss=0.3016, pruned_loss=0.06238, over 16711.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3047, pruned_loss=0.07096, over 3069513.89 frames. ], batch size: 124, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:39:09,508 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 11:39:24,928 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109550.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:39:27,482 INFO [train.py:904] (0/8) Epoch 11, batch 8050, loss[loss=0.2164, simple_loss=0.2987, pruned_loss=0.06706, over 16957.00 frames. ], tot_loss[loss=0.2228, simple_loss=0.3045, pruned_loss=0.0705, over 3075434.34 frames. ], batch size: 41, lr: 6.14e-03, grad_scale: 4.0 2023-04-29 11:39:31,004 INFO [optim.py:368] (0/8) 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 11:39:35,181 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8868, 4.1774, 3.9662, 4.0023, 3.6593, 3.8133, 3.7911, 4.1527], device='cuda:0'), covar=tensor([0.1034, 0.0821, 0.0897, 0.0683, 0.0734, 0.1465, 0.0867, 0.0882], device='cuda:0'), in_proj_covar=tensor([0.0534, 0.0657, 0.0547, 0.0454, 0.0412, 0.0433, 0.0548, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:40:06,468 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109579.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:40:40,896 INFO [train.py:904] (0/8) Epoch 11, batch 8100, loss[loss=0.2172, simple_loss=0.2988, pruned_loss=0.06776, over 16532.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3037, pruned_loss=0.0696, over 3079535.55 frames. ], batch size: 68, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:41:39,253 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109640.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:54,497 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:41:56,619 INFO [train.py:904] (0/8) Epoch 11, batch 8150, loss[loss=0.239, simple_loss=0.3079, pruned_loss=0.08502, over 12007.00 frames. ], tot_loss[loss=0.219, simple_loss=0.3012, pruned_loss=0.06839, over 3096805.44 frames. ], batch size: 248, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:42:01,365 INFO [optim.py:368] (0/8) 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 11:42:08,200 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=109659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:42:09,596 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4274, 3.8049, 3.5064, 1.8257, 2.8382, 2.2409, 3.7402, 3.9591], device='cuda:0'), covar=tensor([0.0220, 0.0558, 0.0631, 0.2175, 0.0970, 0.1061, 0.0603, 0.0810], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0143, 0.0157, 0.0144, 0.0137, 0.0125, 0.0137, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 11:42:30,927 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109674.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:42:59,235 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109693.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:43:12,489 INFO [train.py:904] (0/8) Epoch 11, batch 8200, loss[loss=0.2132, simple_loss=0.2946, pruned_loss=0.06591, over 16762.00 frames. ], tot_loss[loss=0.2174, simple_loss=0.2988, pruned_loss=0.068, over 3086607.90 frames. ], batch size: 134, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:43:41,779 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=109720.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:44:16,311 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109741.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 11:44:33,542 INFO [train.py:904] (0/8) Epoch 11, batch 8250, loss[loss=0.2054, simple_loss=0.2979, pruned_loss=0.05646, over 16653.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2976, pruned_loss=0.06589, over 3070615.27 frames. ], batch size: 134, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:44:38,006 INFO [optim.py:368] (0/8) 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 11:45:52,516 INFO [train.py:904] (0/8) Epoch 11, batch 8300, loss[loss=0.1868, simple_loss=0.2819, pruned_loss=0.04578, over 15445.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2949, pruned_loss=0.06303, over 3050207.52 frames. ], batch size: 191, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:46:56,578 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9430, 5.3728, 5.5774, 5.3709, 5.4219, 5.9534, 5.4932, 5.2349], device='cuda:0'), covar=tensor([0.0870, 0.1694, 0.1819, 0.2060, 0.2637, 0.1008, 0.1323, 0.2332], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0471, 0.0516, 0.0408, 0.0537, 0.0540, 0.0409, 0.0553], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 11:47:09,112 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109850.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:47:11,570 INFO [train.py:904] (0/8) Epoch 11, batch 8350, loss[loss=0.229, simple_loss=0.3202, pruned_loss=0.06887, over 15227.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2939, pruned_loss=0.06075, over 3061667.07 frames. ], batch size: 191, lr: 6.13e-03, grad_scale: 4.0 2023-04-29 11:47:16,940 INFO [optim.py:368] (0/8) 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 11:48:13,997 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7409, 3.2290, 3.3964, 1.7436, 2.9338, 2.2757, 3.3055, 3.3129], device='cuda:0'), covar=tensor([0.0238, 0.0741, 0.0454, 0.2072, 0.0753, 0.0959, 0.0653, 0.0853], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0140, 0.0154, 0.0140, 0.0134, 0.0123, 0.0133, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 11:48:24,628 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109898.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:48:30,282 INFO [train.py:904] (0/8) Epoch 11, batch 8400, loss[loss=0.1862, simple_loss=0.2822, pruned_loss=0.04506, over 16777.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2905, pruned_loss=0.0581, over 3064357.58 frames. ], batch size: 89, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:21,775 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=109935.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:45,409 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:49:47,554 INFO [train.py:904] (0/8) Epoch 11, batch 8450, loss[loss=0.167, simple_loss=0.2552, pruned_loss=0.03943, over 17042.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2884, pruned_loss=0.0564, over 3054690.81 frames. ], batch size: 55, lr: 6.13e-03, grad_scale: 8.0 2023-04-29 11:49:52,359 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.575e+02 2.195e+02 2.664e+02 3.370e+02 8.341e+02, threshold=5.327e+02, percent-clipped=3.0 2023-04-29 11:50:23,577 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=109974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:00,423 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=109998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:03,565 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-110000.pt 2023-04-29 11:51:09,682 INFO [train.py:904] (0/8) Epoch 11, batch 8500, loss[loss=0.1775, simple_loss=0.2655, pruned_loss=0.04475, over 15439.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2844, pruned_loss=0.05384, over 3043534.99 frames. ], batch size: 190, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:51:31,232 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110015.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:33,889 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:51:46,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0057, 2.3684, 2.0188, 2.0843, 2.6251, 2.3459, 2.7911, 2.8587], device='cuda:0'), covar=tensor([0.0107, 0.0284, 0.0367, 0.0346, 0.0197, 0.0299, 0.0190, 0.0186], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0196, 0.0192, 0.0192, 0.0194, 0.0195, 0.0196, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:52:17,093 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110044.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:52:31,128 INFO [train.py:904] (0/8) Epoch 11, batch 8550, loss[loss=0.2041, simple_loss=0.3013, pruned_loss=0.05347, over 16698.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2817, pruned_loss=0.05262, over 3043995.94 frames. ], batch size: 134, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:52:37,012 INFO [optim.py:368] (0/8) 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 11:52:54,252 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0421, 1.8033, 1.6233, 1.5247, 1.9308, 1.6191, 1.7522, 1.9738], device='cuda:0'), covar=tensor([0.0106, 0.0223, 0.0312, 0.0289, 0.0165, 0.0229, 0.0141, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0197, 0.0192, 0.0192, 0.0195, 0.0195, 0.0196, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 11:53:07,441 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3179, 3.1982, 3.3605, 1.6066, 3.5040, 3.5876, 2.8446, 2.7331], device='cuda:0'), covar=tensor([0.0673, 0.0223, 0.0189, 0.1186, 0.0066, 0.0106, 0.0344, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0097, 0.0084, 0.0135, 0.0067, 0.0098, 0.0117, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 11:53:19,585 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110078.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:07,124 INFO [train.py:904] (0/8) Epoch 11, batch 8600, loss[loss=0.1962, simple_loss=0.2946, pruned_loss=0.04894, over 16375.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2823, pruned_loss=0.05167, over 3046721.29 frames. ], batch size: 146, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 11:54:14,109 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:22,986 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110110.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:54:35,040 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110115.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:55:43,464 INFO [train.py:904] (0/8) Epoch 11, batch 8650, loss[loss=0.1768, simple_loss=0.2628, pruned_loss=0.04539, over 11997.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2802, pruned_loss=0.04995, over 3050914.48 frames. ], batch size: 247, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:55:53,868 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.527e+02 3.196e+02 4.321e+02 7.577e+02, threshold=6.393e+02, percent-clipped=5.0 2023-04-29 11:56:28,154 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:56:39,713 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:57:30,820 INFO [train.py:904] (0/8) Epoch 11, batch 8700, loss[loss=0.1892, simple_loss=0.2908, pruned_loss=0.04376, over 16348.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2775, pruned_loss=0.04861, over 3058617.77 frames. ], batch size: 146, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:58:12,225 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.41 vs. limit=5.0 2023-04-29 11:58:34,004 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110235.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 11:59:05,968 INFO [train.py:904] (0/8) Epoch 11, batch 8750, loss[loss=0.1842, simple_loss=0.2874, pruned_loss=0.04052, over 16858.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2766, pruned_loss=0.04773, over 3055330.38 frames. ], batch size: 102, lr: 6.12e-03, grad_scale: 4.0 2023-04-29 11:59:15,717 INFO [optim.py:368] (0/8) 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 12:00:22,507 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:00:22,631 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110283.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:00,296 INFO [train.py:904] (0/8) Epoch 11, batch 8800, loss[loss=0.1954, simple_loss=0.2866, pruned_loss=0.05209, over 16680.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2757, pruned_loss=0.04719, over 3045522.37 frames. ], batch size: 134, lr: 6.12e-03, grad_scale: 8.0 2023-04-29 12:01:03,254 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110303.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:03,531 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-29 12:01:27,463 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:01:34,271 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:02:29,385 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110344.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:02:45,841 INFO [train.py:904] (0/8) Epoch 11, batch 8850, loss[loss=0.1906, simple_loss=0.2921, pruned_loss=0.04458, over 16206.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2778, pruned_loss=0.04643, over 3042090.36 frames. ], batch size: 165, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:02:52,480 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110363.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:11,310 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:29,780 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110373.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:42,475 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110379.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:03:54,838 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6387, 2.6860, 1.7867, 2.8571, 2.1618, 2.8247, 2.0173, 2.4147], device='cuda:0'), covar=tensor([0.0238, 0.0289, 0.1334, 0.0200, 0.0667, 0.0453, 0.1256, 0.0680], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0156, 0.0182, 0.0120, 0.0161, 0.0193, 0.0190, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 12:04:16,390 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6362, 3.8247, 2.2232, 4.1563, 2.8770, 4.0760, 2.3441, 2.9896], device='cuda:0'), covar=tensor([0.0203, 0.0254, 0.1476, 0.0145, 0.0682, 0.0371, 0.1375, 0.0632], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0155, 0.0182, 0.0120, 0.0161, 0.0193, 0.0190, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 12:04:27,980 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110400.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:04:30,362 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3733, 2.0809, 2.1152, 3.9668, 1.9562, 2.4749, 2.1814, 2.2647], device='cuda:0'), covar=tensor([0.0902, 0.3220, 0.2448, 0.0366, 0.3969, 0.2227, 0.3143, 0.3105], device='cuda:0'), in_proj_covar=tensor([0.0350, 0.0383, 0.0323, 0.0310, 0.0405, 0.0434, 0.0344, 0.0447], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:04:30,865 INFO [train.py:904] (0/8) Epoch 11, batch 8900, loss[loss=0.1982, simple_loss=0.2886, pruned_loss=0.05387, over 16764.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2785, pruned_loss=0.04594, over 3050768.38 frames. ], batch size: 124, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:05:57,031 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 12:06:06,515 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9509, 1.8939, 2.2901, 3.2547, 2.0873, 2.1294, 2.1138, 1.9957], device='cuda:0'), covar=tensor([0.0891, 0.3501, 0.1897, 0.0488, 0.3901, 0.2386, 0.3003, 0.3394], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0380, 0.0322, 0.0309, 0.0403, 0.0431, 0.0343, 0.0444], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:06:36,599 INFO [train.py:904] (0/8) Epoch 11, batch 8950, loss[loss=0.1987, simple_loss=0.283, pruned_loss=0.05719, over 12861.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2781, pruned_loss=0.04631, over 3042808.27 frames. ], batch size: 247, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:06:45,508 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.155e+02 2.674e+02 3.353e+02 7.252e+02, threshold=5.349e+02, percent-clipped=1.0 2023-04-29 12:06:55,484 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7938, 1.2567, 1.6211, 1.7325, 1.8275, 1.8507, 1.5729, 1.8327], device='cuda:0'), covar=tensor([0.0166, 0.0278, 0.0141, 0.0192, 0.0205, 0.0146, 0.0279, 0.0087], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0165, 0.0147, 0.0150, 0.0161, 0.0118, 0.0165, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 12:07:07,586 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110466.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:07:17,805 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110471.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:08:26,531 INFO [train.py:904] (0/8) Epoch 11, batch 9000, loss[loss=0.1628, simple_loss=0.2533, pruned_loss=0.03621, over 12109.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2749, pruned_loss=0.04496, over 3048121.66 frames. ], batch size: 250, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:08:26,532 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 12:08:36,935 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 12:08:40,843 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 12:10:00,937 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1756, 4.2097, 4.5980, 4.5673, 4.5746, 4.3146, 4.3152, 4.1678], device='cuda:0'), covar=tensor([0.0255, 0.0542, 0.0357, 0.0368, 0.0346, 0.0305, 0.0669, 0.0400], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0319, 0.0323, 0.0301, 0.0366, 0.0336, 0.0434, 0.0273], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 12:10:21,916 INFO [train.py:904] (0/8) Epoch 11, batch 9050, loss[loss=0.1693, simple_loss=0.2546, pruned_loss=0.04199, over 16836.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2763, pruned_loss=0.04567, over 3058161.09 frames. ], batch size: 83, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:10:28,907 INFO [optim.py:368] (0/8) 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 12:10:35,358 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-29 12:11:24,310 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2181, 4.2324, 4.6459, 4.6233, 4.6129, 4.3397, 4.3152, 4.2033], device='cuda:0'), covar=tensor([0.0268, 0.0499, 0.0353, 0.0375, 0.0406, 0.0312, 0.0786, 0.0396], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0319, 0.0321, 0.0301, 0.0366, 0.0336, 0.0433, 0.0273], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 12:12:01,539 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6330, 3.7992, 2.9224, 2.1656, 2.5080, 2.2837, 4.1446, 3.3627], device='cuda:0'), covar=tensor([0.2618, 0.0719, 0.1601, 0.2372, 0.2495, 0.1868, 0.0406, 0.1054], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0251, 0.0276, 0.0270, 0.0263, 0.0217, 0.0260, 0.0282], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:12:06,400 INFO [train.py:904] (0/8) Epoch 11, batch 9100, loss[loss=0.1898, simple_loss=0.2829, pruned_loss=0.04837, over 16756.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2761, pruned_loss=0.04623, over 3069611.17 frames. ], batch size: 124, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:12:25,166 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110611.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:13:30,143 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8237, 5.1408, 4.9478, 4.9512, 4.6496, 4.6367, 4.5792, 5.1904], device='cuda:0'), covar=tensor([0.1056, 0.0842, 0.0877, 0.0555, 0.0776, 0.0796, 0.0984, 0.0889], device='cuda:0'), in_proj_covar=tensor([0.0525, 0.0646, 0.0535, 0.0451, 0.0408, 0.0426, 0.0546, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:13:36,699 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110639.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:06,393 INFO [train.py:904] (0/8) Epoch 11, batch 9150, loss[loss=0.1796, simple_loss=0.2746, pruned_loss=0.04233, over 15120.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2757, pruned_loss=0.04557, over 3055319.49 frames. ], batch size: 190, lr: 6.11e-03, grad_scale: 8.0 2023-04-29 12:14:15,990 INFO [optim.py:368] (0/8) 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 12:14:23,404 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:52,221 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110672.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:53,810 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:14:56,096 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110674.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:15:45,468 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110700.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:15:48,934 INFO [train.py:904] (0/8) Epoch 11, batch 9200, loss[loss=0.1923, simple_loss=0.2811, pruned_loss=0.05176, over 16792.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.271, pruned_loss=0.04425, over 3069638.05 frames. ], batch size: 124, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:16:08,326 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4165, 4.4409, 4.7844, 4.7679, 4.7961, 4.5052, 4.4623, 4.2968], device='cuda:0'), covar=tensor([0.0256, 0.0510, 0.0434, 0.0429, 0.0391, 0.0401, 0.0797, 0.0386], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0316, 0.0322, 0.0299, 0.0362, 0.0336, 0.0431, 0.0271], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 12:16:08,438 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6812, 1.6370, 2.0906, 2.6377, 2.4712, 2.8955, 1.9369, 2.9381], device='cuda:0'), covar=tensor([0.0143, 0.0373, 0.0243, 0.0189, 0.0217, 0.0124, 0.0358, 0.0092], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0165, 0.0148, 0.0151, 0.0162, 0.0117, 0.0166, 0.0108], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 12:16:24,196 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:16:52,960 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6638, 4.5208, 4.7266, 4.8961, 5.0252, 4.4549, 5.0247, 5.0220], device='cuda:0'), covar=tensor([0.1400, 0.0981, 0.1197, 0.0506, 0.0528, 0.0862, 0.0429, 0.0497], device='cuda:0'), in_proj_covar=tensor([0.0493, 0.0615, 0.0732, 0.0621, 0.0478, 0.0482, 0.0494, 0.0560], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:16:53,007 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110748.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:17:24,352 INFO [train.py:904] (0/8) Epoch 11, batch 9250, loss[loss=0.1677, simple_loss=0.2654, pruned_loss=0.03497, over 16126.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2707, pruned_loss=0.04448, over 3032733.91 frames. ], batch size: 165, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:17:32,957 INFO [optim.py:368] (0/8) 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 12:17:53,110 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:18:05,087 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110771.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:06,557 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110797.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 12:19:16,920 INFO [train.py:904] (0/8) Epoch 11, batch 9300, loss[loss=0.1778, simple_loss=0.2654, pruned_loss=0.04504, over 16733.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2693, pruned_loss=0.04406, over 3018985.89 frames. ], batch size: 124, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:19:45,579 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110814.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:19:50,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3897, 3.3340, 3.4528, 3.5429, 3.5912, 3.2430, 3.5203, 3.6270], device='cuda:0'), covar=tensor([0.1181, 0.0873, 0.1080, 0.0604, 0.0549, 0.2493, 0.0868, 0.0609], device='cuda:0'), in_proj_covar=tensor([0.0488, 0.0608, 0.0725, 0.0617, 0.0472, 0.0476, 0.0489, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:19:59,653 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110819.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:20:37,053 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 12:21:01,159 INFO [train.py:904] (0/8) Epoch 11, batch 9350, loss[loss=0.2085, simple_loss=0.2935, pruned_loss=0.06173, over 12168.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2691, pruned_loss=0.04419, over 3020590.56 frames. ], batch size: 247, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:21:10,066 INFO [optim.py:368] (0/8) 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 12:21:35,996 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:22:41,701 INFO [train.py:904] (0/8) Epoch 11, batch 9400, loss[loss=0.1746, simple_loss=0.2685, pruned_loss=0.04037, over 16951.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2697, pruned_loss=0.044, over 3035082.79 frames. ], batch size: 41, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:22:46,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3972, 3.3736, 3.4651, 3.5497, 3.5837, 3.2686, 3.5592, 3.6353], device='cuda:0'), covar=tensor([0.1228, 0.0780, 0.0943, 0.0543, 0.0598, 0.1960, 0.0758, 0.0649], device='cuda:0'), in_proj_covar=tensor([0.0491, 0.0611, 0.0728, 0.0619, 0.0476, 0.0480, 0.0493, 0.0556], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:23:36,566 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=110929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:23:57,213 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110939.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:13,717 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5448, 3.6315, 3.4018, 3.1971, 3.2712, 3.5571, 3.3303, 3.3558], device='cuda:0'), covar=tensor([0.0537, 0.0369, 0.0254, 0.0220, 0.0547, 0.0353, 0.1229, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0295, 0.0271, 0.0252, 0.0287, 0.0291, 0.0189, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:24:21,281 INFO [train.py:904] (0/8) Epoch 11, batch 9450, loss[loss=0.1888, simple_loss=0.2834, pruned_loss=0.04708, over 15353.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2709, pruned_loss=0.04419, over 3015769.52 frames. ], batch size: 191, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:24:27,346 INFO [optim.py:368] (0/8) 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 12:24:33,131 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=110958.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:34,637 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:24:50,266 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=110967.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:06,794 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=110974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:25:33,501 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=110987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:00,943 INFO [train.py:904] (0/8) Epoch 11, batch 9500, loss[loss=0.1735, simple_loss=0.2649, pruned_loss=0.041, over 12813.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2698, pruned_loss=0.04334, over 3047735.11 frames. ], batch size: 248, lr: 6.10e-03, grad_scale: 8.0 2023-04-29 12:26:15,364 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111007.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:17,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8338, 4.0563, 2.3292, 4.4650, 2.8437, 4.3325, 2.4556, 3.1653], device='cuda:0'), covar=tensor([0.0193, 0.0264, 0.1453, 0.0162, 0.0755, 0.0461, 0.1435, 0.0617], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0155, 0.0181, 0.0120, 0.0162, 0.0192, 0.0191, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 12:26:32,222 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8635, 2.7112, 2.6290, 1.9334, 2.5554, 2.7250, 2.6535, 1.9393], device='cuda:0'), covar=tensor([0.0362, 0.0042, 0.0050, 0.0308, 0.0084, 0.0074, 0.0062, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0124, 0.0065, 0.0066, 0.0122, 0.0075, 0.0085, 0.0074, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 12:26:36,382 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9532, 2.3682, 2.3412, 3.1313, 2.0313, 3.3780, 1.6465, 2.7555], device='cuda:0'), covar=tensor([0.1241, 0.0557, 0.0969, 0.0145, 0.0087, 0.0385, 0.1425, 0.0608], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0154, 0.0175, 0.0137, 0.0184, 0.0201, 0.0178, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-04-29 12:26:37,626 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:26:43,502 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111022.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:27:46,824 INFO [train.py:904] (0/8) Epoch 11, batch 9550, loss[loss=0.1773, simple_loss=0.2765, pruned_loss=0.03909, over 16852.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2701, pruned_loss=0.0436, over 3060589.14 frames. ], batch size: 96, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:27:55,304 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.451e+02 2.456e+02 2.878e+02 3.283e+02 5.727e+02, threshold=5.755e+02, percent-clipped=0.0 2023-04-29 12:28:33,397 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 12:29:10,436 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111092.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 12:29:26,775 INFO [train.py:904] (0/8) Epoch 11, batch 9600, loss[loss=0.2096, simple_loss=0.3028, pruned_loss=0.0582, over 16330.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.271, pruned_loss=0.04382, over 3068521.74 frames. ], batch size: 165, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:29:32,529 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.4793, 2.5676, 2.4292, 4.2276, 2.7473, 4.1272, 1.2413, 2.9163], device='cuda:0'), covar=tensor([0.1538, 0.0756, 0.1140, 0.0119, 0.0139, 0.0345, 0.1705, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0155, 0.0176, 0.0138, 0.0185, 0.0203, 0.0180, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 12:29:47,525 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9507, 1.9644, 2.2842, 3.1932, 2.0691, 2.1677, 2.2282, 2.0357], device='cuda:0'), covar=tensor([0.0899, 0.3281, 0.1951, 0.0532, 0.3865, 0.2333, 0.2789, 0.3237], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0373, 0.0319, 0.0304, 0.0398, 0.0422, 0.0338, 0.0435], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:30:27,312 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9631, 3.1304, 1.7706, 3.3267, 2.1925, 3.2641, 1.8927, 2.5087], device='cuda:0'), covar=tensor([0.0305, 0.0377, 0.1749, 0.0190, 0.0944, 0.0597, 0.1701, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0155, 0.0183, 0.0120, 0.0163, 0.0193, 0.0192, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 12:31:14,976 INFO [train.py:904] (0/8) Epoch 11, batch 9650, loss[loss=0.1684, simple_loss=0.2626, pruned_loss=0.03711, over 16447.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2724, pruned_loss=0.04379, over 3065983.88 frames. ], batch size: 68, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:31:24,140 INFO [optim.py:368] (0/8) 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 12:31:25,495 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 12:32:01,406 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0841, 1.4893, 1.7490, 2.0434, 2.1378, 2.2423, 1.7306, 2.1632], device='cuda:0'), covar=tensor([0.0212, 0.0347, 0.0217, 0.0237, 0.0209, 0.0160, 0.0327, 0.0087], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0166, 0.0147, 0.0152, 0.0160, 0.0118, 0.0165, 0.0107], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 12:33:03,910 INFO [train.py:904] (0/8) Epoch 11, batch 9700, loss[loss=0.191, simple_loss=0.2802, pruned_loss=0.05086, over 16904.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2717, pruned_loss=0.04387, over 3060928.99 frames. ], batch size: 109, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:33:37,735 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3066, 4.2879, 4.1345, 3.7559, 4.1563, 1.6067, 3.9867, 4.0170], device='cuda:0'), covar=tensor([0.0086, 0.0080, 0.0167, 0.0224, 0.0089, 0.2379, 0.0119, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0125, 0.0110, 0.0158, 0.0143, 0.0129, 0.0177, 0.0145, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:33:49,088 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111224.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:33:51,793 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:34:46,303 INFO [train.py:904] (0/8) Epoch 11, batch 9750, loss[loss=0.1736, simple_loss=0.2693, pruned_loss=0.03891, over 15431.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2715, pruned_loss=0.04413, over 3057413.25 frames. ], batch size: 190, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:34:53,731 INFO [optim.py:368] (0/8) 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 12:35:17,251 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111267.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:35:27,707 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7481, 2.5117, 2.1919, 3.5918, 2.4501, 3.7418, 1.3657, 2.7490], device='cuda:0'), covar=tensor([0.1259, 0.0643, 0.1155, 0.0133, 0.0115, 0.0436, 0.1485, 0.0743], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0155, 0.0176, 0.0138, 0.0184, 0.0204, 0.0180, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 12:35:57,931 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:36:09,524 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0477, 3.9358, 4.1370, 4.2482, 4.3767, 3.9042, 4.3602, 4.3964], device='cuda:0'), covar=tensor([0.1403, 0.0989, 0.1151, 0.0614, 0.0456, 0.1294, 0.0501, 0.0558], device='cuda:0'), in_proj_covar=tensor([0.0487, 0.0609, 0.0727, 0.0619, 0.0470, 0.0479, 0.0493, 0.0556], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:36:25,546 INFO [train.py:904] (0/8) Epoch 11, batch 9800, loss[loss=0.1809, simple_loss=0.282, pruned_loss=0.03991, over 16524.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2714, pruned_loss=0.0432, over 3071600.48 frames. ], batch size: 62, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:36:49,121 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111315.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:38:06,135 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5035, 3.4504, 2.7116, 2.0878, 2.1503, 2.2033, 3.5593, 3.0411], device='cuda:0'), covar=tensor([0.2531, 0.0666, 0.1528, 0.2369, 0.2443, 0.1851, 0.0400, 0.1114], device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0248, 0.0274, 0.0269, 0.0255, 0.0216, 0.0256, 0.0279], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:38:11,451 INFO [train.py:904] (0/8) Epoch 11, batch 9850, loss[loss=0.18, simple_loss=0.2761, pruned_loss=0.04197, over 15428.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2726, pruned_loss=0.04298, over 3076766.18 frames. ], batch size: 191, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:38:20,192 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.271e+02 2.860e+02 3.428e+02 8.615e+02, threshold=5.720e+02, percent-clipped=1.0 2023-04-29 12:39:38,664 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2301, 3.3694, 1.8687, 3.7263, 2.4240, 3.6376, 2.0986, 2.6599], device='cuda:0'), covar=tensor([0.0267, 0.0380, 0.1740, 0.0153, 0.0961, 0.0522, 0.1598, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0156, 0.0184, 0.0120, 0.0163, 0.0194, 0.0193, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 12:39:42,367 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111392.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 12:39:58,892 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-04-29 12:39:58,985 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-29 12:40:02,796 INFO [train.py:904] (0/8) Epoch 11, batch 9900, loss[loss=0.1657, simple_loss=0.2751, pruned_loss=0.02819, over 16894.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2733, pruned_loss=0.04313, over 3070244.81 frames. ], batch size: 102, lr: 6.09e-03, grad_scale: 8.0 2023-04-29 12:41:34,262 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111440.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:42:01,114 INFO [train.py:904] (0/8) Epoch 11, batch 9950, loss[loss=0.166, simple_loss=0.2638, pruned_loss=0.03412, over 16843.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2746, pruned_loss=0.04337, over 3067361.82 frames. ], batch size: 90, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:42:11,473 INFO [optim.py:368] (0/8) 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 12:42:40,577 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=111468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:02,039 INFO [train.py:904] (0/8) Epoch 11, batch 10000, loss[loss=0.1757, simple_loss=0.2807, pruned_loss=0.03533, over 16875.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2733, pruned_loss=0.04302, over 3079090.51 frames. ], batch size: 96, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:44:19,754 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8548, 3.2208, 3.3918, 1.9202, 2.9876, 2.1976, 3.3853, 3.3460], device='cuda:0'), covar=tensor([0.0269, 0.0673, 0.0463, 0.1686, 0.0620, 0.0905, 0.0592, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0132, 0.0152, 0.0139, 0.0131, 0.0121, 0.0130, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 12:44:45,197 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:55,188 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:44:55,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3625, 2.0571, 2.1011, 3.9909, 2.0457, 2.5426, 2.2346, 2.2718], device='cuda:0'), covar=tensor([0.0907, 0.3344, 0.2407, 0.0362, 0.3794, 0.2130, 0.3064, 0.3056], device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0376, 0.0321, 0.0305, 0.0400, 0.0424, 0.0341, 0.0436], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:45:42,144 INFO [train.py:904] (0/8) Epoch 11, batch 10050, loss[loss=0.1937, simple_loss=0.2889, pruned_loss=0.04923, over 16321.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2734, pruned_loss=0.04306, over 3076616.15 frames. ], batch size: 166, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:45:50,239 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.518e+02 2.160e+02 2.693e+02 3.599e+02 6.171e+02, threshold=5.385e+02, percent-clipped=1.0 2023-04-29 12:46:22,298 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111572.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:46:38,856 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111581.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:47:12,032 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-29 12:47:14,835 INFO [train.py:904] (0/8) Epoch 11, batch 10100, loss[loss=0.1759, simple_loss=0.2699, pruned_loss=0.04093, over 15312.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2743, pruned_loss=0.04354, over 3072796.67 frames. ], batch size: 190, lr: 6.08e-03, grad_scale: 8.0 2023-04-29 12:47:35,966 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111614.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:48:31,868 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-11.pt 2023-04-29 12:48:58,027 INFO [train.py:904] (0/8) Epoch 12, batch 0, loss[loss=0.2483, simple_loss=0.3207, pruned_loss=0.08794, over 16603.00 frames. ], tot_loss[loss=0.2483, simple_loss=0.3207, pruned_loss=0.08794, over 16603.00 frames. ], batch size: 62, lr: 5.82e-03, grad_scale: 8.0 2023-04-29 12:48:58,028 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 12:49:05,314 INFO [train.py:938] (0/8) Epoch 12, validation: loss=0.1543, simple_loss=0.2577, pruned_loss=0.0254, over 944034.00 frames. 2023-04-29 12:49:05,315 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 12:49:12,538 INFO [optim.py:368] (0/8) 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 12:49:21,582 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:50:15,989 INFO [train.py:904] (0/8) Epoch 12, batch 50, loss[loss=0.2022, simple_loss=0.2757, pruned_loss=0.06436, over 16866.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2875, pruned_loss=0.06644, over 746155.70 frames. ], batch size: 102, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:50:44,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9546, 4.3670, 4.5344, 3.1427, 3.8031, 4.4514, 4.0489, 2.6314], device='cuda:0'), covar=tensor([0.0356, 0.0043, 0.0026, 0.0278, 0.0079, 0.0053, 0.0052, 0.0347], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0067, 0.0068, 0.0124, 0.0077, 0.0086, 0.0075, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 12:51:25,714 INFO [train.py:904] (0/8) Epoch 12, batch 100, loss[loss=0.1925, simple_loss=0.2903, pruned_loss=0.04734, over 17138.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2806, pruned_loss=0.06081, over 1312771.47 frames. ], batch size: 49, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:51:34,350 INFO [optim.py:368] (0/8) 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 12:52:18,875 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 12, batch 150, loss[loss=0.1972, simple_loss=0.2727, pruned_loss=0.06083, over 16700.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2807, pruned_loss=0.05991, over 1752565.25 frames. ], batch size: 134, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:02,376 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=111824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:53:40,729 INFO [train.py:904] (0/8) Epoch 12, batch 200, loss[loss=0.2343, simple_loss=0.3139, pruned_loss=0.0773, over 16672.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2797, pruned_loss=0.05897, over 2098931.20 frames. ], batch size: 62, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:53:41,302 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=111852.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 12:53:50,238 INFO [optim.py:368] (0/8) 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 12:54:21,346 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=111881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:54:49,930 INFO [train.py:904] (0/8) Epoch 12, batch 250, loss[loss=0.2067, simple_loss=0.2722, pruned_loss=0.07062, over 16812.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2777, pruned_loss=0.05873, over 2370967.90 frames. ], batch size: 83, lr: 5.82e-03, grad_scale: 2.0 2023-04-29 12:55:00,656 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7970, 2.8948, 2.3130, 2.6899, 3.2605, 3.0303, 3.5431, 3.4153], device='cuda:0'), covar=tensor([0.0084, 0.0310, 0.0441, 0.0347, 0.0225, 0.0267, 0.0260, 0.0193], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0210, 0.0201, 0.0202, 0.0207, 0.0206, 0.0209, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 12:55:26,516 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=111929.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:55:57,754 INFO [train.py:904] (0/8) Epoch 12, batch 300, loss[loss=0.2009, simple_loss=0.2754, pruned_loss=0.06319, over 16877.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2735, pruned_loss=0.05577, over 2587210.80 frames. ], batch size: 116, lr: 5.82e-03, grad_scale: 1.0 2023-04-29 12:56:09,474 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.376e+02 2.773e+02 3.192e+02 6.381e+02, threshold=5.545e+02, percent-clipped=1.0 2023-04-29 12:56:50,032 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-29 12:57:04,338 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-112000.pt 2023-04-29 12:57:10,674 INFO [train.py:904] (0/8) Epoch 12, batch 350, loss[loss=0.1522, simple_loss=0.2352, pruned_loss=0.03454, over 15859.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2705, pruned_loss=0.0543, over 2741546.21 frames. ], batch size: 35, lr: 5.81e-03, grad_scale: 1.0 2023-04-29 12:58:17,747 INFO [train.py:904] (0/8) Epoch 12, batch 400, loss[loss=0.2216, simple_loss=0.2905, pruned_loss=0.07638, over 16911.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2696, pruned_loss=0.0543, over 2873234.61 frames. ], batch size: 96, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:58:27,674 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.508e+02 2.981e+02 3.635e+02 6.653e+02, threshold=5.962e+02, percent-clipped=1.0 2023-04-29 12:58:36,990 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112066.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 12:59:26,054 INFO [train.py:904] (0/8) Epoch 12, batch 450, loss[loss=0.1741, simple_loss=0.2452, pruned_loss=0.05151, over 16811.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2675, pruned_loss=0.05341, over 2968530.38 frames. ], batch size: 116, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 12:59:44,628 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4473, 4.4801, 4.6201, 4.5042, 4.4533, 5.0853, 4.5601, 4.2383], device='cuda:0'), covar=tensor([0.1355, 0.1972, 0.2066, 0.2219, 0.3054, 0.1162, 0.1717, 0.2787], device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0501, 0.0550, 0.0435, 0.0581, 0.0579, 0.0436, 0.0583], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 12:59:56,013 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:00,103 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:00:06,778 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7675, 3.5548, 3.9986, 2.8858, 3.5079, 3.9704, 3.6939, 2.4920], device='cuda:0'), covar=tensor([0.0330, 0.0186, 0.0034, 0.0258, 0.0079, 0.0072, 0.0064, 0.0320], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0069, 0.0069, 0.0125, 0.0078, 0.0088, 0.0077, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:00:27,076 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112147.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 13:00:33,805 INFO [train.py:904] (0/8) Epoch 12, batch 500, loss[loss=0.1642, simple_loss=0.2594, pruned_loss=0.03455, over 17064.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2653, pruned_loss=0.05177, over 3044728.22 frames. ], batch size: 50, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:00:41,732 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 13:00:45,215 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:01:44,749 INFO [train.py:904] (0/8) Epoch 12, batch 550, loss[loss=0.2177, simple_loss=0.2901, pruned_loss=0.07266, over 15602.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2644, pruned_loss=0.05098, over 3098452.57 frames. ], batch size: 191, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:02:55,025 INFO [train.py:904] (0/8) Epoch 12, batch 600, loss[loss=0.1453, simple_loss=0.2326, pruned_loss=0.02899, over 17175.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2644, pruned_loss=0.05166, over 3149760.81 frames. ], batch size: 43, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:03:06,853 INFO [optim.py:368] (0/8) 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 13:03:44,829 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-04-29 13:04:04,841 INFO [train.py:904] (0/8) Epoch 12, batch 650, loss[loss=0.177, simple_loss=0.2501, pruned_loss=0.05194, over 16780.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.263, pruned_loss=0.05112, over 3187742.62 frames. ], batch size: 83, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:04:16,760 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0264, 1.9396, 2.5606, 3.0424, 2.8427, 3.4082, 2.3632, 3.3925], device='cuda:0'), covar=tensor([0.0172, 0.0353, 0.0226, 0.0226, 0.0229, 0.0154, 0.0325, 0.0112], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0174, 0.0155, 0.0162, 0.0170, 0.0127, 0.0174, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 13:04:49,116 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0567, 3.2110, 3.2490, 2.0443, 2.7572, 2.4057, 3.5526, 3.4636], device='cuda:0'), covar=tensor([0.0306, 0.0871, 0.0593, 0.1665, 0.0882, 0.0914, 0.0602, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0142, 0.0158, 0.0143, 0.0136, 0.0124, 0.0136, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 13:05:14,232 INFO [train.py:904] (0/8) Epoch 12, batch 700, loss[loss=0.172, simple_loss=0.2682, pruned_loss=0.03783, over 17053.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2632, pruned_loss=0.05031, over 3219618.98 frames. ], batch size: 50, lr: 5.81e-03, grad_scale: 2.0 2023-04-29 13:05:26,020 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.678e+02 2.509e+02 2.916e+02 3.539e+02 5.225e+02, threshold=5.832e+02, percent-clipped=0.0 2023-04-29 13:06:06,985 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112389.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:06:11,411 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7767, 3.0488, 2.6262, 4.4418, 3.5931, 4.3338, 1.6886, 3.0930], device='cuda:0'), covar=tensor([0.1315, 0.0586, 0.1019, 0.0178, 0.0223, 0.0360, 0.1393, 0.0713], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0159, 0.0179, 0.0146, 0.0191, 0.0210, 0.0182, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 13:06:24,424 INFO [train.py:904] (0/8) Epoch 12, batch 750, loss[loss=0.1604, simple_loss=0.2459, pruned_loss=0.0374, over 16809.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2634, pruned_loss=0.05027, over 3245613.00 frames. ], batch size: 42, lr: 5.80e-03, grad_scale: 2.0 2023-04-29 13:06:38,211 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9072, 2.0730, 2.3050, 3.1782, 2.1459, 2.2573, 2.2632, 2.1533], device='cuda:0'), covar=tensor([0.1148, 0.3063, 0.2040, 0.0616, 0.3609, 0.2222, 0.2723, 0.3095], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0394, 0.0334, 0.0322, 0.0415, 0.0449, 0.0358, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:06:51,837 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112422.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:06:52,175 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 13:07:28,379 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112447.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:07:32,715 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:07:34,642 INFO [train.py:904] (0/8) Epoch 12, batch 800, loss[loss=0.1395, simple_loss=0.2311, pruned_loss=0.0239, over 17028.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2635, pruned_loss=0.05003, over 3261791.94 frames. ], batch size: 41, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:07:45,054 INFO [optim.py:368] (0/8) 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 13:08:10,482 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7985, 2.2416, 2.3463, 4.6434, 2.2954, 2.7237, 2.4221, 2.4918], device='cuda:0'), covar=tensor([0.0871, 0.3429, 0.2430, 0.0331, 0.3767, 0.2302, 0.2872, 0.3466], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0392, 0.0332, 0.0321, 0.0413, 0.0448, 0.0356, 0.0459], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:08:33,909 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112495.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 13:08:39,210 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4503, 3.4772, 3.6648, 1.8246, 3.8480, 3.8055, 2.9691, 2.7718], device='cuda:0'), covar=tensor([0.0666, 0.0164, 0.0155, 0.1078, 0.0059, 0.0159, 0.0361, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0100, 0.0087, 0.0140, 0.0070, 0.0105, 0.0120, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 13:08:42,886 INFO [train.py:904] (0/8) Epoch 12, batch 850, loss[loss=0.1679, simple_loss=0.2437, pruned_loss=0.04605, over 16250.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2624, pruned_loss=0.04911, over 3278298.75 frames. ], batch size: 165, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:09:52,005 INFO [train.py:904] (0/8) Epoch 12, batch 900, loss[loss=0.173, simple_loss=0.2675, pruned_loss=0.03927, over 17123.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2616, pruned_loss=0.04867, over 3285077.82 frames. ], batch size: 47, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:10:02,366 INFO [optim.py:368] (0/8) 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 13:10:05,186 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112562.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:23,963 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7544, 4.3633, 2.9920, 2.1607, 2.8802, 2.4048, 4.5302, 3.8099], device='cuda:0'), covar=tensor([0.2664, 0.0581, 0.1699, 0.2252, 0.2492, 0.1810, 0.0409, 0.0947], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0259, 0.0284, 0.0277, 0.0275, 0.0223, 0.0266, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:10:26,121 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:10:59,467 INFO [train.py:904] (0/8) Epoch 12, batch 950, loss[loss=0.1843, simple_loss=0.2712, pruned_loss=0.04875, over 17144.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2612, pruned_loss=0.0492, over 3294739.36 frames. ], batch size: 47, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:11:19,487 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 13:11:21,313 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:28,349 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112623.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:11:50,177 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112638.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:11:58,163 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6138, 2.7986, 2.6372, 4.7859, 4.0133, 4.4535, 1.5865, 3.2446], device='cuda:0'), covar=tensor([0.1379, 0.0687, 0.1110, 0.0203, 0.0290, 0.0360, 0.1428, 0.0705], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0159, 0.0179, 0.0146, 0.0191, 0.0209, 0.0180, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 13:12:07,861 INFO [train.py:904] (0/8) Epoch 12, batch 1000, loss[loss=0.1489, simple_loss=0.2266, pruned_loss=0.03557, over 16975.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2594, pruned_loss=0.04894, over 3301713.72 frames. ], batch size: 41, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:12:09,478 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5723, 2.5533, 2.1984, 2.3731, 2.8759, 2.7012, 3.3202, 3.1044], device='cuda:0'), covar=tensor([0.0084, 0.0312, 0.0368, 0.0356, 0.0242, 0.0288, 0.0200, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0211, 0.0204, 0.0204, 0.0210, 0.0210, 0.0215, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:12:18,370 INFO [optim.py:368] (0/8) 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 13:12:43,318 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112678.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 13:13:15,640 INFO [train.py:904] (0/8) Epoch 12, batch 1050, loss[loss=0.1767, simple_loss=0.2641, pruned_loss=0.04464, over 16847.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2599, pruned_loss=0.04838, over 3317464.75 frames. ], batch size: 42, lr: 5.80e-03, grad_scale: 4.0 2023-04-29 13:13:42,872 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=112722.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:14,524 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112745.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:14:23,363 INFO [train.py:904] (0/8) Epoch 12, batch 1100, loss[loss=0.1945, simple_loss=0.2861, pruned_loss=0.0514, over 17055.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2602, pruned_loss=0.04802, over 3328540.88 frames. ], batch size: 55, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:14:34,078 INFO [optim.py:368] (0/8) 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 13:14:43,400 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 13:14:48,135 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=112770.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:15:33,624 INFO [train.py:904] (0/8) Epoch 12, batch 1150, loss[loss=0.172, simple_loss=0.249, pruned_loss=0.0475, over 16201.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2594, pruned_loss=0.04764, over 3328795.15 frames. ], batch size: 164, lr: 5.79e-03, grad_scale: 4.0 2023-04-29 13:16:42,806 INFO [train.py:904] (0/8) Epoch 12, batch 1200, loss[loss=0.1903, simple_loss=0.2667, pruned_loss=0.05694, over 16774.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2587, pruned_loss=0.04752, over 3328701.02 frames. ], batch size: 102, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:16:52,676 INFO [optim.py:368] (0/8) 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 13:17:24,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1882, 4.8888, 5.1884, 5.4641, 5.6018, 4.9346, 5.4993, 5.5208], device='cuda:0'), covar=tensor([0.1538, 0.1236, 0.1578, 0.0568, 0.0458, 0.0676, 0.0461, 0.0627], device='cuda:0'), in_proj_covar=tensor([0.0555, 0.0687, 0.0836, 0.0700, 0.0529, 0.0534, 0.0549, 0.0633], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:17:49,340 INFO [train.py:904] (0/8) Epoch 12, batch 1250, loss[loss=0.1696, simple_loss=0.2558, pruned_loss=0.04165, over 17206.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2595, pruned_loss=0.04799, over 3328912.95 frames. ], batch size: 46, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:18:08,056 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-29 13:18:12,265 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:16,577 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=112921.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:18:32,357 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112933.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 13:18:54,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4556, 4.2633, 4.4842, 4.6894, 4.7990, 4.3520, 4.6646, 4.7532], device='cuda:0'), covar=tensor([0.1532, 0.1062, 0.1379, 0.0634, 0.0599, 0.1023, 0.1251, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0558, 0.0690, 0.0839, 0.0704, 0.0534, 0.0537, 0.0552, 0.0636], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:18:57,927 INFO [train.py:904] (0/8) Epoch 12, batch 1300, loss[loss=0.1728, simple_loss=0.2725, pruned_loss=0.0366, over 17144.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2582, pruned_loss=0.04741, over 3326568.37 frames. ], batch size: 49, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:19:09,594 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.880e+02 2.412e+02 2.831e+02 3.329e+02 6.613e+02, threshold=5.661e+02, percent-clipped=2.0 2023-04-29 13:19:27,117 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=112973.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 13:19:41,049 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=112982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:20:08,554 INFO [train.py:904] (0/8) Epoch 12, batch 1350, loss[loss=0.174, simple_loss=0.2595, pruned_loss=0.04431, over 16414.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.259, pruned_loss=0.04765, over 3325168.62 frames. ], batch size: 75, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:20:15,836 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7166, 3.7766, 2.0168, 4.0295, 2.8148, 4.0037, 2.0750, 2.8941], device='cuda:0'), covar=tensor([0.0202, 0.0269, 0.1627, 0.0271, 0.0712, 0.0405, 0.1582, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0168, 0.0191, 0.0137, 0.0169, 0.0211, 0.0200, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 13:21:07,362 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113045.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:21:16,384 INFO [train.py:904] (0/8) Epoch 12, batch 1400, loss[loss=0.1841, simple_loss=0.2624, pruned_loss=0.05295, over 16732.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2597, pruned_loss=0.04849, over 3327206.82 frames. ], batch size: 124, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:21:26,437 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.826e+02 2.440e+02 3.018e+02 3.818e+02 8.239e+02, threshold=6.035e+02, percent-clipped=8.0 2023-04-29 13:22:12,591 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:22:24,199 INFO [train.py:904] (0/8) Epoch 12, batch 1450, loss[loss=0.1837, simple_loss=0.2545, pruned_loss=0.05646, over 16640.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2588, pruned_loss=0.04818, over 3317142.91 frames. ], batch size: 124, lr: 5.79e-03, grad_scale: 8.0 2023-04-29 13:23:24,521 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6769, 3.9778, 3.0703, 2.2992, 2.6713, 2.3389, 4.0014, 3.5341], device='cuda:0'), covar=tensor([0.2547, 0.0574, 0.1404, 0.2493, 0.2445, 0.1796, 0.0470, 0.1266], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0260, 0.0284, 0.0279, 0.0279, 0.0225, 0.0268, 0.0301], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:23:35,103 INFO [train.py:904] (0/8) Epoch 12, batch 1500, loss[loss=0.18, simple_loss=0.2472, pruned_loss=0.05643, over 16454.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2589, pruned_loss=0.04853, over 3315354.60 frames. ], batch size: 146, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:23:45,778 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.499e+02 2.892e+02 3.478e+02 9.992e+02, threshold=5.783e+02, percent-clipped=3.0 2023-04-29 13:23:50,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3101, 4.2092, 4.1937, 3.9798, 3.9549, 4.2647, 3.9924, 4.0514], device='cuda:0'), covar=tensor([0.0610, 0.0629, 0.0284, 0.0269, 0.0744, 0.0442, 0.0677, 0.0598], device='cuda:0'), in_proj_covar=tensor([0.0262, 0.0341, 0.0311, 0.0288, 0.0331, 0.0336, 0.0212, 0.0360], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:24:43,445 INFO [train.py:904] (0/8) Epoch 12, batch 1550, loss[loss=0.1724, simple_loss=0.2539, pruned_loss=0.04545, over 17213.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2598, pruned_loss=0.04977, over 3318618.04 frames. ], batch size: 44, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:25:06,544 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:26,570 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-04-29 13:25:27,417 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:28,619 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113234.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:25:54,139 INFO [train.py:904] (0/8) Epoch 12, batch 1600, loss[loss=0.1515, simple_loss=0.236, pruned_loss=0.03353, over 17008.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2615, pruned_loss=0.05028, over 3316225.45 frames. ], batch size: 41, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:26:04,707 INFO [optim.py:368] (0/8) 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 13:26:12,642 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113266.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:22,762 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113273.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 13:26:26,932 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0407, 4.0019, 3.9403, 3.3751, 3.9307, 1.7657, 3.7046, 3.4466], device='cuda:0'), covar=tensor([0.0090, 0.0089, 0.0133, 0.0244, 0.0078, 0.2376, 0.0108, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0121, 0.0170, 0.0159, 0.0141, 0.0184, 0.0158, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:26:27,897 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113277.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:33,951 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113281.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:26:53,672 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 12, batch 1650, loss[loss=0.1693, simple_loss=0.2438, pruned_loss=0.04738, over 16761.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2629, pruned_loss=0.0503, over 3325715.88 frames. ], batch size: 83, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:27:29,761 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:27:52,786 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6112, 2.3095, 2.4985, 4.3743, 2.2175, 2.7747, 2.3762, 2.5841], device='cuda:0'), covar=tensor([0.0924, 0.3215, 0.2126, 0.0398, 0.3508, 0.2107, 0.2927, 0.2985], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0396, 0.0335, 0.0323, 0.0412, 0.0453, 0.0359, 0.0464], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:28:12,390 INFO [train.py:904] (0/8) Epoch 12, batch 1700, loss[loss=0.1759, simple_loss=0.2528, pruned_loss=0.04954, over 16794.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2652, pruned_loss=0.05096, over 3327581.14 frames. ], batch size: 102, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:28:23,616 INFO [optim.py:368] (0/8) 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 13:28:45,448 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.45 vs. limit=5.0 2023-04-29 13:29:22,279 INFO [train.py:904] (0/8) Epoch 12, batch 1750, loss[loss=0.1688, simple_loss=0.2482, pruned_loss=0.04474, over 16803.00 frames. ], tot_loss[loss=0.184, simple_loss=0.266, pruned_loss=0.05099, over 3331773.56 frames. ], batch size: 39, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:30:14,622 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1105, 3.9532, 4.3796, 2.1287, 4.5976, 4.6291, 3.1324, 3.4224], device='cuda:0'), covar=tensor([0.0591, 0.0207, 0.0209, 0.1103, 0.0055, 0.0110, 0.0406, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0142, 0.0071, 0.0108, 0.0123, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 13:30:16,889 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5857, 5.9570, 5.6570, 5.7991, 5.3120, 5.2339, 5.3270, 6.0793], device='cuda:0'), covar=tensor([0.1091, 0.0764, 0.1008, 0.0630, 0.0852, 0.0707, 0.1085, 0.0780], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0715, 0.0583, 0.0500, 0.0448, 0.0461, 0.0599, 0.0548], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:30:18,218 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5375, 3.5640, 3.2241, 3.0337, 3.1908, 3.5024, 3.2613, 3.3447], device='cuda:0'), covar=tensor([0.0552, 0.0509, 0.0309, 0.0279, 0.0579, 0.0405, 0.1538, 0.0443], device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0342, 0.0311, 0.0290, 0.0331, 0.0335, 0.0213, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:30:32,315 INFO [train.py:904] (0/8) Epoch 12, batch 1800, loss[loss=0.2404, simple_loss=0.3126, pruned_loss=0.08409, over 12505.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2673, pruned_loss=0.05067, over 3317152.33 frames. ], batch size: 247, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:30:43,447 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.362e+02 2.904e+02 3.594e+02 5.616e+02, threshold=5.809e+02, percent-clipped=0.0 2023-04-29 13:31:42,916 INFO [train.py:904] (0/8) Epoch 12, batch 1850, loss[loss=0.182, simple_loss=0.271, pruned_loss=0.04648, over 17131.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2686, pruned_loss=0.05142, over 3306230.48 frames. ], batch size: 49, lr: 5.78e-03, grad_scale: 8.0 2023-04-29 13:32:53,590 INFO [train.py:904] (0/8) Epoch 12, batch 1900, loss[loss=0.1768, simple_loss=0.2543, pruned_loss=0.04967, over 16809.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.268, pruned_loss=0.05019, over 3298351.74 frames. ], batch size: 102, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:33:04,797 INFO [optim.py:368] (0/8) 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 13:33:29,834 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:33:48,977 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:33:54,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4313, 3.3177, 3.5148, 1.7907, 3.6941, 3.6363, 2.9242, 2.7487], device='cuda:0'), covar=tensor([0.0682, 0.0192, 0.0177, 0.1109, 0.0066, 0.0165, 0.0398, 0.0401], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0101, 0.0089, 0.0140, 0.0070, 0.0108, 0.0121, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 13:34:00,809 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0178, 5.4234, 5.1510, 5.2019, 4.8729, 4.7917, 4.8538, 5.5187], device='cuda:0'), covar=tensor([0.1026, 0.0841, 0.0847, 0.0713, 0.0728, 0.0905, 0.0986, 0.0802], device='cuda:0'), in_proj_covar=tensor([0.0576, 0.0718, 0.0584, 0.0503, 0.0448, 0.0463, 0.0599, 0.0550], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:34:05,634 INFO [train.py:904] (0/8) Epoch 12, batch 1950, loss[loss=0.1907, simple_loss=0.281, pruned_loss=0.05024, over 17064.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2672, pruned_loss=0.0497, over 3309320.50 frames. ], batch size: 53, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:34:27,436 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113617.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:34:38,487 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=113625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:06,103 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3747, 4.3479, 4.7787, 4.7908, 4.8013, 4.4628, 4.4610, 4.3338], device='cuda:0'), covar=tensor([0.0294, 0.0534, 0.0354, 0.0371, 0.0385, 0.0357, 0.0808, 0.0518], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0361, 0.0364, 0.0339, 0.0404, 0.0383, 0.0485, 0.0310], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 13:35:08,073 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=113646.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:35:16,116 INFO [train.py:904] (0/8) Epoch 12, batch 2000, loss[loss=0.1554, simple_loss=0.2353, pruned_loss=0.03779, over 17033.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.267, pruned_loss=0.04929, over 3313328.53 frames. ], batch size: 41, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:35:27,902 INFO [optim.py:368] (0/8) 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 13:35:52,995 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113678.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:36:25,551 INFO [train.py:904] (0/8) Epoch 12, batch 2050, loss[loss=0.1566, simple_loss=0.2386, pruned_loss=0.03732, over 15888.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2667, pruned_loss=0.04934, over 3318848.07 frames. ], batch size: 35, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:36:32,490 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=113707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:37:33,729 INFO [train.py:904] (0/8) Epoch 12, batch 2100, loss[loss=0.218, simple_loss=0.2883, pruned_loss=0.07384, over 16833.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2681, pruned_loss=0.05092, over 3327443.05 frames. ], batch size: 116, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:37:45,430 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.602e+02 3.022e+02 3.669e+02 5.748e+02, threshold=6.044e+02, percent-clipped=1.0 2023-04-29 13:38:30,005 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1315, 5.1287, 4.9057, 4.2264, 4.9471, 1.6649, 4.6811, 4.8348], device='cuda:0'), covar=tensor([0.0081, 0.0061, 0.0177, 0.0399, 0.0089, 0.2659, 0.0129, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0125, 0.0175, 0.0162, 0.0145, 0.0188, 0.0163, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:38:44,791 INFO [train.py:904] (0/8) Epoch 12, batch 2150, loss[loss=0.1765, simple_loss=0.2558, pruned_loss=0.0486, over 16821.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2689, pruned_loss=0.05176, over 3321935.97 frames. ], batch size: 83, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:39:45,220 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0603, 5.0665, 5.5679, 5.5197, 5.5126, 5.1826, 5.1115, 4.9551], device='cuda:0'), covar=tensor([0.0304, 0.0458, 0.0318, 0.0413, 0.0483, 0.0296, 0.0876, 0.0360], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0361, 0.0366, 0.0340, 0.0406, 0.0383, 0.0487, 0.0311], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 13:39:54,117 INFO [train.py:904] (0/8) Epoch 12, batch 2200, loss[loss=0.2233, simple_loss=0.2956, pruned_loss=0.07551, over 16893.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2688, pruned_loss=0.05214, over 3324646.87 frames. ], batch size: 109, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:40:05,120 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.389e+02 2.779e+02 3.462e+02 6.586e+02, threshold=5.558e+02, percent-clipped=1.0 2023-04-29 13:40:48,907 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=113890.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:41:03,709 INFO [train.py:904] (0/8) Epoch 12, batch 2250, loss[loss=0.1677, simple_loss=0.2529, pruned_loss=0.04128, over 17165.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2688, pruned_loss=0.05185, over 3320914.71 frames. ], batch size: 46, lr: 5.77e-03, grad_scale: 8.0 2023-04-29 13:41:54,647 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 12, batch 2300, loss[loss=0.2006, simple_loss=0.2844, pruned_loss=0.05838, over 16650.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2695, pruned_loss=0.05227, over 3319382.56 frames. ], batch size: 68, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:42:24,217 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.424e+02 2.413e+02 2.819e+02 3.398e+02 7.599e+02, threshold=5.637e+02, percent-clipped=2.0 2023-04-29 13:42:43,137 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=113973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:20,147 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-114000.pt 2023-04-29 13:43:25,577 INFO [train.py:904] (0/8) Epoch 12, batch 2350, loss[loss=0.2181, simple_loss=0.3062, pruned_loss=0.06498, over 16738.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2707, pruned_loss=0.05291, over 3307451.15 frames. ], batch size: 62, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:43:25,843 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:43:26,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6341, 2.6425, 1.8203, 2.7507, 2.1650, 2.7986, 2.1005, 2.3777], device='cuda:0'), covar=tensor([0.0256, 0.0329, 0.1219, 0.0216, 0.0657, 0.0405, 0.1070, 0.0630], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0169, 0.0190, 0.0139, 0.0169, 0.0212, 0.0198, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 13:44:05,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4837, 5.2375, 5.3328, 4.9889, 4.9799, 5.3046, 5.3748, 4.9446], device='cuda:0'), covar=tensor([0.0501, 0.0426, 0.0236, 0.0242, 0.0921, 0.0315, 0.0244, 0.0649], device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0346, 0.0313, 0.0293, 0.0332, 0.0338, 0.0214, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:44:35,989 INFO [train.py:904] (0/8) Epoch 12, batch 2400, loss[loss=0.1688, simple_loss=0.2677, pruned_loss=0.03496, over 17137.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2718, pruned_loss=0.05307, over 3315908.53 frames. ], batch size: 48, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:44:48,487 INFO [optim.py:368] (0/8) 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 13:45:41,900 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2627, 2.0061, 2.5871, 3.1218, 3.0216, 3.5223, 1.9235, 3.5314], device='cuda:0'), covar=tensor([0.0148, 0.0392, 0.0226, 0.0177, 0.0186, 0.0128, 0.0441, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0177, 0.0158, 0.0165, 0.0172, 0.0130, 0.0175, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 13:45:49,053 INFO [train.py:904] (0/8) Epoch 12, batch 2450, loss[loss=0.1863, simple_loss=0.283, pruned_loss=0.04475, over 17280.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2717, pruned_loss=0.05253, over 3313517.27 frames. ], batch size: 52, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:45:53,176 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-29 13:46:13,477 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 13:46:23,342 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114127.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:46:53,831 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2818, 2.0987, 2.3047, 4.0347, 2.0976, 2.4950, 2.2046, 2.2875], device='cuda:0'), covar=tensor([0.1096, 0.3314, 0.2238, 0.0458, 0.3507, 0.2332, 0.3121, 0.2935], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0397, 0.0336, 0.0325, 0.0413, 0.0460, 0.0362, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:46:57,393 INFO [train.py:904] (0/8) Epoch 12, batch 2500, loss[loss=0.2121, simple_loss=0.2886, pruned_loss=0.06777, over 16475.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2711, pruned_loss=0.05219, over 3309609.97 frames. ], batch size: 146, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:47:09,685 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.290e+02 2.668e+02 3.158e+02 5.614e+02, threshold=5.335e+02, percent-clipped=0.0 2023-04-29 13:47:48,666 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114188.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:47:55,501 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9181, 4.0419, 3.0441, 2.2485, 2.9438, 2.4728, 4.5950, 3.8505], device='cuda:0'), covar=tensor([0.2412, 0.0802, 0.1630, 0.2477, 0.2579, 0.1789, 0.0350, 0.1082], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0261, 0.0287, 0.0283, 0.0285, 0.0226, 0.0273, 0.0305], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:48:06,983 INFO [train.py:904] (0/8) Epoch 12, batch 2550, loss[loss=0.165, simple_loss=0.2468, pruned_loss=0.04166, over 16717.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2712, pruned_loss=0.05169, over 3309277.20 frames. ], batch size: 89, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:48:35,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7435, 2.6974, 2.3981, 3.9330, 3.1479, 3.9341, 1.4315, 2.8047], device='cuda:0'), covar=tensor([0.1290, 0.0625, 0.1095, 0.0143, 0.0180, 0.0369, 0.1437, 0.0749], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0161, 0.0182, 0.0152, 0.0198, 0.0213, 0.0182, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 13:48:36,748 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9045, 4.2629, 4.2801, 3.0057, 3.6208, 4.1579, 3.8925, 2.4919], device='cuda:0'), covar=tensor([0.0349, 0.0059, 0.0032, 0.0279, 0.0084, 0.0082, 0.0064, 0.0353], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0071, 0.0070, 0.0124, 0.0081, 0.0089, 0.0079, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:49:15,397 INFO [train.py:904] (0/8) Epoch 12, batch 2600, loss[loss=0.1797, simple_loss=0.2726, pruned_loss=0.04336, over 16183.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2718, pruned_loss=0.05172, over 3306581.92 frames. ], batch size: 35, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:49:25,935 INFO [optim.py:368] (0/8) 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 13:49:45,723 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114273.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:03,215 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7821, 4.7073, 4.6644, 4.4113, 4.3650, 4.7487, 4.6254, 4.4433], device='cuda:0'), covar=tensor([0.0672, 0.0667, 0.0305, 0.0306, 0.0851, 0.0456, 0.0370, 0.0737], device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0348, 0.0317, 0.0293, 0.0334, 0.0340, 0.0215, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:50:16,384 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8669, 2.6921, 2.2993, 2.5176, 3.0232, 2.8784, 3.5626, 3.3011], device='cuda:0'), covar=tensor([0.0084, 0.0333, 0.0371, 0.0361, 0.0212, 0.0276, 0.0213, 0.0203], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0215, 0.0206, 0.0207, 0.0215, 0.0214, 0.0223, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:50:24,323 INFO [train.py:904] (0/8) Epoch 12, batch 2650, loss[loss=0.1829, simple_loss=0.2646, pruned_loss=0.0506, over 16168.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2724, pruned_loss=0.05122, over 3308508.16 frames. ], batch size: 164, lr: 5.76e-03, grad_scale: 16.0 2023-04-29 13:50:24,610 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114302.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:50:45,323 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1402, 1.9760, 2.2201, 3.7551, 1.9984, 2.3742, 2.0685, 2.1403], device='cuda:0'), covar=tensor([0.1115, 0.3483, 0.2365, 0.0518, 0.3671, 0.2407, 0.3347, 0.2942], device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0399, 0.0337, 0.0327, 0.0415, 0.0462, 0.0363, 0.0470], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:50:51,417 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:51:04,934 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-04-29 13:51:12,589 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 13:51:32,416 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 12, batch 2700, loss[loss=0.1742, simple_loss=0.2604, pruned_loss=0.044, over 17033.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2725, pruned_loss=0.05081, over 3316296.31 frames. ], batch size: 53, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:51:45,484 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.425e+02 2.890e+02 3.523e+02 1.000e+03, threshold=5.781e+02, percent-clipped=5.0 2023-04-29 13:52:01,061 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 13:52:41,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6080, 3.6458, 2.8644, 2.1836, 2.4986, 2.1141, 3.8047, 3.3102], device='cuda:0'), covar=tensor([0.2505, 0.0622, 0.1423, 0.2388, 0.2306, 0.1862, 0.0485, 0.1255], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0259, 0.0284, 0.0281, 0.0282, 0.0224, 0.0270, 0.0303], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:52:44,859 INFO [train.py:904] (0/8) Epoch 12, batch 2750, loss[loss=0.1954, simple_loss=0.2726, pruned_loss=0.05912, over 16738.00 frames. ], tot_loss[loss=0.187, simple_loss=0.2722, pruned_loss=0.05087, over 3317364.04 frames. ], batch size: 134, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:53:05,135 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3956, 5.2878, 5.2227, 4.8330, 4.8225, 5.2635, 5.2778, 4.8598], device='cuda:0'), covar=tensor([0.0520, 0.0385, 0.0257, 0.0259, 0.1005, 0.0331, 0.0226, 0.0750], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0349, 0.0318, 0.0294, 0.0336, 0.0341, 0.0216, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:53:54,637 INFO [train.py:904] (0/8) Epoch 12, batch 2800, loss[loss=0.1897, simple_loss=0.2695, pruned_loss=0.05491, over 16732.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2715, pruned_loss=0.05061, over 3310829.26 frames. ], batch size: 134, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:54:06,078 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.386e+02 2.760e+02 3.486e+02 6.287e+02, threshold=5.520e+02, percent-clipped=1.0 2023-04-29 13:54:38,613 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=114483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 13:55:04,049 INFO [train.py:904] (0/8) Epoch 12, batch 2850, loss[loss=0.1751, simple_loss=0.254, pruned_loss=0.0481, over 16750.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2719, pruned_loss=0.05122, over 3301854.00 frames. ], batch size: 83, lr: 5.75e-03, grad_scale: 16.0 2023-04-29 13:55:38,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6586, 2.5967, 2.0788, 2.3789, 2.9397, 2.6771, 3.3800, 3.1763], device='cuda:0'), covar=tensor([0.0097, 0.0309, 0.0389, 0.0361, 0.0212, 0.0290, 0.0201, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0214, 0.0203, 0.0206, 0.0213, 0.0212, 0.0222, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:56:13,213 INFO [train.py:904] (0/8) Epoch 12, batch 2900, loss[loss=0.1422, simple_loss=0.2322, pruned_loss=0.02605, over 17204.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2722, pruned_loss=0.05222, over 3295741.72 frames. ], batch size: 44, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:56:24,540 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.629e+02 2.459e+02 2.961e+02 3.423e+02 5.764e+02, threshold=5.923e+02, percent-clipped=1.0 2023-04-29 13:56:44,500 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2165, 3.3412, 3.4527, 2.1776, 3.0079, 2.4835, 3.7264, 3.6059], device='cuda:0'), covar=tensor([0.0234, 0.0816, 0.0525, 0.1624, 0.0711, 0.0897, 0.0483, 0.0844], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0148, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 13:57:11,087 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 13:57:20,463 INFO [train.py:904] (0/8) Epoch 12, batch 2950, loss[loss=0.1976, simple_loss=0.2734, pruned_loss=0.06097, over 16697.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2715, pruned_loss=0.0526, over 3307307.26 frames. ], batch size: 134, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:23,557 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6763, 4.6824, 4.8560, 4.7068, 4.7300, 5.3214, 4.8779, 4.5590], device='cuda:0'), covar=tensor([0.1320, 0.1893, 0.1817, 0.2139, 0.2668, 0.1030, 0.1410, 0.2446], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0517, 0.0560, 0.0440, 0.0596, 0.0587, 0.0446, 0.0596], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 13:58:25,943 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 13:58:26,804 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8329, 2.4624, 1.8970, 2.3579, 2.8469, 2.6109, 3.0070, 3.0197], device='cuda:0'), covar=tensor([0.0141, 0.0258, 0.0366, 0.0295, 0.0167, 0.0250, 0.0176, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0213, 0.0203, 0.0205, 0.0213, 0.0212, 0.0221, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 13:58:28,578 INFO [train.py:904] (0/8) Epoch 12, batch 3000, loss[loss=0.1932, simple_loss=0.2665, pruned_loss=0.05995, over 16841.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2703, pruned_loss=0.05218, over 3314462.54 frames. ], batch size: 116, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 13:58:28,578 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 13:58:38,476 INFO [train.py:938] (0/8) Epoch 12, validation: loss=0.14, simple_loss=0.2459, pruned_loss=0.01708, over 944034.00 frames. 2023-04-29 13:58:38,477 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 13:58:50,232 INFO [optim.py:368] (0/8) 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 13:59:48,675 INFO [train.py:904] (0/8) Epoch 12, batch 3050, loss[loss=0.216, simple_loss=0.2802, pruned_loss=0.07586, over 16846.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2687, pruned_loss=0.05128, over 3327488.23 frames. ], batch size: 116, lr: 5.75e-03, grad_scale: 8.0 2023-04-29 14:00:08,911 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114716.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:00:56,626 INFO [train.py:904] (0/8) Epoch 12, batch 3100, loss[loss=0.2, simple_loss=0.2667, pruned_loss=0.06663, over 16908.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2676, pruned_loss=0.05056, over 3331977.88 frames. ], batch size: 116, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:01:10,360 INFO [optim.py:368] (0/8) 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 14:01:31,910 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:01:40,449 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=114783.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:02:05,275 INFO [train.py:904] (0/8) Epoch 12, batch 3150, loss[loss=0.1681, simple_loss=0.2501, pruned_loss=0.04308, over 16789.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2666, pruned_loss=0.0507, over 3328925.44 frames. ], batch size: 39, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:02:17,890 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6091, 2.5148, 2.1579, 2.3613, 2.8464, 2.6162, 3.3088, 3.1040], device='cuda:0'), covar=tensor([0.0096, 0.0327, 0.0386, 0.0371, 0.0221, 0.0311, 0.0207, 0.0207], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0211, 0.0203, 0.0205, 0.0213, 0.0213, 0.0221, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:02:30,764 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7048, 4.7753, 5.0259, 4.8064, 4.7931, 5.4471, 5.0362, 4.6716], device='cuda:0'), covar=tensor([0.1177, 0.1841, 0.1876, 0.1973, 0.2681, 0.0985, 0.1413, 0.2315], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0522, 0.0565, 0.0445, 0.0604, 0.0588, 0.0449, 0.0603], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 14:02:45,192 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=114831.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:02:47,972 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 14:03:14,011 INFO [train.py:904] (0/8) Epoch 12, batch 3200, loss[loss=0.1707, simple_loss=0.2558, pruned_loss=0.04278, over 16815.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.266, pruned_loss=0.05034, over 3331959.44 frames. ], batch size: 102, lr: 5.74e-03, grad_scale: 8.0 2023-04-29 14:03:26,055 INFO [optim.py:368] (0/8) 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 14:03:40,765 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=114871.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:03:59,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3599, 4.1797, 4.3893, 4.5801, 4.6885, 4.2952, 4.4848, 4.6487], device='cuda:0'), covar=tensor([0.1424, 0.1120, 0.1451, 0.0627, 0.0569, 0.1095, 0.1617, 0.0618], device='cuda:0'), in_proj_covar=tensor([0.0584, 0.0729, 0.0883, 0.0739, 0.0555, 0.0569, 0.0580, 0.0663], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:04:22,407 INFO [train.py:904] (0/8) Epoch 12, batch 3250, loss[loss=0.1995, simple_loss=0.2769, pruned_loss=0.06106, over 16482.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2673, pruned_loss=0.05092, over 3332087.51 frames. ], batch size: 146, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:05,239 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=114932.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:05:32,352 INFO [train.py:904] (0/8) Epoch 12, batch 3300, loss[loss=0.1744, simple_loss=0.2692, pruned_loss=0.03984, over 16738.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2678, pruned_loss=0.05058, over 3328739.31 frames. ], batch size: 57, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:05:45,364 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.351e+02 3.077e+02 3.881e+02 7.792e+02, threshold=6.153e+02, percent-clipped=7.0 2023-04-29 14:06:42,024 INFO [train.py:904] (0/8) Epoch 12, batch 3350, loss[loss=0.2147, simple_loss=0.2926, pruned_loss=0.06841, over 16323.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2686, pruned_loss=0.05065, over 3330947.49 frames. ], batch size: 165, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:07:13,187 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 14:07:50,797 INFO [train.py:904] (0/8) Epoch 12, batch 3400, loss[loss=0.205, simple_loss=0.288, pruned_loss=0.061, over 16703.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2677, pruned_loss=0.05012, over 3327922.69 frames. ], batch size: 62, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:08:04,045 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.344e+02 2.703e+02 3.237e+02 1.034e+03, threshold=5.406e+02, percent-clipped=1.0 2023-04-29 14:08:12,120 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0373, 3.8562, 4.2926, 2.0262, 4.5112, 4.4840, 3.2908, 3.5152], device='cuda:0'), covar=tensor([0.0652, 0.0232, 0.0254, 0.1111, 0.0077, 0.0135, 0.0383, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0102, 0.0090, 0.0139, 0.0071, 0.0110, 0.0123, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 14:08:18,297 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:08:37,790 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115086.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:09:00,025 INFO [train.py:904] (0/8) Epoch 12, batch 3450, loss[loss=0.1495, simple_loss=0.245, pruned_loss=0.02697, over 17113.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.267, pruned_loss=0.05013, over 3323735.98 frames. ], batch size: 49, lr: 5.74e-03, grad_scale: 4.0 2023-04-29 14:09:15,330 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-04-29 14:09:56,094 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-29 14:10:01,700 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:10:08,329 INFO [train.py:904] (0/8) Epoch 12, batch 3500, loss[loss=0.1997, simple_loss=0.2714, pruned_loss=0.06403, over 16869.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2658, pruned_loss=0.05003, over 3316524.93 frames. ], batch size: 109, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:10:23,311 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.367e+02 2.772e+02 3.300e+02 1.193e+03, threshold=5.543e+02, percent-clipped=2.0 2023-04-29 14:11:19,458 INFO [train.py:904] (0/8) Epoch 12, batch 3550, loss[loss=0.1807, simple_loss=0.2712, pruned_loss=0.04509, over 17076.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2649, pruned_loss=0.04928, over 3314030.93 frames. ], batch size: 53, lr: 5.73e-03, grad_scale: 4.0 2023-04-29 14:11:37,752 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-29 14:11:53,660 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115227.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:12:05,660 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 14:12:28,666 INFO [train.py:904] (0/8) Epoch 12, batch 3600, loss[loss=0.1607, simple_loss=0.2555, pruned_loss=0.03298, over 17186.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2638, pruned_loss=0.0489, over 3320797.46 frames. ], batch size: 46, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:12:43,851 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.756e+02 2.259e+02 2.631e+02 3.384e+02 1.021e+03, threshold=5.262e+02, percent-clipped=2.0 2023-04-29 14:12:57,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2776, 3.3454, 3.4124, 2.3524, 3.2998, 3.5945, 3.3250, 1.6966], device='cuda:0'), covar=tensor([0.0390, 0.0083, 0.0060, 0.0316, 0.0088, 0.0110, 0.0097, 0.0487], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0072, 0.0072, 0.0127, 0.0082, 0.0092, 0.0081, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 14:13:40,333 INFO [train.py:904] (0/8) Epoch 12, batch 3650, loss[loss=0.1772, simple_loss=0.2504, pruned_loss=0.052, over 16445.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2629, pruned_loss=0.0499, over 3327708.44 frames. ], batch size: 146, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:14:12,137 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5102, 2.9260, 3.1044, 2.0301, 2.6587, 2.1572, 3.1188, 3.1342], device='cuda:0'), covar=tensor([0.0250, 0.0754, 0.0481, 0.1726, 0.0787, 0.0908, 0.0544, 0.0821], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0149, 0.0159, 0.0144, 0.0137, 0.0124, 0.0138, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 14:14:28,949 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8553, 5.2192, 5.4755, 5.2634, 5.2313, 5.8361, 5.4150, 5.1282], device='cuda:0'), covar=tensor([0.1221, 0.1867, 0.1553, 0.1856, 0.2337, 0.0912, 0.1239, 0.2318], device='cuda:0'), in_proj_covar=tensor([0.0362, 0.0515, 0.0560, 0.0437, 0.0594, 0.0581, 0.0442, 0.0595], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 14:14:50,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3663, 4.3173, 4.3265, 3.7629, 4.2862, 1.7705, 4.1368, 4.0144], device='cuda:0'), covar=tensor([0.0157, 0.0114, 0.0157, 0.0354, 0.0134, 0.2489, 0.0137, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0128, 0.0177, 0.0167, 0.0148, 0.0188, 0.0166, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:14:55,146 INFO [train.py:904] (0/8) Epoch 12, batch 3700, loss[loss=0.1934, simple_loss=0.2684, pruned_loss=0.05922, over 15553.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2609, pruned_loss=0.0506, over 3293169.95 frames. ], batch size: 190, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:15:09,334 INFO [optim.py:368] (0/8) 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 14:15:21,758 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7142, 2.9425, 2.4972, 4.3025, 3.4471, 4.1105, 1.5712, 2.9061], device='cuda:0'), covar=tensor([0.1475, 0.0647, 0.1143, 0.0163, 0.0211, 0.0392, 0.1547, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0160, 0.0181, 0.0153, 0.0199, 0.0212, 0.0181, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 14:15:22,919 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:24,724 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115371.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:25,863 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115372.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:15:35,416 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4499, 4.4684, 4.8523, 4.8510, 4.8548, 4.5680, 4.5543, 4.3958], device='cuda:0'), covar=tensor([0.0316, 0.0555, 0.0329, 0.0369, 0.0393, 0.0350, 0.0740, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0371, 0.0370, 0.0345, 0.0416, 0.0387, 0.0496, 0.0313], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 14:15:46,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3422, 4.3525, 4.2356, 4.0474, 3.9803, 4.3146, 4.0002, 4.1002], device='cuda:0'), covar=tensor([0.0580, 0.0544, 0.0275, 0.0231, 0.0769, 0.0416, 0.0696, 0.0537], device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0346, 0.0314, 0.0291, 0.0334, 0.0338, 0.0212, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 14:16:10,078 INFO [train.py:904] (0/8) Epoch 12, batch 3750, loss[loss=0.1794, simple_loss=0.2529, pruned_loss=0.05294, over 16843.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2617, pruned_loss=0.0525, over 3289270.85 frames. ], batch size: 102, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:16:36,700 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115420.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:37,996 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6816, 2.7191, 2.4040, 4.0297, 3.2325, 4.0062, 1.5163, 2.8042], device='cuda:0'), covar=tensor([0.1390, 0.0645, 0.1071, 0.0161, 0.0164, 0.0378, 0.1465, 0.0804], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0162, 0.0182, 0.0155, 0.0202, 0.0215, 0.0184, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 14:16:47,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3592, 4.3221, 4.2544, 4.0896, 4.0041, 4.3065, 3.9755, 4.0839], device='cuda:0'), covar=tensor([0.0562, 0.0505, 0.0261, 0.0231, 0.0788, 0.0449, 0.0753, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0345, 0.0314, 0.0291, 0.0334, 0.0338, 0.0212, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 14:16:54,245 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:16:55,346 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:09,438 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115442.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:17:23,457 INFO [train.py:904] (0/8) Epoch 12, batch 3800, loss[loss=0.1922, simple_loss=0.2607, pruned_loss=0.06183, over 16870.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2635, pruned_loss=0.05421, over 3289638.55 frames. ], batch size: 109, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:17:38,968 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 2.466e+02 2.817e+02 3.424e+02 6.065e+02, threshold=5.633e+02, percent-clipped=2.0 2023-04-29 14:17:53,007 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9622, 3.0767, 2.5224, 4.6389, 3.8287, 4.2731, 1.7028, 3.2437], device='cuda:0'), covar=tensor([0.1185, 0.0624, 0.1166, 0.0141, 0.0301, 0.0347, 0.1390, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0155, 0.0202, 0.0215, 0.0184, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 14:18:37,594 INFO [train.py:904] (0/8) Epoch 12, batch 3850, loss[loss=0.1794, simple_loss=0.2467, pruned_loss=0.05608, over 16665.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2634, pruned_loss=0.05473, over 3290720.21 frames. ], batch size: 83, lr: 5.73e-03, grad_scale: 8.0 2023-04-29 14:18:50,947 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-29 14:19:16,982 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:19:52,717 INFO [train.py:904] (0/8) Epoch 12, batch 3900, loss[loss=0.1865, simple_loss=0.2578, pruned_loss=0.05765, over 16904.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2627, pruned_loss=0.05479, over 3298654.30 frames. ], batch size: 116, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:20:07,963 INFO [optim.py:368] (0/8) 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 14:20:29,521 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:21:08,918 INFO [train.py:904] (0/8) Epoch 12, batch 3950, loss[loss=0.1787, simple_loss=0.2513, pruned_loss=0.05303, over 16444.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2625, pruned_loss=0.05502, over 3294857.63 frames. ], batch size: 75, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:21:52,653 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9188, 3.9652, 4.3008, 2.3026, 3.6612, 2.8407, 4.1035, 4.2852], device='cuda:0'), covar=tensor([0.0144, 0.0577, 0.0392, 0.1691, 0.0602, 0.0711, 0.0407, 0.0629], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 14:22:21,470 INFO [train.py:904] (0/8) Epoch 12, batch 4000, loss[loss=0.2171, simple_loss=0.2935, pruned_loss=0.07035, over 12579.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2624, pruned_loss=0.05528, over 3284928.55 frames. ], batch size: 246, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:22:34,737 INFO [optim.py:368] (0/8) 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 14:23:35,785 INFO [train.py:904] (0/8) Epoch 12, batch 4050, loss[loss=0.2002, simple_loss=0.2789, pruned_loss=0.06074, over 17179.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2627, pruned_loss=0.0541, over 3275049.51 frames. ], batch size: 46, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:24:00,354 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 14:24:12,138 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:13,335 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=115727.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:27,819 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4606, 2.2671, 2.2734, 4.2148, 2.1322, 2.6906, 2.3658, 2.4689], device='cuda:0'), covar=tensor([0.0972, 0.3187, 0.2318, 0.0389, 0.3572, 0.2173, 0.3024, 0.3070], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0407, 0.0339, 0.0327, 0.0416, 0.0468, 0.0369, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:24:35,720 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=115742.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:24:50,501 INFO [train.py:904] (0/8) Epoch 12, batch 4100, loss[loss=0.2185, simple_loss=0.3008, pruned_loss=0.06812, over 15376.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2647, pruned_loss=0.05364, over 3270599.95 frames. ], batch size: 190, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:25:05,532 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.445e+02 1.888e+02 2.192e+02 2.609e+02 4.553e+02, threshold=4.384e+02, percent-clipped=0.0 2023-04-29 14:25:28,295 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5796, 2.6002, 2.3518, 3.9949, 3.0200, 3.8500, 1.3989, 2.7848], device='cuda:0'), covar=tensor([0.1345, 0.0724, 0.1274, 0.0150, 0.0324, 0.0410, 0.1596, 0.0891], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0156, 0.0203, 0.0214, 0.0184, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 14:25:48,283 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=115790.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:26:05,817 INFO [train.py:904] (0/8) Epoch 12, batch 4150, loss[loss=0.2116, simple_loss=0.3032, pruned_loss=0.05997, over 16655.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2723, pruned_loss=0.05635, over 3257477.64 frames. ], batch size: 134, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:01,989 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1505, 3.2058, 1.8064, 3.4873, 2.4218, 3.5192, 1.9449, 2.5474], device='cuda:0'), covar=tensor([0.0259, 0.0348, 0.1617, 0.0126, 0.0737, 0.0423, 0.1528, 0.0715], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0166, 0.0188, 0.0136, 0.0168, 0.0210, 0.0195, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 14:27:09,946 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 14:27:23,051 INFO [train.py:904] (0/8) Epoch 12, batch 4200, loss[loss=0.21, simple_loss=0.2999, pruned_loss=0.06008, over 16669.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2803, pruned_loss=0.05873, over 3208494.62 frames. ], batch size: 134, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:27:37,143 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.546e+02 2.889e+02 3.538e+02 7.743e+02, threshold=5.778e+02, percent-clipped=11.0 2023-04-29 14:27:40,776 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4792, 3.5650, 3.2160, 3.0144, 3.1624, 3.4523, 3.2739, 3.2079], device='cuda:0'), covar=tensor([0.0551, 0.0524, 0.0273, 0.0275, 0.0624, 0.0440, 0.1248, 0.0475], device='cuda:0'), in_proj_covar=tensor([0.0252, 0.0328, 0.0300, 0.0277, 0.0320, 0.0322, 0.0202, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:28:03,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8563, 4.7277, 4.8756, 5.0587, 5.2415, 4.5976, 5.1929, 5.2262], device='cuda:0'), covar=tensor([0.1431, 0.1011, 0.1518, 0.0612, 0.0462, 0.0794, 0.0573, 0.0506], device='cuda:0'), in_proj_covar=tensor([0.0539, 0.0673, 0.0804, 0.0683, 0.0514, 0.0532, 0.0533, 0.0617], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:28:22,513 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=115892.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:28:25,135 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6585, 3.9519, 2.9593, 2.2474, 2.6787, 2.3442, 4.1366, 3.4895], device='cuda:0'), covar=tensor([0.2430, 0.0577, 0.1493, 0.2387, 0.2523, 0.1817, 0.0462, 0.0916], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0256, 0.0283, 0.0278, 0.0284, 0.0222, 0.0267, 0.0298], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:28:36,619 INFO [train.py:904] (0/8) Epoch 12, batch 4250, loss[loss=0.1857, simple_loss=0.2795, pruned_loss=0.04597, over 16753.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2831, pruned_loss=0.05864, over 3191711.17 frames. ], batch size: 76, lr: 5.72e-03, grad_scale: 8.0 2023-04-29 14:29:26,214 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 14:29:49,147 INFO [train.py:904] (0/8) Epoch 12, batch 4300, loss[loss=0.1968, simple_loss=0.2921, pruned_loss=0.05074, over 16859.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2833, pruned_loss=0.05685, over 3186737.52 frames. ], batch size: 96, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:29:50,884 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=115953.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 14:30:04,742 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.520e+02 2.885e+02 3.320e+02 6.529e+02, threshold=5.769e+02, percent-clipped=1.0 2023-04-29 14:30:42,726 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-29 14:31:01,711 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-116000.pt 2023-04-29 14:31:07,416 INFO [train.py:904] (0/8) Epoch 12, batch 4350, loss[loss=0.2154, simple_loss=0.304, pruned_loss=0.06344, over 16847.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2869, pruned_loss=0.0581, over 3184188.13 frames. ], batch size: 109, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:31:10,220 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4361, 5.4001, 5.2997, 5.0841, 4.9338, 5.3116, 5.2602, 4.9779], device='cuda:0'), covar=tensor([0.0496, 0.0204, 0.0213, 0.0180, 0.0869, 0.0291, 0.0179, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0326, 0.0299, 0.0276, 0.0317, 0.0318, 0.0202, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:31:27,426 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116014.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:43,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2917, 3.2690, 3.3422, 3.4496, 3.4968, 3.2342, 3.4472, 3.5481], device='cuda:0'), covar=tensor([0.1073, 0.0951, 0.1078, 0.0553, 0.0581, 0.2571, 0.0957, 0.0689], device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0669, 0.0796, 0.0679, 0.0508, 0.0528, 0.0528, 0.0613], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:31:46,036 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116026.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:31:46,136 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8189, 2.6921, 2.5603, 1.8460, 2.5681, 2.7498, 2.5642, 1.8390], device='cuda:0'), covar=tensor([0.0380, 0.0050, 0.0049, 0.0335, 0.0085, 0.0093, 0.0084, 0.0324], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0071, 0.0072, 0.0127, 0.0081, 0.0090, 0.0081, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 14:31:47,318 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:22,084 INFO [train.py:904] (0/8) Epoch 12, batch 4400, loss[loss=0.1935, simple_loss=0.2823, pruned_loss=0.05235, over 17050.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2885, pruned_loss=0.05891, over 3195269.11 frames. ], batch size: 55, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:32:37,554 INFO [optim.py:368] (0/8) 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 14:32:54,764 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116074.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:56,779 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:32:56,923 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116075.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:33:35,427 INFO [train.py:904] (0/8) Epoch 12, batch 4450, loss[loss=0.2265, simple_loss=0.3143, pruned_loss=0.06937, over 16837.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2924, pruned_loss=0.06029, over 3201916.03 frames. ], batch size: 116, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:33:40,171 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116105.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:33:56,381 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116116.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:34:49,333 INFO [train.py:904] (0/8) Epoch 12, batch 4500, loss[loss=0.2004, simple_loss=0.2889, pruned_loss=0.05589, over 16717.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2926, pruned_loss=0.06068, over 3209765.17 frames. ], batch size: 89, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:35:03,474 INFO [optim.py:368] (0/8) 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 14:35:10,209 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116166.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:35:25,600 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116177.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:36:02,118 INFO [train.py:904] (0/8) Epoch 12, batch 4550, loss[loss=0.2349, simple_loss=0.3108, pruned_loss=0.07948, over 16858.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2931, pruned_loss=0.06116, over 3227770.52 frames. ], batch size: 116, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:36:06,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4821, 3.7536, 3.8980, 2.0803, 3.0979, 2.4209, 3.8679, 3.8918], device='cuda:0'), covar=tensor([0.0202, 0.0584, 0.0478, 0.1862, 0.0772, 0.0980, 0.0537, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0149, 0.0158, 0.0144, 0.0136, 0.0124, 0.0137, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 14:37:08,908 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116248.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:37:14,081 INFO [train.py:904] (0/8) Epoch 12, batch 4600, loss[loss=0.1963, simple_loss=0.2871, pruned_loss=0.05272, over 16751.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2937, pruned_loss=0.06132, over 3225118.12 frames. ], batch size: 89, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:37:29,435 INFO [optim.py:368] (0/8) 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 14:38:26,068 INFO [train.py:904] (0/8) Epoch 12, batch 4650, loss[loss=0.2141, simple_loss=0.2973, pruned_loss=0.06547, over 16710.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2927, pruned_loss=0.06121, over 3230358.71 frames. ], batch size: 134, lr: 5.71e-03, grad_scale: 8.0 2023-04-29 14:38:55,784 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116323.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:39:38,315 INFO [train.py:904] (0/8) Epoch 12, batch 4700, loss[loss=0.1957, simple_loss=0.2774, pruned_loss=0.057, over 16566.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2902, pruned_loss=0.06044, over 3213543.43 frames. ], batch size: 62, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:39:53,947 INFO [optim.py:368] (0/8) 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 14:40:05,780 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116370.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:40:27,539 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:40:46,772 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-04-29 14:40:54,108 INFO [train.py:904] (0/8) Epoch 12, batch 4750, loss[loss=0.1882, simple_loss=0.2789, pruned_loss=0.04879, over 16299.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2863, pruned_loss=0.05831, over 3204671.75 frames. ], batch size: 165, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:41:29,791 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-29 14:41:53,546 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-29 14:41:58,942 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116446.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:05,744 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 14:42:07,129 INFO [train.py:904] (0/8) Epoch 12, batch 4800, loss[loss=0.1922, simple_loss=0.2835, pruned_loss=0.05048, over 16627.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2823, pruned_loss=0.05618, over 3203987.99 frames. ], batch size: 134, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:42:22,129 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116461.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:23,022 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.261e+02 1.880e+02 2.223e+02 2.728e+02 4.441e+02, threshold=4.446e+02, percent-clipped=0.0 2023-04-29 14:42:39,549 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:42:48,759 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 14:43:23,406 INFO [train.py:904] (0/8) Epoch 12, batch 4850, loss[loss=0.1972, simple_loss=0.2918, pruned_loss=0.05131, over 16428.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2825, pruned_loss=0.0552, over 3197494.35 frames. ], batch size: 146, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:43:27,706 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116505.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:43:30,768 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116507.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:43:50,308 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6662, 4.6300, 4.5646, 3.8864, 4.5613, 1.7507, 4.3658, 4.3880], device='cuda:0'), covar=tensor([0.0084, 0.0071, 0.0114, 0.0415, 0.0085, 0.2246, 0.0120, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0119, 0.0164, 0.0155, 0.0136, 0.0178, 0.0154, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:44:07,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-29 14:44:31,803 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116548.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:44:38,151 INFO [train.py:904] (0/8) Epoch 12, batch 4900, loss[loss=0.2011, simple_loss=0.2913, pruned_loss=0.05541, over 16858.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2827, pruned_loss=0.05426, over 3189183.94 frames. ], batch size: 116, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:44:52,631 INFO [optim.py:368] (0/8) 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 14:44:58,572 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116566.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:45:42,267 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116596.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 14:45:52,035 INFO [train.py:904] (0/8) Epoch 12, batch 4950, loss[loss=0.2016, simple_loss=0.2904, pruned_loss=0.05641, over 17111.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2821, pruned_loss=0.05358, over 3192632.02 frames. ], batch size: 49, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:46:16,320 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116618.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:46:18,608 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1506, 4.2162, 3.9936, 3.7794, 3.6887, 4.1337, 3.8781, 3.8477], device='cuda:0'), covar=tensor([0.0551, 0.0448, 0.0316, 0.0281, 0.0948, 0.0401, 0.0783, 0.0617], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0320, 0.0291, 0.0270, 0.0311, 0.0311, 0.0198, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:47:04,274 INFO [train.py:904] (0/8) Epoch 12, batch 5000, loss[loss=0.1905, simple_loss=0.2809, pruned_loss=0.0501, over 17138.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2841, pruned_loss=0.05367, over 3198305.12 frames. ], batch size: 49, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:47:17,032 INFO [optim.py:368] (0/8) 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 14:47:30,355 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116670.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 14:47:42,672 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:47:42,810 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116679.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:15,691 INFO [train.py:904] (0/8) Epoch 12, batch 5050, loss[loss=0.1806, simple_loss=0.2672, pruned_loss=0.04704, over 16842.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2845, pruned_loss=0.05373, over 3198917.12 frames. ], batch size: 42, lr: 5.70e-03, grad_scale: 8.0 2023-04-29 14:48:18,342 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:48:38,169 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:24,621 INFO [train.py:904] (0/8) Epoch 12, batch 5100, loss[loss=0.152, simple_loss=0.2473, pruned_loss=0.0283, over 16837.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2827, pruned_loss=0.053, over 3193343.20 frames. ], batch size: 102, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:49:36,778 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116761.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:38,225 INFO [optim.py:368] (0/8) 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 14:49:43,678 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:52,010 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:49:58,658 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8620, 3.9150, 4.2578, 4.2325, 4.2279, 3.9787, 3.9672, 3.9299], device='cuda:0'), covar=tensor([0.0340, 0.0562, 0.0380, 0.0409, 0.0472, 0.0360, 0.0892, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0356, 0.0356, 0.0335, 0.0406, 0.0375, 0.0482, 0.0301], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 14:50:35,777 INFO [train.py:904] (0/8) Epoch 12, batch 5150, loss[loss=0.1837, simple_loss=0.2888, pruned_loss=0.03924, over 16692.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2826, pruned_loss=0.0528, over 3160896.97 frames. ], batch size: 134, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:50:36,195 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116802.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 14:50:36,286 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=116802.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:50:47,784 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116809.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:03,104 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=116820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:51:11,894 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 14:51:15,214 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6066, 2.3414, 2.3387, 4.4381, 2.1716, 2.7077, 2.3957, 2.5956], device='cuda:0'), covar=tensor([0.0951, 0.3086, 0.2308, 0.0322, 0.3555, 0.2069, 0.3088, 0.2435], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0397, 0.0331, 0.0318, 0.0412, 0.0455, 0.0361, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:51:47,923 INFO [train.py:904] (0/8) Epoch 12, batch 5200, loss[loss=0.1912, simple_loss=0.2864, pruned_loss=0.04799, over 16644.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2804, pruned_loss=0.05198, over 3180538.15 frames. ], batch size: 89, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:52:00,676 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116861.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:01,644 INFO [optim.py:368] (0/8) 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 14:52:03,388 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=116863.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:52:58,583 INFO [train.py:904] (0/8) Epoch 12, batch 5250, loss[loss=0.1981, simple_loss=0.2849, pruned_loss=0.05562, over 16767.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2773, pruned_loss=0.05107, over 3203365.64 frames. ], batch size: 124, lr: 5.69e-03, grad_scale: 16.0 2023-04-29 14:54:01,884 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 14:54:11,425 INFO [train.py:904] (0/8) Epoch 12, batch 5300, loss[loss=0.1661, simple_loss=0.2531, pruned_loss=0.03956, over 15315.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2733, pruned_loss=0.04973, over 3209276.82 frames. ], batch size: 190, lr: 5.69e-03, grad_scale: 8.0 2023-04-29 14:54:27,261 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.433e+02 2.208e+02 2.588e+02 3.045e+02 5.450e+02, threshold=5.175e+02, percent-clipped=1.0 2023-04-29 14:54:42,045 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=116974.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:54:48,053 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1223, 5.1194, 4.8876, 4.3090, 4.9526, 1.6717, 4.6930, 4.7769], device='cuda:0'), covar=tensor([0.0069, 0.0063, 0.0123, 0.0374, 0.0082, 0.2490, 0.0107, 0.0161], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0120, 0.0164, 0.0157, 0.0137, 0.0179, 0.0154, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:54:49,813 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=116979.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:55:21,992 INFO [train.py:904] (0/8) Epoch 12, batch 5350, loss[loss=0.2148, simple_loss=0.3005, pruned_loss=0.06457, over 16399.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2719, pruned_loss=0.04912, over 3217434.79 frames. ], batch size: 146, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:55:58,444 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117027.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:31,811 INFO [train.py:904] (0/8) Epoch 12, batch 5400, loss[loss=0.202, simple_loss=0.2885, pruned_loss=0.05774, over 16297.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2748, pruned_loss=0.05021, over 3196017.38 frames. ], batch size: 35, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:56:43,923 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:56:49,070 INFO [optim.py:368] (0/8) 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 14:57:46,209 INFO [train.py:904] (0/8) Epoch 12, batch 5450, loss[loss=0.2116, simple_loss=0.2996, pruned_loss=0.06178, over 16868.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.278, pruned_loss=0.0517, over 3199791.56 frames. ], batch size: 116, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:57:46,717 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:57:54,860 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0661, 5.0118, 4.8612, 4.5997, 4.4424, 4.8901, 4.9226, 4.5836], device='cuda:0'), covar=tensor([0.0607, 0.0696, 0.0326, 0.0275, 0.1159, 0.0531, 0.0294, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0330, 0.0300, 0.0278, 0.0319, 0.0324, 0.0203, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-29 14:58:57,083 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117150.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:00,352 INFO [train.py:904] (0/8) Epoch 12, batch 5500, loss[loss=0.2642, simple_loss=0.3425, pruned_loss=0.09295, over 16395.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.286, pruned_loss=0.05673, over 3183233.17 frames. ], batch size: 35, lr: 5.69e-03, grad_scale: 4.0 2023-04-29 14:59:09,315 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117158.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:13,822 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117161.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 14:59:18,174 INFO [optim.py:368] (0/8) 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 14:59:47,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2910, 3.5108, 3.1298, 2.8517, 2.8670, 3.3470, 3.2182, 3.0513], device='cuda:0'), covar=tensor([0.0794, 0.0568, 0.0408, 0.0332, 0.0979, 0.0478, 0.1406, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0327, 0.0297, 0.0276, 0.0317, 0.0321, 0.0200, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:00:18,016 INFO [train.py:904] (0/8) Epoch 12, batch 5550, loss[loss=0.2924, simple_loss=0.3431, pruned_loss=0.1209, over 11283.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2933, pruned_loss=0.06182, over 3166758.77 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:00:30,336 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117209.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:00:43,466 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 15:00:49,810 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:01:39,155 INFO [train.py:904] (0/8) Epoch 12, batch 5600, loss[loss=0.3122, simple_loss=0.3624, pruned_loss=0.131, over 11372.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2989, pruned_loss=0.06644, over 3145206.01 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 8.0 2023-04-29 15:01:56,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7615, 1.7255, 1.5534, 1.6021, 1.8147, 1.5850, 1.5881, 1.8708], device='cuda:0'), covar=tensor([0.0140, 0.0178, 0.0291, 0.0272, 0.0151, 0.0185, 0.0158, 0.0154], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0205, 0.0200, 0.0199, 0.0205, 0.0203, 0.0210, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:01:58,905 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.404e+02 3.697e+02 4.299e+02 5.006e+02 8.998e+02, threshold=8.599e+02, percent-clipped=4.0 2023-04-29 15:02:16,219 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117274.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:02:30,367 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117282.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:02,081 INFO [train.py:904] (0/8) Epoch 12, batch 5650, loss[loss=0.2496, simple_loss=0.3195, pruned_loss=0.0899, over 16240.00 frames. ], tot_loss[loss=0.2226, simple_loss=0.3039, pruned_loss=0.07064, over 3114481.33 frames. ], batch size: 165, lr: 5.68e-03, grad_scale: 4.0 2023-04-29 15:03:10,866 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2390, 2.0389, 1.6898, 1.7412, 2.2672, 1.9411, 2.0958, 2.3855], device='cuda:0'), covar=tensor([0.0122, 0.0262, 0.0369, 0.0334, 0.0160, 0.0254, 0.0165, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0206, 0.0201, 0.0200, 0.0206, 0.0204, 0.0211, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:03:32,809 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7154, 1.7739, 1.5749, 1.5464, 1.8708, 1.5521, 1.6380, 1.8970], device='cuda:0'), covar=tensor([0.0127, 0.0195, 0.0277, 0.0247, 0.0133, 0.0182, 0.0142, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0206, 0.0201, 0.0200, 0.0205, 0.0203, 0.0210, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:03:33,932 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117322.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:45,040 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=117329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:03:53,241 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5801, 2.6187, 1.8737, 2.7391, 2.1547, 2.7676, 2.0438, 2.4046], device='cuda:0'), covar=tensor([0.0304, 0.0452, 0.1279, 0.0231, 0.0712, 0.0500, 0.1333, 0.0659], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0166, 0.0189, 0.0131, 0.0167, 0.0206, 0.0197, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:04:09,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6773, 4.6757, 4.5195, 3.8699, 4.5800, 1.6848, 4.3925, 4.3915], device='cuda:0'), covar=tensor([0.0075, 0.0066, 0.0119, 0.0316, 0.0073, 0.2428, 0.0096, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0117, 0.0161, 0.0155, 0.0134, 0.0176, 0.0150, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:04:14,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3681, 3.3249, 3.3653, 3.4806, 3.5058, 3.2796, 3.4593, 3.5491], device='cuda:0'), covar=tensor([0.0997, 0.0873, 0.1075, 0.0551, 0.0638, 0.2145, 0.0947, 0.0699], device='cuda:0'), in_proj_covar=tensor([0.0519, 0.0655, 0.0784, 0.0663, 0.0498, 0.0516, 0.0519, 0.0602], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:04:18,939 INFO [train.py:904] (0/8) Epoch 12, batch 5700, loss[loss=0.2193, simple_loss=0.3083, pruned_loss=0.06518, over 16609.00 frames. ], tot_loss[loss=0.2241, simple_loss=0.305, pruned_loss=0.07158, over 3111467.45 frames. ], batch size: 76, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:04:32,712 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117360.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:04:41,589 INFO [optim.py:368] (0/8) 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 15:05:21,412 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=117390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:05:39,185 INFO [train.py:904] (0/8) Epoch 12, batch 5750, loss[loss=0.2489, simple_loss=0.31, pruned_loss=0.09386, over 11052.00 frames. ], tot_loss[loss=0.2274, simple_loss=0.3079, pruned_loss=0.07341, over 3078362.34 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:05:48,657 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117408.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:07:00,140 INFO [train.py:904] (0/8) Epoch 12, batch 5800, loss[loss=0.2204, simple_loss=0.2947, pruned_loss=0.07308, over 11818.00 frames. ], tot_loss[loss=0.2265, simple_loss=0.3078, pruned_loss=0.07266, over 3070361.81 frames. ], batch size: 247, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:07:09,815 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117458.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:07:21,330 INFO [optim.py:368] (0/8) 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 15:08:16,572 INFO [train.py:904] (0/8) Epoch 12, batch 5850, loss[loss=0.2288, simple_loss=0.3053, pruned_loss=0.0761, over 15542.00 frames. ], tot_loss[loss=0.2242, simple_loss=0.3057, pruned_loss=0.07134, over 3068884.20 frames. ], batch size: 190, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:08:24,196 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:08:37,712 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1130, 2.0545, 2.1473, 3.8559, 1.9222, 2.4384, 2.1466, 2.2306], device='cuda:0'), covar=tensor([0.1164, 0.3275, 0.2482, 0.0448, 0.3907, 0.2315, 0.3095, 0.3109], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0395, 0.0331, 0.0316, 0.0412, 0.0453, 0.0359, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:09:37,165 INFO [train.py:904] (0/8) Epoch 12, batch 5900, loss[loss=0.2806, simple_loss=0.339, pruned_loss=0.1111, over 11501.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3053, pruned_loss=0.07134, over 3066883.29 frames. ], batch size: 248, lr: 5.68e-03, grad_scale: 2.0 2023-04-29 15:09:38,510 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 15:10:01,567 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117577.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:10:55,837 INFO [train.py:904] (0/8) Epoch 12, batch 5950, loss[loss=0.1856, simple_loss=0.2804, pruned_loss=0.04545, over 16803.00 frames. ], tot_loss[loss=0.2231, simple_loss=0.3059, pruned_loss=0.07017, over 3058967.18 frames. ], batch size: 83, lr: 5.67e-03, grad_scale: 2.0 2023-04-29 15:11:14,496 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-29 15:12:14,117 INFO [train.py:904] (0/8) Epoch 12, batch 6000, loss[loss=0.2175, simple_loss=0.3005, pruned_loss=0.06722, over 16780.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3058, pruned_loss=0.0704, over 3052181.75 frames. ], batch size: 124, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:12:14,118 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 15:12:25,318 INFO [train.py:938] (0/8) Epoch 12, validation: loss=0.161, simple_loss=0.2739, pruned_loss=0.02405, over 944034.00 frames. 2023-04-29 15:12:25,319 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 15:12:46,506 INFO [optim.py:368] (0/8) 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 15:13:18,118 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=117685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:13:46,113 INFO [train.py:904] (0/8) Epoch 12, batch 6050, loss[loss=0.2112, simple_loss=0.3081, pruned_loss=0.05713, over 16467.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3042, pruned_loss=0.06915, over 3081077.55 frames. ], batch size: 75, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:14:29,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7049, 3.6320, 3.8815, 3.6155, 3.7859, 4.2121, 3.8700, 3.6246], device='cuda:0'), covar=tensor([0.2283, 0.2307, 0.2071, 0.2754, 0.2878, 0.1791, 0.1615, 0.2906], device='cuda:0'), in_proj_covar=tensor([0.0355, 0.0492, 0.0539, 0.0423, 0.0580, 0.0564, 0.0429, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 15:14:54,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8014, 2.5568, 2.3334, 4.1769, 2.6604, 3.9675, 1.5320, 2.7117], device='cuda:0'), covar=tensor([0.1308, 0.0785, 0.1300, 0.0151, 0.0285, 0.0428, 0.1564, 0.0937], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0163, 0.0182, 0.0151, 0.0199, 0.0208, 0.0183, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:15:02,184 INFO [train.py:904] (0/8) Epoch 12, batch 6100, loss[loss=0.1963, simple_loss=0.2852, pruned_loss=0.05374, over 16999.00 frames. ], tot_loss[loss=0.2206, simple_loss=0.3041, pruned_loss=0.06856, over 3081562.85 frames. ], batch size: 55, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:15:13,782 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4885, 3.5647, 3.3334, 3.1025, 3.1704, 3.4503, 3.2658, 3.2283], device='cuda:0'), covar=tensor([0.0579, 0.0493, 0.0271, 0.0266, 0.0559, 0.0371, 0.1245, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0244, 0.0323, 0.0290, 0.0271, 0.0312, 0.0312, 0.0200, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:15:24,824 INFO [optim.py:368] (0/8) 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 15:16:07,621 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2023-04-29 15:16:12,631 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7936, 5.0652, 5.2416, 5.0349, 5.0657, 5.6104, 5.1048, 4.8930], device='cuda:0'), covar=tensor([0.0996, 0.1686, 0.1772, 0.1798, 0.2359, 0.1032, 0.1552, 0.2303], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0490, 0.0540, 0.0422, 0.0579, 0.0564, 0.0429, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 15:16:19,688 INFO [train.py:904] (0/8) Epoch 12, batch 6150, loss[loss=0.1847, simple_loss=0.2734, pruned_loss=0.04805, over 16743.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3024, pruned_loss=0.06833, over 3058494.49 frames. ], batch size: 83, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:17:38,950 INFO [train.py:904] (0/8) Epoch 12, batch 6200, loss[loss=0.2246, simple_loss=0.2923, pruned_loss=0.07849, over 11683.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3002, pruned_loss=0.06765, over 3060921.73 frames. ], batch size: 248, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:18:00,668 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.030e+02 3.055e+02 3.626e+02 4.280e+02 7.203e+02, threshold=7.253e+02, percent-clipped=0.0 2023-04-29 15:18:17,388 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:18:25,713 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8687, 4.1593, 3.9324, 3.9806, 3.6050, 3.7619, 3.7980, 4.1165], device='cuda:0'), covar=tensor([0.1013, 0.0814, 0.0963, 0.0722, 0.0811, 0.1638, 0.0926, 0.1025], device='cuda:0'), in_proj_covar=tensor([0.0554, 0.0689, 0.0568, 0.0486, 0.0438, 0.0449, 0.0571, 0.0532], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:18:52,888 INFO [train.py:904] (0/8) Epoch 12, batch 6250, loss[loss=0.2191, simple_loss=0.3042, pruned_loss=0.06694, over 16734.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3001, pruned_loss=0.06709, over 3082867.57 frames. ], batch size: 57, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:19:27,993 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=117925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:19:46,175 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6319, 6.0516, 5.7177, 5.8021, 5.4104, 5.3329, 5.4168, 6.1375], device='cuda:0'), covar=tensor([0.1080, 0.0757, 0.0913, 0.0689, 0.0780, 0.0634, 0.0957, 0.0860], device='cuda:0'), in_proj_covar=tensor([0.0555, 0.0689, 0.0570, 0.0487, 0.0438, 0.0449, 0.0571, 0.0534], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:20:06,486 INFO [train.py:904] (0/8) Epoch 12, batch 6300, loss[loss=0.2177, simple_loss=0.3054, pruned_loss=0.06506, over 16846.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2996, pruned_loss=0.06614, over 3091380.41 frames. ], batch size: 116, lr: 5.67e-03, grad_scale: 4.0 2023-04-29 15:20:28,835 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.927e+02 3.047e+02 3.596e+02 4.345e+02 1.152e+03, threshold=7.193e+02, percent-clipped=4.0 2023-04-29 15:20:57,513 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=117985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:21:17,289 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.43 vs. limit=5.0 2023-04-29 15:21:19,318 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-118000.pt 2023-04-29 15:21:23,230 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 15:21:25,216 INFO [train.py:904] (0/8) Epoch 12, batch 6350, loss[loss=0.2079, simple_loss=0.2911, pruned_loss=0.06235, over 16594.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3013, pruned_loss=0.06877, over 3058833.42 frames. ], batch size: 62, lr: 5.66e-03, grad_scale: 4.0 2023-04-29 15:22:11,807 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:21,214 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:22:25,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3668, 5.5758, 5.2721, 5.0703, 4.7661, 5.3416, 5.3660, 4.9886], device='cuda:0'), covar=tensor([0.0694, 0.0299, 0.0309, 0.0280, 0.1048, 0.0402, 0.0219, 0.0707], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0324, 0.0289, 0.0269, 0.0308, 0.0311, 0.0200, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:22:34,931 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-29 15:22:37,553 INFO [train.py:904] (0/8) Epoch 12, batch 6400, loss[loss=0.1902, simple_loss=0.2758, pruned_loss=0.05231, over 17030.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.3012, pruned_loss=0.06885, over 3074862.88 frames. ], batch size: 53, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:22:57,665 INFO [optim.py:368] (0/8) 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 15:23:32,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7409, 3.6432, 3.7771, 3.9716, 3.9701, 3.6547, 3.9675, 4.0407], device='cuda:0'), covar=tensor([0.1489, 0.1209, 0.1530, 0.0710, 0.0805, 0.1925, 0.0996, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0527, 0.0657, 0.0789, 0.0669, 0.0503, 0.0518, 0.0529, 0.0605], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:23:49,779 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118101.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:23:50,964 INFO [train.py:904] (0/8) Epoch 12, batch 6450, loss[loss=0.1839, simple_loss=0.2751, pruned_loss=0.04632, over 17266.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.3005, pruned_loss=0.06808, over 3067891.43 frames. ], batch size: 52, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:23:57,578 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3727, 5.3869, 5.1872, 4.0658, 5.2516, 1.7425, 4.9461, 4.9889], device='cuda:0'), covar=tensor([0.0096, 0.0101, 0.0138, 0.0523, 0.0089, 0.2694, 0.0161, 0.0214], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0119, 0.0162, 0.0156, 0.0135, 0.0178, 0.0152, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:25:08,004 INFO [train.py:904] (0/8) Epoch 12, batch 6500, loss[loss=0.2212, simple_loss=0.2995, pruned_loss=0.07147, over 15277.00 frames. ], tot_loss[loss=0.216, simple_loss=0.298, pruned_loss=0.06702, over 3069821.81 frames. ], batch size: 190, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:25:29,347 INFO [optim.py:368] (0/8) 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 15:25:30,404 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-29 15:25:32,627 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118168.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:26:28,550 INFO [train.py:904] (0/8) Epoch 12, batch 6550, loss[loss=0.2025, simple_loss=0.3122, pruned_loss=0.04637, over 16491.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.301, pruned_loss=0.06756, over 3089131.12 frames. ], batch size: 75, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:27:10,838 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118229.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 15:27:44,412 INFO [train.py:904] (0/8) Epoch 12, batch 6600, loss[loss=0.2179, simple_loss=0.2971, pruned_loss=0.0694, over 16992.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3038, pruned_loss=0.06832, over 3098578.75 frames. ], batch size: 55, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:28:05,474 INFO [optim.py:368] (0/8) 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 15:28:10,240 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:28:30,010 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8454, 3.3469, 3.2962, 2.0670, 3.1074, 3.3295, 3.0819, 1.8724], device='cuda:0'), covar=tensor([0.0529, 0.0034, 0.0050, 0.0391, 0.0072, 0.0095, 0.0085, 0.0399], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0069, 0.0072, 0.0126, 0.0081, 0.0092, 0.0081, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 15:28:41,379 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118289.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:00,297 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0920, 4.1486, 4.5507, 4.5263, 4.5461, 4.2344, 4.2668, 4.1453], device='cuda:0'), covar=tensor([0.0302, 0.0579, 0.0328, 0.0384, 0.0423, 0.0370, 0.0885, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0343, 0.0355, 0.0357, 0.0340, 0.0406, 0.0379, 0.0480, 0.0304], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 15:29:01,040 INFO [train.py:904] (0/8) Epoch 12, batch 6650, loss[loss=0.192, simple_loss=0.281, pruned_loss=0.05152, over 16745.00 frames. ], tot_loss[loss=0.2216, simple_loss=0.3041, pruned_loss=0.06957, over 3074896.00 frames. ], batch size: 124, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:29:43,671 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118329.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:29:45,544 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118330.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:16,406 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:30:18,843 INFO [train.py:904] (0/8) Epoch 12, batch 6700, loss[loss=0.2723, simple_loss=0.3272, pruned_loss=0.1087, over 11161.00 frames. ], tot_loss[loss=0.2212, simple_loss=0.3029, pruned_loss=0.06976, over 3079278.17 frames. ], batch size: 248, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:30:39,918 INFO [optim.py:368] (0/8) 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 15:31:17,436 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118390.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:26,562 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:31:35,208 INFO [train.py:904] (0/8) Epoch 12, batch 6750, loss[loss=0.2204, simple_loss=0.2994, pruned_loss=0.07072, over 16259.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3013, pruned_loss=0.06908, over 3088929.25 frames. ], batch size: 35, lr: 5.66e-03, grad_scale: 8.0 2023-04-29 15:32:42,974 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0486, 5.0644, 5.5520, 5.4704, 5.5001, 5.1032, 5.1145, 4.7421], device='cuda:0'), covar=tensor([0.0292, 0.0406, 0.0241, 0.0385, 0.0454, 0.0284, 0.1011, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0347, 0.0359, 0.0359, 0.0343, 0.0412, 0.0383, 0.0486, 0.0307], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 15:32:49,918 INFO [train.py:904] (0/8) Epoch 12, batch 6800, loss[loss=0.2236, simple_loss=0.3117, pruned_loss=0.06781, over 17100.00 frames. ], tot_loss[loss=0.2195, simple_loss=0.3013, pruned_loss=0.06882, over 3087765.36 frames. ], batch size: 49, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:33:10,449 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7460, 1.2853, 1.6146, 1.6238, 1.7328, 1.8606, 1.5260, 1.8026], device='cuda:0'), covar=tensor([0.0196, 0.0295, 0.0148, 0.0214, 0.0193, 0.0146, 0.0280, 0.0090], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0154, 0.0157, 0.0169, 0.0125, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 15:33:11,648 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.917e+02 3.071e+02 3.777e+02 4.759e+02 7.416e+02, threshold=7.554e+02, percent-clipped=1.0 2023-04-29 15:34:04,801 INFO [train.py:904] (0/8) Epoch 12, batch 6850, loss[loss=0.2243, simple_loss=0.3185, pruned_loss=0.06504, over 16352.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3026, pruned_loss=0.06974, over 3070654.83 frames. ], batch size: 165, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:34:35,615 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118524.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 15:35:15,879 INFO [train.py:904] (0/8) Epoch 12, batch 6900, loss[loss=0.2095, simple_loss=0.3052, pruned_loss=0.05693, over 16831.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3048, pruned_loss=0.06893, over 3095519.25 frames. ], batch size: 102, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:35:36,848 INFO [optim.py:368] (0/8) 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 15:36:30,546 INFO [train.py:904] (0/8) Epoch 12, batch 6950, loss[loss=0.2918, simple_loss=0.3445, pruned_loss=0.1195, over 11744.00 frames. ], tot_loss[loss=0.2251, simple_loss=0.3073, pruned_loss=0.07141, over 3066088.67 frames. ], batch size: 248, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:37:03,127 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1797, 2.3867, 1.8506, 2.2123, 2.7681, 2.4203, 2.9449, 3.0153], device='cuda:0'), covar=tensor([0.0101, 0.0349, 0.0461, 0.0368, 0.0195, 0.0290, 0.0174, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0203, 0.0200, 0.0199, 0.0205, 0.0202, 0.0208, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:37:04,563 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118625.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:14,695 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=118631.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:34,934 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118645.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:37:44,799 INFO [train.py:904] (0/8) Epoch 12, batch 7000, loss[loss=0.2228, simple_loss=0.3159, pruned_loss=0.06485, over 16342.00 frames. ], tot_loss[loss=0.224, simple_loss=0.3071, pruned_loss=0.07039, over 3081037.97 frames. ], batch size: 146, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:37:49,951 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2499, 5.5323, 5.2653, 5.2625, 4.9781, 4.9071, 4.9286, 5.6027], device='cuda:0'), covar=tensor([0.0975, 0.0820, 0.0992, 0.0767, 0.0836, 0.0731, 0.1141, 0.0922], device='cuda:0'), in_proj_covar=tensor([0.0557, 0.0695, 0.0574, 0.0490, 0.0442, 0.0455, 0.0577, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:38:05,446 INFO [optim.py:368] (0/8) 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 15:38:33,784 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118685.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:44,108 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=118692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:50,386 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118696.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:38:59,538 INFO [train.py:904] (0/8) Epoch 12, batch 7050, loss[loss=0.2019, simple_loss=0.2939, pruned_loss=0.055, over 16737.00 frames. ], tot_loss[loss=0.2245, simple_loss=0.3079, pruned_loss=0.07052, over 3068621.40 frames. ], batch size: 124, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:39:15,261 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-29 15:39:23,200 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4689, 4.2674, 4.5149, 4.6902, 4.8208, 4.3388, 4.8104, 4.7848], device='cuda:0'), covar=tensor([0.1492, 0.1200, 0.1415, 0.0622, 0.0534, 0.0975, 0.0499, 0.0602], device='cuda:0'), in_proj_covar=tensor([0.0526, 0.0659, 0.0790, 0.0671, 0.0504, 0.0519, 0.0532, 0.0606], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:40:01,768 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118744.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:40:14,611 INFO [train.py:904] (0/8) Epoch 12, batch 7100, loss[loss=0.1986, simple_loss=0.2859, pruned_loss=0.05561, over 16801.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3056, pruned_loss=0.06955, over 3078308.95 frames. ], batch size: 83, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:40:28,475 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 15:40:36,863 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.757e+02 3.120e+02 3.747e+02 4.649e+02 1.761e+03, threshold=7.495e+02, percent-clipped=5.0 2023-04-29 15:40:53,364 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-29 15:41:29,297 INFO [train.py:904] (0/8) Epoch 12, batch 7150, loss[loss=0.225, simple_loss=0.3067, pruned_loss=0.07166, over 16309.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3034, pruned_loss=0.06927, over 3074592.68 frames. ], batch size: 146, lr: 5.65e-03, grad_scale: 8.0 2023-04-29 15:42:01,924 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:42:12,331 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8230, 5.1039, 4.8807, 4.8628, 4.5629, 4.5870, 4.5630, 5.1649], device='cuda:0'), covar=tensor([0.0938, 0.0791, 0.0921, 0.0730, 0.0763, 0.0816, 0.0954, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0550, 0.0684, 0.0566, 0.0483, 0.0435, 0.0450, 0.0568, 0.0530], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:42:41,521 INFO [train.py:904] (0/8) Epoch 12, batch 7200, loss[loss=0.1894, simple_loss=0.2793, pruned_loss=0.04974, over 16879.00 frames. ], tot_loss[loss=0.2177, simple_loss=0.3009, pruned_loss=0.06727, over 3073352.49 frames. ], batch size: 96, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:42:45,003 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-29 15:43:03,913 INFO [optim.py:368] (0/8) 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 15:43:12,504 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:43:39,737 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 15:43:47,075 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4297, 3.0252, 2.9734, 1.8246, 2.6479, 2.1278, 3.0186, 3.1596], device='cuda:0'), covar=tensor([0.0309, 0.0644, 0.0588, 0.1932, 0.0851, 0.0942, 0.0690, 0.0863], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0147, 0.0161, 0.0146, 0.0139, 0.0126, 0.0139, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:44:00,079 INFO [train.py:904] (0/8) Epoch 12, batch 7250, loss[loss=0.1906, simple_loss=0.2775, pruned_loss=0.0519, over 16713.00 frames. ], tot_loss[loss=0.2157, simple_loss=0.299, pruned_loss=0.06619, over 3085691.46 frames. ], batch size: 83, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:44:35,588 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118925.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:44:38,028 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4604, 1.6186, 2.0130, 2.2084, 2.3557, 2.6261, 1.7000, 2.4968], device='cuda:0'), covar=tensor([0.0158, 0.0366, 0.0226, 0.0260, 0.0226, 0.0118, 0.0386, 0.0097], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0154, 0.0157, 0.0170, 0.0124, 0.0171, 0.0117], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 15:45:04,512 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118945.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:14,401 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3387, 3.3370, 3.3716, 3.4725, 3.4749, 3.2429, 3.4799, 3.5253], device='cuda:0'), covar=tensor([0.1030, 0.0814, 0.0990, 0.0514, 0.0597, 0.2318, 0.0925, 0.0733], device='cuda:0'), in_proj_covar=tensor([0.0517, 0.0649, 0.0780, 0.0661, 0.0499, 0.0511, 0.0528, 0.0596], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:45:15,125 INFO [train.py:904] (0/8) Epoch 12, batch 7300, loss[loss=0.2647, simple_loss=0.3227, pruned_loss=0.1033, over 11269.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2985, pruned_loss=0.0666, over 3060765.77 frames. ], batch size: 248, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:45:36,389 INFO [optim.py:368] (0/8) 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 15:45:45,785 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118973.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:45:49,306 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-29 15:46:03,509 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=118985.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:05,806 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=118987.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:15,627 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=118993.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:46:28,526 INFO [train.py:904] (0/8) Epoch 12, batch 7350, loss[loss=0.2055, simple_loss=0.2977, pruned_loss=0.0567, over 16607.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2991, pruned_loss=0.06669, over 3062140.97 frames. ], batch size: 68, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:47:14,410 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119033.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:47:42,942 INFO [train.py:904] (0/8) Epoch 12, batch 7400, loss[loss=0.2763, simple_loss=0.3363, pruned_loss=0.1081, over 11700.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3003, pruned_loss=0.06787, over 3061910.35 frames. ], batch size: 248, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:48:06,307 INFO [optim.py:368] (0/8) 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 15:48:06,776 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3557, 3.3582, 1.7664, 3.8128, 2.4341, 3.6802, 1.8129, 2.5309], device='cuda:0'), covar=tensor([0.0247, 0.0385, 0.1870, 0.0156, 0.0818, 0.0578, 0.1928, 0.0792], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0164, 0.0190, 0.0130, 0.0168, 0.0205, 0.0197, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:48:08,274 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 15:48:19,165 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119075.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:48:47,026 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119093.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:48:47,247 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 15:49:01,220 INFO [train.py:904] (0/8) Epoch 12, batch 7450, loss[loss=0.2088, simple_loss=0.2872, pruned_loss=0.0652, over 16261.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3022, pruned_loss=0.06962, over 3055685.78 frames. ], batch size: 35, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:49:20,222 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 15:49:55,913 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:50:10,456 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-29 15:50:20,478 INFO [train.py:904] (0/8) Epoch 12, batch 7500, loss[loss=0.1966, simple_loss=0.2779, pruned_loss=0.05768, over 16889.00 frames. ], tot_loss[loss=0.2202, simple_loss=0.3022, pruned_loss=0.06908, over 3060520.69 frames. ], batch size: 109, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:50:23,478 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119154.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 15:50:42,267 INFO [optim.py:368] (0/8) 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 15:51:07,576 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1757, 2.9605, 3.0756, 1.9825, 2.8164, 2.0667, 3.0728, 3.1923], device='cuda:0'), covar=tensor([0.0242, 0.0664, 0.0503, 0.1844, 0.0750, 0.0983, 0.0548, 0.0696], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0145, 0.0159, 0.0144, 0.0137, 0.0125, 0.0137, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:51:13,029 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2290, 3.2986, 1.7910, 3.6665, 2.4093, 3.6775, 1.9842, 2.5708], device='cuda:0'), covar=tensor([0.0261, 0.0391, 0.1788, 0.0158, 0.0857, 0.0491, 0.1618, 0.0791], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0163, 0.0189, 0.0130, 0.0167, 0.0203, 0.0196, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:51:35,625 INFO [train.py:904] (0/8) Epoch 12, batch 7550, loss[loss=0.1937, simple_loss=0.2698, pruned_loss=0.05886, over 16678.00 frames. ], tot_loss[loss=0.2186, simple_loss=0.3004, pruned_loss=0.0684, over 3080093.59 frames. ], batch size: 62, lr: 5.64e-03, grad_scale: 4.0 2023-04-29 15:52:50,120 INFO [train.py:904] (0/8) Epoch 12, batch 7600, loss[loss=0.2054, simple_loss=0.2862, pruned_loss=0.06226, over 16630.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2995, pruned_loss=0.06812, over 3099965.93 frames. ], batch size: 62, lr: 5.64e-03, grad_scale: 8.0 2023-04-29 15:53:02,815 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.99 vs. limit=5.0 2023-04-29 15:53:12,407 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.101e+02 2.984e+02 3.563e+02 4.291e+02 1.017e+03, threshold=7.126e+02, percent-clipped=2.0 2023-04-29 15:53:20,565 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5430, 3.6324, 2.9090, 2.1116, 2.5169, 2.3851, 4.0279, 3.3123], device='cuda:0'), covar=tensor([0.2903, 0.0884, 0.1687, 0.2471, 0.2378, 0.1771, 0.0432, 0.1196], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0258, 0.0288, 0.0282, 0.0281, 0.0224, 0.0269, 0.0299], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:53:32,951 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 15:53:43,297 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9117, 4.9172, 4.7016, 4.0831, 4.7552, 1.9361, 4.5256, 4.6005], device='cuda:0'), covar=tensor([0.0077, 0.0061, 0.0139, 0.0326, 0.0076, 0.2213, 0.0104, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0118, 0.0163, 0.0154, 0.0135, 0.0178, 0.0150, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:53:43,301 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119287.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:54:05,877 INFO [train.py:904] (0/8) Epoch 12, batch 7650, loss[loss=0.2714, simple_loss=0.3298, pruned_loss=0.1065, over 11817.00 frames. ], tot_loss[loss=0.218, simple_loss=0.2999, pruned_loss=0.06807, over 3106050.25 frames. ], batch size: 247, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:54:34,988 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0062, 1.9851, 2.1298, 3.5058, 2.0036, 2.3269, 2.1134, 2.1324], device='cuda:0'), covar=tensor([0.1155, 0.3212, 0.2411, 0.0512, 0.3959, 0.2220, 0.3064, 0.3275], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0394, 0.0330, 0.0317, 0.0413, 0.0452, 0.0359, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 15:54:55,612 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:55:20,166 INFO [train.py:904] (0/8) Epoch 12, batch 7700, loss[loss=0.2049, simple_loss=0.299, pruned_loss=0.0554, over 16844.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.2999, pruned_loss=0.06839, over 3102009.89 frames. ], batch size: 96, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:55:42,613 INFO [optim.py:368] (0/8) 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 15:56:19,927 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 15:56:36,121 INFO [train.py:904] (0/8) Epoch 12, batch 7750, loss[loss=0.2206, simple_loss=0.3051, pruned_loss=0.06807, over 16652.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3004, pruned_loss=0.06831, over 3085627.82 frames. ], batch size: 76, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:56:53,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9919, 3.0792, 1.8833, 3.2623, 2.3421, 3.3045, 1.9980, 2.4860], device='cuda:0'), covar=tensor([0.0257, 0.0351, 0.1514, 0.0164, 0.0751, 0.0508, 0.1435, 0.0733], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0164, 0.0190, 0.0130, 0.0168, 0.0204, 0.0196, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:57:09,953 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-04-29 15:57:18,808 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119431.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 15:57:45,523 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119449.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 15:57:48,711 INFO [train.py:904] (0/8) Epoch 12, batch 7800, loss[loss=0.2005, simple_loss=0.2971, pruned_loss=0.05193, over 16945.00 frames. ], tot_loss[loss=0.2201, simple_loss=0.3017, pruned_loss=0.06921, over 3082802.82 frames. ], batch size: 96, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 15:58:11,185 INFO [optim.py:368] (0/8) 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 15:59:01,427 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.66 vs. limit=2.0 2023-04-29 15:59:04,889 INFO [train.py:904] (0/8) Epoch 12, batch 7850, loss[loss=0.209, simple_loss=0.2967, pruned_loss=0.06071, over 16736.00 frames. ], tot_loss[loss=0.221, simple_loss=0.3027, pruned_loss=0.06968, over 3070859.36 frames. ], batch size: 134, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 15:59:19,832 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9810, 3.1038, 1.8715, 3.2808, 2.3728, 3.2957, 1.9815, 2.4693], device='cuda:0'), covar=tensor([0.0261, 0.0347, 0.1559, 0.0178, 0.0763, 0.0583, 0.1516, 0.0810], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0163, 0.0190, 0.0130, 0.0168, 0.0205, 0.0197, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 15:59:23,713 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119514.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:00:21,536 INFO [train.py:904] (0/8) Epoch 12, batch 7900, loss[loss=0.2191, simple_loss=0.2921, pruned_loss=0.07302, over 11771.00 frames. ], tot_loss[loss=0.2199, simple_loss=0.3018, pruned_loss=0.06905, over 3074128.86 frames. ], batch size: 246, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:00:44,371 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2665, 3.4737, 3.5677, 2.4469, 3.3093, 3.5663, 3.4194, 2.0713], device='cuda:0'), covar=tensor([0.0418, 0.0043, 0.0037, 0.0305, 0.0071, 0.0091, 0.0063, 0.0353], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0068, 0.0071, 0.0126, 0.0080, 0.0091, 0.0080, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 16:00:45,732 INFO [optim.py:368] (0/8) 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 16:00:56,291 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119575.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:01:38,585 INFO [train.py:904] (0/8) Epoch 12, batch 7950, loss[loss=0.1959, simple_loss=0.28, pruned_loss=0.05596, over 16840.00 frames. ], tot_loss[loss=0.2203, simple_loss=0.3022, pruned_loss=0.06918, over 3083070.61 frames. ], batch size: 96, lr: 5.63e-03, grad_scale: 4.0 2023-04-29 16:01:40,173 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7169, 6.0750, 5.7300, 5.8256, 5.4327, 5.3496, 5.5508, 6.1591], device='cuda:0'), covar=tensor([0.1045, 0.0707, 0.0915, 0.0653, 0.0716, 0.0541, 0.1025, 0.0752], device='cuda:0'), in_proj_covar=tensor([0.0542, 0.0671, 0.0559, 0.0478, 0.0429, 0.0444, 0.0565, 0.0523], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:02:17,311 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4757, 3.5238, 3.9224, 1.7575, 4.1162, 4.1194, 3.0000, 2.9083], device='cuda:0'), covar=tensor([0.0789, 0.0203, 0.0159, 0.1202, 0.0046, 0.0110, 0.0371, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0101, 0.0088, 0.0138, 0.0069, 0.0105, 0.0121, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 16:02:53,354 INFO [train.py:904] (0/8) Epoch 12, batch 8000, loss[loss=0.2001, simple_loss=0.3004, pruned_loss=0.04993, over 16861.00 frames. ], tot_loss[loss=0.2208, simple_loss=0.3026, pruned_loss=0.06953, over 3086157.71 frames. ], batch size: 42, lr: 5.63e-03, grad_scale: 8.0 2023-04-29 16:03:12,809 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=119665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:03:17,121 INFO [optim.py:368] (0/8) 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 16:03:17,620 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6381, 4.7128, 4.4737, 4.2173, 4.1311, 4.5843, 4.4308, 4.2470], device='cuda:0'), covar=tensor([0.0596, 0.0459, 0.0290, 0.0284, 0.0949, 0.0424, 0.0379, 0.0617], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0324, 0.0285, 0.0264, 0.0303, 0.0307, 0.0197, 0.0332], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:03:57,124 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1809, 1.4163, 1.8383, 2.1367, 2.2463, 2.4378, 1.5964, 2.3518], device='cuda:0'), covar=tensor([0.0159, 0.0404, 0.0233, 0.0261, 0.0260, 0.0145, 0.0374, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0173, 0.0157, 0.0160, 0.0173, 0.0125, 0.0172, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 16:04:07,956 INFO [train.py:904] (0/8) Epoch 12, batch 8050, loss[loss=0.2177, simple_loss=0.3079, pruned_loss=0.06378, over 16794.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3024, pruned_loss=0.06927, over 3093468.98 frames. ], batch size: 102, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:04:42,933 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=119726.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:04:49,758 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119731.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:05:17,345 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=119749.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 16:05:21,196 INFO [train.py:904] (0/8) Epoch 12, batch 8100, loss[loss=0.2263, simple_loss=0.293, pruned_loss=0.07984, over 11795.00 frames. ], tot_loss[loss=0.2196, simple_loss=0.3016, pruned_loss=0.06883, over 3094297.20 frames. ], batch size: 248, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:05:44,429 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 16:05:47,745 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.274e+02 3.055e+02 3.528e+02 4.290e+02 7.598e+02, threshold=7.056e+02, percent-clipped=1.0 2023-04-29 16:06:00,176 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:28,059 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=119797.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:06:34,783 INFO [train.py:904] (0/8) Epoch 12, batch 8150, loss[loss=0.177, simple_loss=0.2654, pruned_loss=0.04427, over 16857.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2985, pruned_loss=0.06732, over 3106891.52 frames. ], batch size: 102, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:07:50,608 INFO [train.py:904] (0/8) Epoch 12, batch 8200, loss[loss=0.1926, simple_loss=0.281, pruned_loss=0.05207, over 16689.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.2968, pruned_loss=0.06748, over 3098125.25 frames. ], batch size: 62, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:08:06,415 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7867, 3.4624, 2.9741, 1.7487, 2.6102, 1.9574, 3.1717, 3.4359], device='cuda:0'), covar=tensor([0.0316, 0.0665, 0.0784, 0.2443, 0.1132, 0.1335, 0.0780, 0.0829], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0145, 0.0160, 0.0144, 0.0137, 0.0125, 0.0137, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 16:08:18,227 INFO [optim.py:368] (0/8) 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 16:08:18,682 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=119870.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:09:09,080 INFO [train.py:904] (0/8) Epoch 12, batch 8250, loss[loss=0.2082, simple_loss=0.3066, pruned_loss=0.05488, over 16375.00 frames. ], tot_loss[loss=0.213, simple_loss=0.296, pruned_loss=0.06499, over 3082163.59 frames. ], batch size: 146, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:09:35,869 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-29 16:09:43,865 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7658, 1.3294, 1.5936, 1.7247, 1.8606, 1.8291, 1.5546, 1.8684], device='cuda:0'), covar=tensor([0.0175, 0.0289, 0.0136, 0.0210, 0.0214, 0.0150, 0.0287, 0.0091], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0155, 0.0158, 0.0170, 0.0124, 0.0169, 0.0116], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 16:10:28,030 INFO [train.py:904] (0/8) Epoch 12, batch 8300, loss[loss=0.1734, simple_loss=0.253, pruned_loss=0.04695, over 11972.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2931, pruned_loss=0.0621, over 3052721.56 frames. ], batch size: 247, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:10:40,657 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9759, 4.1217, 2.5425, 4.7125, 3.0775, 4.6650, 2.6829, 3.2338], device='cuda:0'), covar=tensor([0.0187, 0.0249, 0.1344, 0.0140, 0.0667, 0.0338, 0.1365, 0.0597], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0160, 0.0187, 0.0128, 0.0165, 0.0201, 0.0195, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 16:10:47,538 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-29 16:10:57,565 INFO [optim.py:368] (0/8) 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 16:11:01,524 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.48 vs. limit=5.0 2023-04-29 16:11:40,270 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5295, 3.5410, 2.8501, 1.9690, 2.2382, 2.1684, 3.7526, 3.2980], device='cuda:0'), covar=tensor([0.2697, 0.0589, 0.1447, 0.2740, 0.2684, 0.2132, 0.0372, 0.1017], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0253, 0.0283, 0.0279, 0.0277, 0.0220, 0.0265, 0.0293], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:11:46,702 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-120000.pt 2023-04-29 16:11:52,973 INFO [train.py:904] (0/8) Epoch 12, batch 8350, loss[loss=0.1811, simple_loss=0.2785, pruned_loss=0.04183, over 16474.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2918, pruned_loss=0.05959, over 3055168.15 frames. ], batch size: 75, lr: 5.62e-03, grad_scale: 2.0 2023-04-29 16:12:24,541 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 12, batch 8400, loss[loss=0.1909, simple_loss=0.2824, pruned_loss=0.04969, over 16919.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2887, pruned_loss=0.05718, over 3051771.32 frames. ], batch size: 109, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:13:42,957 INFO [optim.py:368] (0/8) 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 16:14:31,508 INFO [train.py:904] (0/8) Epoch 12, batch 8450, loss[loss=0.1675, simple_loss=0.2505, pruned_loss=0.04226, over 12316.00 frames. ], tot_loss[loss=0.198, simple_loss=0.286, pruned_loss=0.05506, over 3036018.94 frames. ], batch size: 246, lr: 5.62e-03, grad_scale: 4.0 2023-04-29 16:14:58,333 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 16:15:41,980 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0583, 4.0344, 3.9274, 3.3732, 3.9334, 1.8151, 3.7555, 3.5432], device='cuda:0'), covar=tensor([0.0090, 0.0084, 0.0137, 0.0238, 0.0081, 0.2328, 0.0109, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0116, 0.0161, 0.0152, 0.0133, 0.0178, 0.0149, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:15:47,457 INFO [train.py:904] (0/8) Epoch 12, batch 8500, loss[loss=0.1831, simple_loss=0.2754, pruned_loss=0.04543, over 16377.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2819, pruned_loss=0.05277, over 3012275.58 frames. ], batch size: 146, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:16:15,430 INFO [optim.py:368] (0/8) 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 16:16:15,895 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120170.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:17:07,735 INFO [train.py:904] (0/8) Epoch 12, batch 8550, loss[loss=0.195, simple_loss=0.2693, pruned_loss=0.06034, over 12076.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2795, pruned_loss=0.05147, over 2999117.22 frames. ], batch size: 247, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:17:36,932 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=120218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:18:41,893 INFO [train.py:904] (0/8) Epoch 12, batch 8600, loss[loss=0.1733, simple_loss=0.2657, pruned_loss=0.04046, over 16676.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2796, pruned_loss=0.05011, over 3003743.31 frames. ], batch size: 62, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:18:45,448 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.06 vs. limit=2.0 2023-04-29 16:19:19,445 INFO [optim.py:368] (0/8) 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 16:20:03,482 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0307, 4.7043, 4.7607, 3.3451, 3.8348, 4.5096, 4.1046, 3.0767], device='cuda:0'), covar=tensor([0.0340, 0.0018, 0.0016, 0.0247, 0.0070, 0.0052, 0.0042, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0067, 0.0068, 0.0124, 0.0079, 0.0088, 0.0079, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 16:20:19,209 INFO [train.py:904] (0/8) Epoch 12, batch 8650, loss[loss=0.1688, simple_loss=0.2714, pruned_loss=0.03308, over 16358.00 frames. ], tot_loss[loss=0.1869, simple_loss=0.2773, pruned_loss=0.04824, over 3004968.83 frames. ], batch size: 165, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:21:01,265 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=120321.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:22:02,146 INFO [train.py:904] (0/8) Epoch 12, batch 8700, loss[loss=0.1686, simple_loss=0.2565, pruned_loss=0.0403, over 12540.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2746, pruned_loss=0.04688, over 3003129.11 frames. ], batch size: 248, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:22:33,097 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 2.303e+02 2.810e+02 3.260e+02 5.191e+02, threshold=5.620e+02, percent-clipped=0.0 2023-04-29 16:23:36,540 INFO [train.py:904] (0/8) Epoch 12, batch 8750, loss[loss=0.18, simple_loss=0.2827, pruned_loss=0.0387, over 16523.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2747, pruned_loss=0.04635, over 3019559.33 frames. ], batch size: 68, lr: 5.61e-03, grad_scale: 4.0 2023-04-29 16:23:38,030 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 16:24:48,668 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 16:25:30,730 INFO [train.py:904] (0/8) Epoch 12, batch 8800, loss[loss=0.1876, simple_loss=0.2796, pruned_loss=0.0478, over 16822.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2731, pruned_loss=0.04529, over 3038873.66 frames. ], batch size: 124, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:25:41,808 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 16:26:07,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7213, 2.0429, 1.7418, 1.8058, 2.4133, 1.9922, 2.3699, 2.5750], device='cuda:0'), covar=tensor([0.0103, 0.0326, 0.0412, 0.0364, 0.0230, 0.0308, 0.0153, 0.0206], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0202, 0.0197, 0.0197, 0.0203, 0.0200, 0.0201, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:26:08,433 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.720e+02 2.538e+02 3.135e+02 3.629e+02 6.620e+02, threshold=6.271e+02, percent-clipped=2.0 2023-04-29 16:26:30,861 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 16:27:17,126 INFO [train.py:904] (0/8) Epoch 12, batch 8850, loss[loss=0.1923, simple_loss=0.2928, pruned_loss=0.04586, over 16365.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2757, pruned_loss=0.04497, over 3028685.40 frames. ], batch size: 146, lr: 5.61e-03, grad_scale: 8.0 2023-04-29 16:28:33,060 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 16:29:04,475 INFO [train.py:904] (0/8) Epoch 12, batch 8900, loss[loss=0.1746, simple_loss=0.2749, pruned_loss=0.03714, over 16597.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2763, pruned_loss=0.04478, over 3035052.11 frames. ], batch size: 89, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:29:39,306 INFO [optim.py:368] (0/8) 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 16:31:10,998 INFO [train.py:904] (0/8) Epoch 12, batch 8950, loss[loss=0.1629, simple_loss=0.2562, pruned_loss=0.03484, over 16705.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2758, pruned_loss=0.04507, over 3033847.14 frames. ], batch size: 134, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:31:50,914 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 16:33:00,373 INFO [train.py:904] (0/8) Epoch 12, batch 9000, loss[loss=0.1712, simple_loss=0.2614, pruned_loss=0.04049, over 12108.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2726, pruned_loss=0.04357, over 3050834.80 frames. ], batch size: 250, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:33:00,375 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 16:33:10,348 INFO [train.py:938] (0/8) Epoch 12, validation: loss=0.1532, simple_loss=0.2571, pruned_loss=0.02465, over 944034.00 frames. 2023-04-29 16:33:10,348 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 16:33:46,688 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 16:33:49,107 INFO [optim.py:368] (0/8) 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 16:34:01,184 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=120676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:34:14,614 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7917, 1.2951, 1.6262, 1.7227, 1.8475, 1.8474, 1.4841, 1.8640], device='cuda:0'), covar=tensor([0.0207, 0.0325, 0.0183, 0.0256, 0.0240, 0.0174, 0.0346, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0156, 0.0156, 0.0168, 0.0122, 0.0170, 0.0115], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 16:34:54,272 INFO [train.py:904] (0/8) Epoch 12, batch 9050, loss[loss=0.1973, simple_loss=0.2744, pruned_loss=0.06007, over 12423.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2733, pruned_loss=0.04403, over 3054491.76 frames. ], batch size: 250, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:35:30,349 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5734, 1.6568, 1.9747, 2.5888, 2.4625, 2.7976, 1.8027, 2.7500], device='cuda:0'), covar=tensor([0.0153, 0.0405, 0.0311, 0.0247, 0.0234, 0.0132, 0.0386, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0170, 0.0156, 0.0156, 0.0168, 0.0121, 0.0170, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 16:36:04,381 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=120737.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:36:39,415 INFO [train.py:904] (0/8) Epoch 12, batch 9100, loss[loss=0.1731, simple_loss=0.2608, pruned_loss=0.04271, over 12295.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.273, pruned_loss=0.04457, over 3071697.08 frames. ], batch size: 247, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:37:15,448 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.896e+02 2.513e+02 2.910e+02 3.503e+02 5.095e+02, threshold=5.819e+02, percent-clipped=0.0 2023-04-29 16:38:36,942 INFO [train.py:904] (0/8) Epoch 12, batch 9150, loss[loss=0.1817, simple_loss=0.2722, pruned_loss=0.04561, over 16948.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2735, pruned_loss=0.04425, over 3070470.33 frames. ], batch size: 109, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:39:33,701 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.07 vs. limit=5.0 2023-04-29 16:39:52,037 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5813, 3.5702, 3.5402, 3.0618, 3.4487, 1.9754, 3.3122, 3.0240], device='cuda:0'), covar=tensor([0.0129, 0.0114, 0.0163, 0.0194, 0.0103, 0.2043, 0.0110, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0116, 0.0159, 0.0147, 0.0132, 0.0178, 0.0147, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:40:21,524 INFO [train.py:904] (0/8) Epoch 12, batch 9200, loss[loss=0.157, simple_loss=0.2388, pruned_loss=0.03763, over 12120.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2683, pruned_loss=0.04279, over 3060232.82 frames. ], batch size: 248, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:40:22,269 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8148, 2.0863, 1.7963, 1.9835, 2.5753, 2.2176, 2.5904, 2.7231], device='cuda:0'), covar=tensor([0.0112, 0.0344, 0.0411, 0.0372, 0.0212, 0.0297, 0.0150, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0204, 0.0199, 0.0198, 0.0204, 0.0201, 0.0201, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:40:27,916 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4706, 2.9970, 3.0772, 1.8660, 2.6732, 2.1630, 2.9932, 3.0772], device='cuda:0'), covar=tensor([0.0316, 0.0813, 0.0514, 0.1901, 0.0818, 0.0942, 0.0717, 0.0934], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0139, 0.0155, 0.0141, 0.0134, 0.0123, 0.0133, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 16:40:40,286 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4649, 4.4007, 4.8945, 4.8813, 4.8926, 4.5944, 4.5401, 4.4623], device='cuda:0'), covar=tensor([0.0340, 0.0680, 0.0372, 0.0400, 0.0420, 0.0381, 0.0936, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0336, 0.0337, 0.0319, 0.0382, 0.0361, 0.0447, 0.0288], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 16:40:55,542 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.400e+02 2.881e+02 3.588e+02 6.775e+02, threshold=5.761e+02, percent-clipped=4.0 2023-04-29 16:42:00,504 INFO [train.py:904] (0/8) Epoch 12, batch 9250, loss[loss=0.1478, simple_loss=0.2294, pruned_loss=0.03317, over 11964.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2684, pruned_loss=0.04285, over 3064445.49 frames. ], batch size: 248, lr: 5.60e-03, grad_scale: 8.0 2023-04-29 16:42:35,450 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-29 16:42:47,517 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-29 16:43:49,111 INFO [train.py:904] (0/8) Epoch 12, batch 9300, loss[loss=0.1639, simple_loss=0.242, pruned_loss=0.04286, over 12560.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2673, pruned_loss=0.04251, over 3053225.21 frames. ], batch size: 248, lr: 5.60e-03, grad_scale: 4.0 2023-04-29 16:44:08,430 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3702, 4.6549, 4.4679, 4.4525, 4.1460, 4.0907, 4.1218, 4.6814], device='cuda:0'), covar=tensor([0.0968, 0.0922, 0.0929, 0.0665, 0.0754, 0.1449, 0.1014, 0.0896], device='cuda:0'), in_proj_covar=tensor([0.0524, 0.0654, 0.0537, 0.0462, 0.0417, 0.0434, 0.0544, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:44:31,800 INFO [optim.py:368] (0/8) 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 16:45:32,159 INFO [train.py:904] (0/8) Epoch 12, batch 9350, loss[loss=0.1809, simple_loss=0.2655, pruned_loss=0.04809, over 12415.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2679, pruned_loss=0.04276, over 3055761.24 frames. ], batch size: 247, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:45:35,066 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121003.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:45:50,763 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3450, 3.4239, 3.6508, 3.6436, 3.6588, 3.4539, 3.5054, 3.5244], device='cuda:0'), covar=tensor([0.0335, 0.0630, 0.0526, 0.0506, 0.0480, 0.0447, 0.0790, 0.0461], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0332, 0.0333, 0.0318, 0.0380, 0.0356, 0.0443, 0.0287], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 16:46:33,104 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:46:55,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3871, 4.6577, 4.3904, 4.1332, 3.8189, 4.5490, 4.4162, 4.2269], device='cuda:0'), covar=tensor([0.0905, 0.0553, 0.0419, 0.0329, 0.1229, 0.0559, 0.0470, 0.0671], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0312, 0.0276, 0.0256, 0.0291, 0.0298, 0.0191, 0.0322], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:47:00,520 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5617, 3.6571, 3.3902, 3.1542, 3.2412, 3.5496, 3.3061, 3.3364], device='cuda:0'), covar=tensor([0.0588, 0.0419, 0.0253, 0.0225, 0.0512, 0.0441, 0.1163, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0312, 0.0276, 0.0256, 0.0291, 0.0298, 0.0191, 0.0322], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:47:13,260 INFO [train.py:904] (0/8) Epoch 12, batch 9400, loss[loss=0.1394, simple_loss=0.227, pruned_loss=0.02587, over 12293.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2678, pruned_loss=0.04245, over 3046204.31 frames. ], batch size: 247, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:47:14,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4851, 5.4717, 5.3393, 4.9479, 4.9494, 5.3749, 5.3154, 5.1201], device='cuda:0'), covar=tensor([0.0599, 0.0547, 0.0243, 0.0257, 0.0936, 0.0453, 0.0208, 0.0642], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0312, 0.0277, 0.0256, 0.0291, 0.0298, 0.0191, 0.0321], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:47:39,144 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.435e+02 2.897e+02 3.662e+02 7.614e+02, threshold=5.793e+02, percent-clipped=2.0 2023-04-29 16:48:34,560 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 16:48:55,047 INFO [train.py:904] (0/8) Epoch 12, batch 9450, loss[loss=0.1801, simple_loss=0.2651, pruned_loss=0.04753, over 12632.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2692, pruned_loss=0.04261, over 3036976.69 frames. ], batch size: 247, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:49:12,922 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-29 16:50:34,755 INFO [train.py:904] (0/8) Epoch 12, batch 9500, loss[loss=0.2105, simple_loss=0.302, pruned_loss=0.05949, over 17002.00 frames. ], tot_loss[loss=0.177, simple_loss=0.269, pruned_loss=0.04247, over 3040819.58 frames. ], batch size: 109, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:51:01,768 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121164.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:51:13,621 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.296e+02 2.757e+02 3.587e+02 6.751e+02, threshold=5.515e+02, percent-clipped=2.0 2023-04-29 16:51:15,324 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0154, 2.7730, 2.8650, 2.0056, 2.6877, 2.2068, 2.6760, 2.8761], device='cuda:0'), covar=tensor([0.0310, 0.0757, 0.0466, 0.1660, 0.0715, 0.0894, 0.0647, 0.0777], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0138, 0.0155, 0.0140, 0.0133, 0.0122, 0.0132, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 16:51:32,026 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-04-29 16:51:38,473 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-29 16:52:20,008 INFO [train.py:904] (0/8) Epoch 12, batch 9550, loss[loss=0.2015, simple_loss=0.2915, pruned_loss=0.05574, over 16900.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2683, pruned_loss=0.04221, over 3052491.07 frames. ], batch size: 116, lr: 5.59e-03, grad_scale: 4.0 2023-04-29 16:53:08,748 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121225.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:53:32,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4282, 2.9337, 2.6861, 2.1724, 2.1985, 2.2033, 2.9038, 2.9298], device='cuda:0'), covar=tensor([0.2249, 0.0645, 0.1344, 0.2263, 0.2041, 0.1729, 0.0410, 0.1064], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0247, 0.0277, 0.0271, 0.0258, 0.0215, 0.0258, 0.0283], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 16:53:33,701 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2539, 3.3011, 3.4975, 1.5303, 3.5913, 3.7091, 2.9551, 2.7982], device='cuda:0'), covar=tensor([0.0853, 0.0193, 0.0145, 0.1348, 0.0080, 0.0116, 0.0355, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0097, 0.0084, 0.0135, 0.0067, 0.0101, 0.0118, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 16:53:59,935 INFO [train.py:904] (0/8) Epoch 12, batch 9600, loss[loss=0.1877, simple_loss=0.2667, pruned_loss=0.0543, over 12663.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.27, pruned_loss=0.04312, over 3042709.19 frames. ], batch size: 250, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:54:12,314 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 16:54:35,190 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.441e+02 3.037e+02 3.482e+02 7.490e+02, threshold=6.075e+02, percent-clipped=2.0 2023-04-29 16:55:45,652 INFO [train.py:904] (0/8) Epoch 12, batch 9650, loss[loss=0.1524, simple_loss=0.2523, pruned_loss=0.02628, over 16549.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2717, pruned_loss=0.04329, over 3047357.79 frames. ], batch size: 68, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:55:56,692 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7322, 3.3595, 3.3467, 1.9211, 2.9939, 2.3190, 3.2831, 3.3534], device='cuda:0'), covar=tensor([0.0252, 0.0685, 0.0476, 0.1874, 0.0687, 0.0864, 0.0708, 0.0896], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0137, 0.0154, 0.0140, 0.0133, 0.0122, 0.0132, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 16:56:09,926 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121311.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 16:56:52,036 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121332.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:57:30,754 INFO [train.py:904] (0/8) Epoch 12, batch 9700, loss[loss=0.1761, simple_loss=0.2621, pruned_loss=0.04509, over 12367.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2714, pruned_loss=0.04343, over 3063571.97 frames. ], batch size: 248, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:57:44,436 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121359.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:57:50,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6526, 2.5753, 2.2289, 3.5489, 1.9411, 3.6067, 1.5360, 2.7580], device='cuda:0'), covar=tensor([0.1516, 0.0717, 0.1256, 0.0149, 0.0109, 0.0374, 0.1743, 0.0787], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0157, 0.0179, 0.0143, 0.0185, 0.0202, 0.0181, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 16:58:05,442 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1509, 2.1026, 2.5031, 3.2566, 2.9815, 3.4491, 2.1003, 3.4464], device='cuda:0'), covar=tensor([0.0129, 0.0334, 0.0242, 0.0150, 0.0177, 0.0110, 0.0340, 0.0086], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0168, 0.0154, 0.0156, 0.0165, 0.0120, 0.0168, 0.0113], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 16:58:07,292 INFO [optim.py:368] (0/8) 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 16:58:10,419 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121372.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 16:58:31,092 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 16:59:14,345 INFO [train.py:904] (0/8) Epoch 12, batch 9750, loss[loss=0.1984, simple_loss=0.2736, pruned_loss=0.06156, over 12725.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2703, pruned_loss=0.04363, over 3059114.90 frames. ], batch size: 248, lr: 5.59e-03, grad_scale: 8.0 2023-04-29 16:59:23,475 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121407.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:00:53,753 INFO [train.py:904] (0/8) Epoch 12, batch 9800, loss[loss=0.1824, simple_loss=0.2863, pruned_loss=0.03929, over 16236.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2705, pruned_loss=0.04252, over 3073746.20 frames. ], batch size: 165, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:01:18,659 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0492, 3.6462, 3.3646, 1.8420, 2.7374, 2.1923, 3.5648, 3.5953], device='cuda:0'), covar=tensor([0.0248, 0.0589, 0.0621, 0.2112, 0.0998, 0.1094, 0.0670, 0.0861], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0137, 0.0155, 0.0140, 0.0133, 0.0122, 0.0131, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 17:01:24,644 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=121468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:01:29,886 INFO [optim.py:368] (0/8) 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 17:02:38,492 INFO [train.py:904] (0/8) Epoch 12, batch 9850, loss[loss=0.1715, simple_loss=0.2545, pruned_loss=0.04424, over 12408.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2717, pruned_loss=0.04256, over 3046566.51 frames. ], batch size: 250, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:03:17,194 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:03:17,389 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3995, 1.9922, 1.7141, 1.7250, 2.2268, 1.9273, 2.1205, 2.3531], device='cuda:0'), covar=tensor([0.0092, 0.0300, 0.0385, 0.0367, 0.0205, 0.0283, 0.0148, 0.0189], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0203, 0.0197, 0.0197, 0.0203, 0.0201, 0.0198, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:04:03,259 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7254, 1.7454, 2.2539, 2.7674, 2.5864, 2.9530, 2.1058, 2.9221], device='cuda:0'), covar=tensor([0.0164, 0.0410, 0.0249, 0.0201, 0.0226, 0.0124, 0.0331, 0.0118], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0170, 0.0155, 0.0157, 0.0167, 0.0122, 0.0170, 0.0114], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0001], device='cuda:0') 2023-04-29 17:04:28,366 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 17:04:30,204 INFO [train.py:904] (0/8) Epoch 12, batch 9900, loss[loss=0.193, simple_loss=0.2902, pruned_loss=0.04793, over 16385.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2721, pruned_loss=0.04231, over 3051631.10 frames. ], batch size: 146, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:04:50,382 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 17:05:12,986 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.385e+02 2.886e+02 3.421e+02 6.283e+02, threshold=5.771e+02, percent-clipped=3.0 2023-04-29 17:06:26,654 INFO [train.py:904] (0/8) Epoch 12, batch 9950, loss[loss=0.1653, simple_loss=0.2632, pruned_loss=0.03365, over 16784.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2739, pruned_loss=0.04282, over 3043648.01 frames. ], batch size: 83, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:07:03,243 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2921, 5.1824, 5.3340, 5.5513, 5.6994, 4.9222, 5.6629, 5.6783], device='cuda:0'), covar=tensor([0.1457, 0.0988, 0.1403, 0.0529, 0.0452, 0.0638, 0.0403, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0498, 0.0625, 0.0749, 0.0638, 0.0482, 0.0492, 0.0505, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:08:27,933 INFO [train.py:904] (0/8) Epoch 12, batch 10000, loss[loss=0.1756, simple_loss=0.276, pruned_loss=0.03762, over 16882.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2718, pruned_loss=0.04188, over 3073262.44 frames. ], batch size: 102, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:08:28,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3413, 3.8170, 3.9552, 2.0715, 3.2316, 2.5839, 3.7707, 3.8401], device='cuda:0'), covar=tensor([0.0200, 0.0575, 0.0440, 0.1772, 0.0626, 0.0788, 0.0619, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0137, 0.0155, 0.0141, 0.0134, 0.0122, 0.0131, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 17:08:44,241 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121659.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:08:57,388 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0853, 3.9214, 4.1061, 4.2880, 4.4035, 4.0454, 4.3874, 4.4002], device='cuda:0'), covar=tensor([0.1440, 0.1149, 0.1567, 0.0735, 0.0638, 0.1038, 0.0720, 0.0688], device='cuda:0'), in_proj_covar=tensor([0.0497, 0.0625, 0.0748, 0.0638, 0.0480, 0.0491, 0.0504, 0.0578], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:08:59,975 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121667.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:09:06,355 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.215e+02 2.322e+02 2.719e+02 3.378e+02 7.215e+02, threshold=5.438e+02, percent-clipped=2.0 2023-04-29 17:09:10,297 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9047, 2.3139, 1.9582, 2.0674, 2.6236, 2.3767, 2.6987, 2.8071], device='cuda:0'), covar=tensor([0.0103, 0.0350, 0.0439, 0.0399, 0.0229, 0.0315, 0.0207, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0206, 0.0200, 0.0200, 0.0206, 0.0203, 0.0200, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:10:02,933 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 17:10:10,842 INFO [train.py:904] (0/8) Epoch 12, batch 10050, loss[loss=0.1703, simple_loss=0.2708, pruned_loss=0.03496, over 16895.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2717, pruned_loss=0.04152, over 3078449.76 frames. ], batch size: 96, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:10:21,426 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121707.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:10:33,268 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 17:10:57,712 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 17:11:13,724 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 17:11:21,663 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.02 vs. limit=5.0 2023-04-29 17:11:22,916 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 17:11:46,089 INFO [train.py:904] (0/8) Epoch 12, batch 10100, loss[loss=0.1742, simple_loss=0.2539, pruned_loss=0.04729, over 12704.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2718, pruned_loss=0.04183, over 3070860.38 frames. ], batch size: 247, lr: 5.58e-03, grad_scale: 8.0 2023-04-29 17:12:05,576 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=121763.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:12:23,726 INFO [optim.py:368] (0/8) 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 17:13:05,195 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-12.pt 2023-04-29 17:13:30,561 INFO [train.py:904] (0/8) Epoch 13, batch 0, loss[loss=0.2135, simple_loss=0.3004, pruned_loss=0.06328, over 16738.00 frames. ], tot_loss[loss=0.2135, simple_loss=0.3004, pruned_loss=0.06328, over 16738.00 frames. ], batch size: 57, lr: 5.36e-03, grad_scale: 8.0 2023-04-29 17:13:30,562 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 17:13:38,110 INFO [train.py:938] (0/8) Epoch 13, validation: loss=0.1523, simple_loss=0.2559, pruned_loss=0.02431, over 944034.00 frames. 2023-04-29 17:13:38,110 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 17:14:04,094 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121820.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:14:23,123 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 17:14:49,509 INFO [train.py:904] (0/8) Epoch 13, batch 50, loss[loss=0.2164, simple_loss=0.2863, pruned_loss=0.07323, over 16793.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2855, pruned_loss=0.05991, over 750187.43 frames. ], batch size: 124, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:15:11,433 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=121868.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:15:18,837 INFO [optim.py:368] (0/8) 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 17:15:53,879 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8191, 2.6903, 2.5485, 2.0029, 2.6077, 2.7521, 2.6427, 1.8247], device='cuda:0'), covar=tensor([0.0373, 0.0062, 0.0043, 0.0302, 0.0092, 0.0070, 0.0067, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0070, 0.0070, 0.0127, 0.0080, 0.0088, 0.0079, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 17:15:58,278 INFO [train.py:904] (0/8) Epoch 13, batch 100, loss[loss=0.1797, simple_loss=0.2775, pruned_loss=0.04101, over 17111.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2821, pruned_loss=0.05832, over 1313612.56 frames. ], batch size: 48, lr: 5.36e-03, grad_scale: 2.0 2023-04-29 17:16:03,237 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2023-04-29 17:17:04,456 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=121950.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:17:07,083 INFO [train.py:904] (0/8) Epoch 13, batch 150, loss[loss=0.1518, simple_loss=0.2372, pruned_loss=0.03316, over 16795.00 frames. ], tot_loss[loss=0.1954, simple_loss=0.2782, pruned_loss=0.05634, over 1762483.20 frames. ], batch size: 39, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:17:27,853 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=121967.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:17:35,633 INFO [optim.py:368] (0/8) 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 17:18:12,590 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-122000.pt 2023-04-29 17:18:18,195 INFO [train.py:904] (0/8) Epoch 13, batch 200, loss[loss=0.2051, simple_loss=0.2715, pruned_loss=0.06935, over 16735.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2766, pruned_loss=0.05561, over 2104825.40 frames. ], batch size: 124, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:18:30,556 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122011.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:18:35,255 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122015.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:19:25,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8507, 4.0061, 4.2759, 3.0937, 3.7583, 4.2242, 3.9723, 2.3997], device='cuda:0'), covar=tensor([0.0351, 0.0064, 0.0030, 0.0261, 0.0082, 0.0066, 0.0060, 0.0375], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0071, 0.0071, 0.0129, 0.0081, 0.0090, 0.0080, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 17:19:26,248 INFO [train.py:904] (0/8) Epoch 13, batch 250, loss[loss=0.1895, simple_loss=0.2745, pruned_loss=0.05231, over 16777.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2753, pruned_loss=0.05608, over 2369163.51 frames. ], batch size: 57, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:19:41,322 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122063.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:19:53,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8685, 2.2881, 2.4320, 4.6996, 2.2411, 2.7818, 2.4551, 2.5605], device='cuda:0'), covar=tensor([0.0907, 0.3505, 0.2432, 0.0364, 0.3811, 0.2239, 0.3148, 0.3453], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0393, 0.0333, 0.0316, 0.0411, 0.0448, 0.0359, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:19:54,206 INFO [optim.py:368] (0/8) 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 17:20:34,750 INFO [train.py:904] (0/8) Epoch 13, batch 300, loss[loss=0.1863, simple_loss=0.2773, pruned_loss=0.04771, over 17109.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2723, pruned_loss=0.05444, over 2579995.38 frames. ], batch size: 48, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:20:47,421 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122111.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:20:55,792 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:21:42,729 INFO [train.py:904] (0/8) Epoch 13, batch 350, loss[loss=0.1988, simple_loss=0.2856, pruned_loss=0.056, over 16791.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.2701, pruned_loss=0.05251, over 2745559.95 frames. ], batch size: 57, lr: 5.35e-03, grad_scale: 2.0 2023-04-29 17:22:13,945 INFO [optim.py:368] (0/8) 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 17:22:20,612 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122178.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:22:43,863 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122195.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:22:53,145 INFO [train.py:904] (0/8) Epoch 13, batch 400, loss[loss=0.1749, simple_loss=0.2663, pruned_loss=0.04171, over 17135.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2683, pruned_loss=0.05093, over 2875048.16 frames. ], batch size: 48, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:23:15,324 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3221, 3.5058, 3.7790, 2.2100, 3.0749, 2.4142, 3.6325, 3.6957], device='cuda:0'), covar=tensor([0.0224, 0.0783, 0.0485, 0.1715, 0.0750, 0.0917, 0.0581, 0.0907], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0142, 0.0157, 0.0143, 0.0137, 0.0124, 0.0134, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 17:24:03,766 INFO [train.py:904] (0/8) Epoch 13, batch 450, loss[loss=0.1896, simple_loss=0.2636, pruned_loss=0.05782, over 16883.00 frames. ], tot_loss[loss=0.183, simple_loss=0.266, pruned_loss=0.04995, over 2967276.89 frames. ], batch size: 116, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:24:09,627 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122256.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:24:32,944 INFO [optim.py:368] (0/8) 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 17:25:13,898 INFO [train.py:904] (0/8) Epoch 13, batch 500, loss[loss=0.183, simple_loss=0.2646, pruned_loss=0.05066, over 17246.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2639, pruned_loss=0.04896, over 3052604.19 frames. ], batch size: 45, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:25:18,863 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122306.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:25:21,337 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.10 vs. limit=2.0 2023-04-29 17:25:33,776 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3568, 4.3752, 4.8257, 4.8159, 4.8596, 4.5101, 4.5002, 4.3513], device='cuda:0'), covar=tensor([0.0359, 0.0665, 0.0458, 0.0437, 0.0436, 0.0398, 0.0829, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0359, 0.0359, 0.0341, 0.0406, 0.0382, 0.0476, 0.0309], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 17:25:36,916 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:26:21,451 INFO [train.py:904] (0/8) Epoch 13, batch 550, loss[loss=0.1582, simple_loss=0.2504, pruned_loss=0.033, over 17211.00 frames. ], tot_loss[loss=0.18, simple_loss=0.263, pruned_loss=0.0485, over 3106537.81 frames. ], batch size: 45, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:26:40,377 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8717, 2.2863, 2.3616, 4.6712, 2.3404, 2.7095, 2.4351, 2.4623], device='cuda:0'), covar=tensor([0.0929, 0.3449, 0.2496, 0.0374, 0.3772, 0.2412, 0.2967, 0.3608], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0399, 0.0338, 0.0323, 0.0417, 0.0457, 0.0366, 0.0467], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:26:50,249 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.266e+02 2.671e+02 3.065e+02 5.763e+02, threshold=5.342e+02, percent-clipped=1.0 2023-04-29 17:27:00,780 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:27:22,549 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0967, 1.9931, 2.5195, 3.0456, 2.8217, 3.4078, 2.4568, 3.4009], device='cuda:0'), covar=tensor([0.0176, 0.0365, 0.0246, 0.0210, 0.0220, 0.0147, 0.0307, 0.0101], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0175, 0.0160, 0.0163, 0.0174, 0.0128, 0.0174, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 17:27:30,043 INFO [train.py:904] (0/8) Epoch 13, batch 600, loss[loss=0.1817, simple_loss=0.2702, pruned_loss=0.04657, over 16756.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2616, pruned_loss=0.04799, over 3153989.14 frames. ], batch size: 57, lr: 5.35e-03, grad_scale: 4.0 2023-04-29 17:27:39,191 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6444, 3.6720, 1.9239, 3.9149, 2.7364, 3.8486, 2.0072, 2.8824], device='cuda:0'), covar=tensor([0.0217, 0.0304, 0.1686, 0.0238, 0.0764, 0.0493, 0.1714, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0164, 0.0188, 0.0136, 0.0170, 0.0205, 0.0197, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 17:27:49,683 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8461, 3.1793, 2.9546, 5.1621, 4.4479, 4.6985, 1.7400, 3.3950], device='cuda:0'), covar=tensor([0.1284, 0.0641, 0.0996, 0.0143, 0.0168, 0.0299, 0.1493, 0.0648], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0157, 0.0179, 0.0147, 0.0189, 0.0204, 0.0180, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 17:28:16,626 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.54 vs. limit=5.0 2023-04-29 17:28:37,803 INFO [train.py:904] (0/8) Epoch 13, batch 650, loss[loss=0.1908, simple_loss=0.2549, pruned_loss=0.06337, over 16455.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.26, pruned_loss=0.04759, over 3188361.37 frames. ], batch size: 146, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:29:01,231 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122468.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:29:07,723 INFO [optim.py:368] (0/8) 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 17:29:08,037 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122473.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:29:47,553 INFO [train.py:904] (0/8) Epoch 13, batch 700, loss[loss=0.1999, simple_loss=0.2755, pruned_loss=0.06212, over 16749.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2603, pruned_loss=0.04775, over 3224966.52 frames. ], batch size: 124, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:30:25,437 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:56,261 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122551.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:30:57,600 INFO [train.py:904] (0/8) Epoch 13, batch 750, loss[loss=0.1889, simple_loss=0.2609, pruned_loss=0.05849, over 16863.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.261, pruned_loss=0.04803, over 3255201.00 frames. ], batch size: 96, lr: 5.34e-03, grad_scale: 4.0 2023-04-29 17:31:27,840 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.426e+02 2.324e+02 2.801e+02 3.357e+02 5.795e+02, threshold=5.603e+02, percent-clipped=1.0 2023-04-29 17:32:09,553 INFO [train.py:904] (0/8) Epoch 13, batch 800, loss[loss=0.1858, simple_loss=0.2785, pruned_loss=0.04655, over 16651.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2606, pruned_loss=0.04752, over 3277279.85 frames. ], batch size: 62, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:32:15,172 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122606.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:15,742 INFO [train.py:904] (0/8) Epoch 13, batch 850, loss[loss=0.1505, simple_loss=0.2323, pruned_loss=0.03434, over 15951.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.26, pruned_loss=0.04738, over 3285854.77 frames. ], batch size: 35, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:33:18,243 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:20,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.60 vs. limit=5.0 2023-04-29 17:33:44,229 INFO [optim.py:368] (0/8) 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 17:33:46,993 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122675.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:33:53,180 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7900, 4.1053, 2.5396, 4.5372, 3.0770, 4.5134, 2.3518, 3.0409], device='cuda:0'), covar=tensor([0.0291, 0.0301, 0.1329, 0.0250, 0.0752, 0.0468, 0.1442, 0.0744], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0164, 0.0189, 0.0137, 0.0169, 0.0206, 0.0196, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 17:34:01,621 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 17:34:23,507 INFO [train.py:904] (0/8) Epoch 13, batch 900, loss[loss=0.1655, simple_loss=0.2588, pruned_loss=0.03614, over 17079.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2595, pruned_loss=0.04677, over 3302130.00 frames. ], batch size: 53, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:35:33,156 INFO [train.py:904] (0/8) Epoch 13, batch 950, loss[loss=0.1737, simple_loss=0.2725, pruned_loss=0.03743, over 17121.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2601, pruned_loss=0.047, over 3300619.00 frames. ], batch size: 48, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:35:39,074 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8406, 5.1910, 4.8945, 4.8644, 4.6601, 4.6002, 4.6772, 5.2777], device='cuda:0'), covar=tensor([0.1169, 0.0887, 0.1116, 0.0761, 0.0784, 0.1063, 0.1075, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0571, 0.0718, 0.0587, 0.0512, 0.0456, 0.0467, 0.0602, 0.0549], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:36:02,671 INFO [optim.py:368] (0/8) 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 17:36:03,090 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122773.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 17:36:43,328 INFO [train.py:904] (0/8) Epoch 13, batch 1000, loss[loss=0.1777, simple_loss=0.2483, pruned_loss=0.05354, over 16322.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2586, pruned_loss=0.04635, over 3305599.76 frames. ], batch size: 145, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:37:09,708 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122821.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:14,624 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=122824.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:39,800 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7577, 2.6787, 2.2909, 2.5438, 3.0522, 2.9192, 3.5927, 3.2973], device='cuda:0'), covar=tensor([0.0081, 0.0318, 0.0391, 0.0325, 0.0200, 0.0270, 0.0154, 0.0207], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0218, 0.0211, 0.0210, 0.0217, 0.0215, 0.0222, 0.0205], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:37:51,945 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122851.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:37:52,810 INFO [train.py:904] (0/8) Epoch 13, batch 1050, loss[loss=0.1721, simple_loss=0.2435, pruned_loss=0.05034, over 16273.00 frames. ], tot_loss[loss=0.175, simple_loss=0.258, pruned_loss=0.04605, over 3312801.57 frames. ], batch size: 165, lr: 5.34e-03, grad_scale: 8.0 2023-04-29 17:38:20,458 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=122872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:38:21,184 INFO [optim.py:368] (0/8) 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 17:38:57,672 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=122899.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:39:01,459 INFO [train.py:904] (0/8) Epoch 13, batch 1100, loss[loss=0.1611, simple_loss=0.2416, pruned_loss=0.04028, over 16636.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2571, pruned_loss=0.04582, over 3309915.64 frames. ], batch size: 76, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:39:45,849 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=122933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:40:11,778 INFO [train.py:904] (0/8) Epoch 13, batch 1150, loss[loss=0.1874, simple_loss=0.2618, pruned_loss=0.05649, over 16847.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2567, pruned_loss=0.04523, over 3309902.49 frames. ], batch size: 116, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:40:39,515 INFO [optim.py:368] (0/8) 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 17:40:43,355 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=122975.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:41:08,253 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.74 vs. limit=5.0 2023-04-29 17:41:20,387 INFO [train.py:904] (0/8) Epoch 13, batch 1200, loss[loss=0.1573, simple_loss=0.233, pruned_loss=0.0408, over 16733.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2565, pruned_loss=0.04528, over 3312155.93 frames. ], batch size: 83, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:41:28,703 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 17:41:50,203 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123023.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:42:04,626 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 17:42:30,145 INFO [train.py:904] (0/8) Epoch 13, batch 1250, loss[loss=0.1472, simple_loss=0.2326, pruned_loss=0.03089, over 17191.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2557, pruned_loss=0.04508, over 3319272.77 frames. ], batch size: 44, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:42:59,823 INFO [optim.py:368] (0/8) 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 17:43:07,826 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-29 17:43:40,470 INFO [train.py:904] (0/8) Epoch 13, batch 1300, loss[loss=0.1416, simple_loss=0.2322, pruned_loss=0.02548, over 16947.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2558, pruned_loss=0.04488, over 3322629.47 frames. ], batch size: 41, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:44:12,385 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123124.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:44:49,693 INFO [train.py:904] (0/8) Epoch 13, batch 1350, loss[loss=0.1747, simple_loss=0.2729, pruned_loss=0.03828, over 16704.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2568, pruned_loss=0.04547, over 3323072.63 frames. ], batch size: 62, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:45:17,082 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:45:17,979 INFO [optim.py:368] (0/8) 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 17:45:58,577 INFO [train.py:904] (0/8) Epoch 13, batch 1400, loss[loss=0.2059, simple_loss=0.2721, pruned_loss=0.06981, over 16832.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2566, pruned_loss=0.04547, over 3325919.23 frames. ], batch size: 96, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:46:16,213 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-04-29 17:46:35,860 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:47:09,281 INFO [train.py:904] (0/8) Epoch 13, batch 1450, loss[loss=0.1635, simple_loss=0.2365, pruned_loss=0.04527, over 12253.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2563, pruned_loss=0.04534, over 3318448.20 frames. ], batch size: 247, lr: 5.33e-03, grad_scale: 8.0 2023-04-29 17:47:38,913 INFO [optim.py:368] (0/8) 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 17:47:43,455 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 17:48:19,764 INFO [train.py:904] (0/8) Epoch 13, batch 1500, loss[loss=0.1825, simple_loss=0.2758, pruned_loss=0.04463, over 17111.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2558, pruned_loss=0.04604, over 3315312.28 frames. ], batch size: 49, lr: 5.33e-03, grad_scale: 4.0 2023-04-29 17:48:37,121 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123314.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:49:08,160 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6537, 6.0588, 5.7774, 5.8737, 5.3913, 5.3825, 5.4787, 6.1750], device='cuda:0'), covar=tensor([0.1241, 0.0928, 0.1045, 0.0681, 0.0738, 0.0635, 0.1047, 0.0994], device='cuda:0'), in_proj_covar=tensor([0.0578, 0.0733, 0.0591, 0.0520, 0.0463, 0.0473, 0.0611, 0.0561], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:49:30,721 INFO [train.py:904] (0/8) Epoch 13, batch 1550, loss[loss=0.2178, simple_loss=0.2824, pruned_loss=0.07663, over 16518.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2575, pruned_loss=0.04728, over 3322329.93 frames. ], batch size: 68, lr: 5.32e-03, grad_scale: 4.0 2023-04-29 17:50:00,254 INFO [optim.py:368] (0/8) 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 17:50:01,802 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123375.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:50:39,399 INFO [train.py:904] (0/8) Epoch 13, batch 1600, loss[loss=0.1723, simple_loss=0.2635, pruned_loss=0.04058, over 17242.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2596, pruned_loss=0.04791, over 3324734.95 frames. ], batch size: 44, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:51:14,556 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=123428.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:51:43,230 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-29 17:51:47,308 INFO [train.py:904] (0/8) Epoch 13, batch 1650, loss[loss=0.1739, simple_loss=0.2638, pruned_loss=0.04199, over 17037.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2615, pruned_loss=0.04888, over 3314862.47 frames. ], batch size: 53, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:52:18,027 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.454e+02 2.938e+02 3.518e+02 6.758e+02, threshold=5.875e+02, percent-clipped=2.0 2023-04-29 17:52:19,614 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7019, 2.9957, 3.1557, 2.0037, 2.7408, 2.1292, 3.2371, 3.2617], device='cuda:0'), covar=tensor([0.0228, 0.0796, 0.0486, 0.1670, 0.0771, 0.1000, 0.0529, 0.0809], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0148, 0.0158, 0.0144, 0.0137, 0.0125, 0.0136, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 17:52:38,218 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123489.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:52:55,812 INFO [train.py:904] (0/8) Epoch 13, batch 1700, loss[loss=0.2443, simple_loss=0.323, pruned_loss=0.08283, over 11929.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2635, pruned_loss=0.04941, over 3311380.93 frames. ], batch size: 247, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:53:31,948 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:54:04,791 INFO [train.py:904] (0/8) Epoch 13, batch 1750, loss[loss=0.1664, simple_loss=0.2528, pruned_loss=0.03996, over 17216.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2639, pruned_loss=0.049, over 3315197.57 frames. ], batch size: 43, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:54:17,158 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9299, 4.9838, 5.3966, 5.3931, 5.4179, 5.0852, 5.0247, 4.7972], device='cuda:0'), covar=tensor([0.0277, 0.0499, 0.0411, 0.0396, 0.0384, 0.0318, 0.0811, 0.0367], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0380, 0.0379, 0.0358, 0.0426, 0.0403, 0.0500, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 17:54:34,131 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=123576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:55:14,415 INFO [train.py:904] (0/8) Epoch 13, batch 1800, loss[loss=0.1951, simple_loss=0.2691, pruned_loss=0.06058, over 16791.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2653, pruned_loss=0.0493, over 3315718.92 frames. ], batch size: 102, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:56:23,380 INFO [train.py:904] (0/8) Epoch 13, batch 1850, loss[loss=0.1941, simple_loss=0.2917, pruned_loss=0.04827, over 17062.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2662, pruned_loss=0.04938, over 3310785.04 frames. ], batch size: 55, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:56:47,482 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123670.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:56:52,512 INFO [optim.py:368] (0/8) 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 17:57:33,306 INFO [train.py:904] (0/8) Epoch 13, batch 1900, loss[loss=0.1951, simple_loss=0.2709, pruned_loss=0.05969, over 16701.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2658, pruned_loss=0.04876, over 3312202.59 frames. ], batch size: 124, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:17,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2503, 5.6704, 5.4232, 5.4835, 5.0938, 4.9575, 5.0936, 5.8064], device='cuda:0'), covar=tensor([0.1170, 0.0842, 0.0985, 0.0728, 0.0775, 0.0701, 0.1041, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0581, 0.0731, 0.0592, 0.0520, 0.0462, 0.0471, 0.0609, 0.0560], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:58:30,328 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-29 17:58:39,346 INFO [train.py:904] (0/8) Epoch 13, batch 1950, loss[loss=0.1581, simple_loss=0.2462, pruned_loss=0.03499, over 16776.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.265, pruned_loss=0.04807, over 3316060.88 frames. ], batch size: 42, lr: 5.32e-03, grad_scale: 8.0 2023-04-29 17:58:56,952 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-29 17:59:09,942 INFO [optim.py:368] (0/8) 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 17:59:23,614 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7816, 4.1844, 3.0519, 2.2231, 2.7423, 2.4704, 4.3585, 3.6228], device='cuda:0'), covar=tensor([0.2559, 0.0515, 0.1603, 0.2442, 0.2579, 0.1860, 0.0367, 0.1117], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0262, 0.0289, 0.0284, 0.0283, 0.0228, 0.0272, 0.0306], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 17:59:25,268 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=123784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 17:59:48,923 INFO [train.py:904] (0/8) Epoch 13, batch 2000, loss[loss=0.1829, simple_loss=0.2658, pruned_loss=0.05004, over 17261.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2649, pruned_loss=0.0479, over 3316341.02 frames. ], batch size: 52, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:00:59,637 INFO [train.py:904] (0/8) Epoch 13, batch 2050, loss[loss=0.2074, simple_loss=0.2933, pruned_loss=0.06069, over 16731.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2649, pruned_loss=0.04826, over 3304798.20 frames. ], batch size: 62, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:01:28,767 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 18:02:09,977 INFO [train.py:904] (0/8) Epoch 13, batch 2100, loss[loss=0.2112, simple_loss=0.3004, pruned_loss=0.06098, over 17073.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2658, pruned_loss=0.04871, over 3305397.92 frames. ], batch size: 53, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:02:54,514 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=123933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:20,336 INFO [train.py:904] (0/8) Epoch 13, batch 2150, loss[loss=0.1757, simple_loss=0.2618, pruned_loss=0.04476, over 17085.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2666, pruned_loss=0.04924, over 3318233.14 frames. ], batch size: 53, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:03:26,794 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0845, 5.0188, 4.8459, 4.0364, 4.8821, 1.7921, 4.6073, 4.7144], device='cuda:0'), covar=tensor([0.0092, 0.0083, 0.0177, 0.0485, 0.0111, 0.2711, 0.0164, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0132, 0.0178, 0.0166, 0.0150, 0.0192, 0.0168, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:03:44,930 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=123970.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:03:50,031 INFO [optim.py:368] (0/8) 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 18:04:02,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3597, 5.2813, 5.2026, 4.7648, 4.7983, 5.2017, 5.2535, 4.8335], device='cuda:0'), covar=tensor([0.0532, 0.0472, 0.0226, 0.0268, 0.0960, 0.0442, 0.0275, 0.0653], device='cuda:0'), in_proj_covar=tensor([0.0270, 0.0363, 0.0322, 0.0300, 0.0342, 0.0345, 0.0217, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:04:25,047 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-124000.pt 2023-04-29 18:04:30,743 INFO [train.py:904] (0/8) Epoch 13, batch 2200, loss[loss=0.1988, simple_loss=0.2735, pruned_loss=0.06205, over 16818.00 frames. ], tot_loss[loss=0.183, simple_loss=0.267, pruned_loss=0.04953, over 3329126.46 frames. ], batch size: 102, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:04:45,255 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 18:04:53,410 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:19,621 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:05:40,214 INFO [train.py:904] (0/8) Epoch 13, batch 2250, loss[loss=0.2001, simple_loss=0.289, pruned_loss=0.05561, over 16717.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.267, pruned_loss=0.04961, over 3332155.64 frames. ], batch size: 57, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:05:53,617 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124062.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:09,555 INFO [optim.py:368] (0/8) 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 18:06:10,137 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7646, 2.2149, 2.3431, 4.5738, 2.2367, 2.7298, 2.3963, 2.4719], device='cuda:0'), covar=tensor([0.0954, 0.3388, 0.2467, 0.0390, 0.3895, 0.2306, 0.3122, 0.3369], device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0405, 0.0341, 0.0327, 0.0418, 0.0467, 0.0371, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:06:22,680 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:42,862 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124097.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:06:48,033 INFO [train.py:904] (0/8) Epoch 13, batch 2300, loss[loss=0.2166, simple_loss=0.2937, pruned_loss=0.0697, over 15653.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2669, pruned_loss=0.04924, over 3337916.97 frames. ], batch size: 190, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:07:17,837 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124123.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:07:28,760 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124132.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:07:56,751 INFO [train.py:904] (0/8) Epoch 13, batch 2350, loss[loss=0.1907, simple_loss=0.2829, pruned_loss=0.04925, over 17062.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2671, pruned_loss=0.04951, over 3345378.15 frames. ], batch size: 55, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:08:26,418 INFO [optim.py:368] (0/8) 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 18:09:06,178 INFO [train.py:904] (0/8) Epoch 13, batch 2400, loss[loss=0.2596, simple_loss=0.3321, pruned_loss=0.09353, over 12139.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2671, pruned_loss=0.04924, over 3338904.27 frames. ], batch size: 246, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:09:42,802 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:10:15,683 INFO [train.py:904] (0/8) Epoch 13, batch 2450, loss[loss=0.1773, simple_loss=0.2809, pruned_loss=0.03681, over 17285.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2683, pruned_loss=0.04959, over 3337050.14 frames. ], batch size: 52, lr: 5.31e-03, grad_scale: 8.0 2023-04-29 18:10:43,290 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124272.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:10:45,732 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 18:10:46,036 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.400e+02 2.819e+02 3.354e+02 7.582e+02, threshold=5.638e+02, percent-clipped=2.0 2023-04-29 18:11:20,661 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-29 18:11:24,243 INFO [train.py:904] (0/8) Epoch 13, batch 2500, loss[loss=0.1987, simple_loss=0.2906, pruned_loss=0.05336, over 17061.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2673, pruned_loss=0.04884, over 3333613.02 frames. ], batch size: 53, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:12:03,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4578, 3.5187, 2.0951, 3.7690, 2.6507, 3.6799, 2.1528, 2.7924], device='cuda:0'), covar=tensor([0.0247, 0.0361, 0.1406, 0.0273, 0.0762, 0.0662, 0.1296, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0170, 0.0191, 0.0146, 0.0172, 0.0215, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 18:12:09,463 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124333.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:12:18,730 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4934, 4.4018, 4.3864, 4.1425, 4.1064, 4.4663, 4.2328, 4.1812], device='cuda:0'), covar=tensor([0.0709, 0.0718, 0.0285, 0.0260, 0.0820, 0.0494, 0.0667, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0367, 0.0324, 0.0303, 0.0346, 0.0346, 0.0219, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:12:31,533 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124348.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:12:37,103 INFO [train.py:904] (0/8) Epoch 13, batch 2550, loss[loss=0.1947, simple_loss=0.2637, pruned_loss=0.06292, over 16884.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2675, pruned_loss=0.04934, over 3332720.24 frames. ], batch size: 109, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:06,418 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.367e+02 2.275e+02 2.738e+02 3.268e+02 6.282e+02, threshold=5.476e+02, percent-clipped=1.0 2023-04-29 18:13:24,654 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-29 18:13:29,912 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-04-29 18:13:31,401 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124392.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:13:44,315 INFO [train.py:904] (0/8) Epoch 13, batch 2600, loss[loss=0.2129, simple_loss=0.3097, pruned_loss=0.0581, over 17031.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2679, pruned_loss=0.0494, over 3331710.54 frames. ], batch size: 55, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:13:54,876 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124409.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 18:14:06,292 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:14:54,848 INFO [train.py:904] (0/8) Epoch 13, batch 2650, loss[loss=0.1921, simple_loss=0.2917, pruned_loss=0.04631, over 17076.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2694, pruned_loss=0.04975, over 3322984.29 frames. ], batch size: 55, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:15:22,355 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9052, 3.2653, 3.1289, 2.1857, 2.7412, 2.3657, 3.4150, 3.4608], device='cuda:0'), covar=tensor([0.0281, 0.0806, 0.0673, 0.1679, 0.0851, 0.0919, 0.0609, 0.0846], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0138, 0.0127, 0.0139, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 18:15:25,990 INFO [optim.py:368] (0/8) 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 18:15:52,771 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:16:04,558 INFO [train.py:904] (0/8) Epoch 13, batch 2700, loss[loss=0.1585, simple_loss=0.2482, pruned_loss=0.03437, over 16813.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2691, pruned_loss=0.04912, over 3323648.86 frames. ], batch size: 39, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:16:40,366 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:16:43,835 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7268, 4.6707, 5.1148, 5.1324, 5.1824, 4.7622, 4.7873, 4.5259], device='cuda:0'), covar=tensor([0.0299, 0.0495, 0.0375, 0.0344, 0.0429, 0.0345, 0.0856, 0.0466], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0383, 0.0381, 0.0360, 0.0431, 0.0406, 0.0504, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 18:17:13,323 INFO [train.py:904] (0/8) Epoch 13, batch 2750, loss[loss=0.188, simple_loss=0.2679, pruned_loss=0.05407, over 16727.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2679, pruned_loss=0.04812, over 3327820.25 frames. ], batch size: 124, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:17:24,711 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-29 18:17:44,015 INFO [optim.py:368] (0/8) 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 18:17:46,593 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:17:52,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5497, 4.4352, 4.4708, 4.7505, 4.8645, 4.4239, 4.7764, 4.8576], device='cuda:0'), covar=tensor([0.1477, 0.1144, 0.1840, 0.0802, 0.0715, 0.1012, 0.1357, 0.0761], device='cuda:0'), in_proj_covar=tensor([0.0582, 0.0724, 0.0876, 0.0734, 0.0551, 0.0582, 0.0584, 0.0676], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:18:22,960 INFO [train.py:904] (0/8) Epoch 13, batch 2800, loss[loss=0.1757, simple_loss=0.27, pruned_loss=0.04072, over 17050.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2674, pruned_loss=0.04808, over 3326994.12 frames. ], batch size: 55, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:18:23,505 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6852, 2.2685, 2.2832, 4.5954, 2.2378, 2.8093, 2.4406, 2.4860], device='cuda:0'), covar=tensor([0.1047, 0.3515, 0.2588, 0.0350, 0.3846, 0.2292, 0.3178, 0.3333], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0406, 0.0339, 0.0325, 0.0417, 0.0468, 0.0369, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:18:45,965 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1905, 5.7015, 5.8732, 5.5899, 5.6775, 6.2402, 5.6859, 5.4114], device='cuda:0'), covar=tensor([0.0800, 0.1878, 0.2028, 0.2222, 0.2566, 0.0977, 0.1469, 0.2404], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0535, 0.0582, 0.0465, 0.0628, 0.0609, 0.0461, 0.0615], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:18:59,126 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124628.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:07,693 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-29 18:19:12,843 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:19:31,063 INFO [train.py:904] (0/8) Epoch 13, batch 2850, loss[loss=0.1583, simple_loss=0.2393, pruned_loss=0.03865, over 15900.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2668, pruned_loss=0.0479, over 3325014.83 frames. ], batch size: 35, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:19:46,071 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=124663.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:00,132 INFO [optim.py:368] (0/8) 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 18:20:06,737 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8583, 3.8471, 4.2024, 4.1952, 4.2409, 3.9581, 3.9972, 3.9124], device='cuda:0'), covar=tensor([0.0424, 0.0667, 0.0515, 0.0541, 0.0580, 0.0541, 0.0908, 0.0660], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0384, 0.0384, 0.0363, 0.0433, 0.0408, 0.0506, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 18:20:24,624 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124692.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:33,933 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124699.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:20:37,526 INFO [train.py:904] (0/8) Epoch 13, batch 2900, loss[loss=0.1924, simple_loss=0.2967, pruned_loss=0.04406, over 17134.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2661, pruned_loss=0.0479, over 3331303.74 frames. ], batch size: 49, lr: 5.30e-03, grad_scale: 8.0 2023-04-29 18:20:40,112 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124704.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:20:58,954 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124718.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:07,446 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124724.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:29,784 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:21:46,245 INFO [train.py:904] (0/8) Epoch 13, batch 2950, loss[loss=0.2879, simple_loss=0.3565, pruned_loss=0.1096, over 12089.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2652, pruned_loss=0.04868, over 3320017.18 frames. ], batch size: 246, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:01,277 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.57 vs. limit=5.0 2023-04-29 18:22:05,798 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124766.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:16,913 INFO [optim.py:368] (0/8) 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 18:22:26,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6953, 2.9033, 2.8989, 4.9194, 4.0181, 4.4159, 1.6482, 3.2360], device='cuda:0'), covar=tensor([0.1348, 0.0697, 0.1038, 0.0190, 0.0253, 0.0383, 0.1464, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0185, 0.0159, 0.0200, 0.0211, 0.0185, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 18:22:35,745 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:22:55,206 INFO [train.py:904] (0/8) Epoch 13, batch 3000, loss[loss=0.1604, simple_loss=0.2514, pruned_loss=0.03473, over 17229.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2666, pruned_loss=0.04961, over 3319028.81 frames. ], batch size: 44, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:22:55,207 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 18:23:04,001 INFO [train.py:938] (0/8) Epoch 13, validation: loss=0.1391, simple_loss=0.2452, pruned_loss=0.01648, over 944034.00 frames. 2023-04-29 18:23:04,002 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 18:23:29,533 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 13, batch 3050, loss[loss=0.2044, simple_loss=0.2761, pruned_loss=0.06639, over 16769.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2659, pruned_loss=0.04939, over 3315091.22 frames. ], batch size: 102, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:24:33,565 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-29 18:24:44,777 INFO [optim.py:368] (0/8) 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 18:24:45,235 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5145, 3.4401, 3.6991, 1.9743, 3.7736, 3.7742, 3.0699, 2.9096], device='cuda:0'), covar=tensor([0.0685, 0.0186, 0.0152, 0.1044, 0.0067, 0.0143, 0.0352, 0.0391], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0102, 0.0090, 0.0138, 0.0071, 0.0112, 0.0122, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 18:24:55,378 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=124881.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:25:25,290 INFO [train.py:904] (0/8) Epoch 13, batch 3100, loss[loss=0.1555, simple_loss=0.2421, pruned_loss=0.03449, over 17260.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2644, pruned_loss=0.04843, over 3328785.24 frames. ], batch size: 44, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:26:01,299 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=124928.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:26:34,838 INFO [train.py:904] (0/8) Epoch 13, batch 3150, loss[loss=0.162, simple_loss=0.2568, pruned_loss=0.03367, over 17215.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2638, pruned_loss=0.04818, over 3331849.89 frames. ], batch size: 45, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:05,840 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.452e+02 2.205e+02 2.626e+02 3.111e+02 5.476e+02, threshold=5.252e+02, percent-clipped=0.0 2023-04-29 18:27:09,072 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=124976.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:34,285 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=124994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:27:34,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0500, 4.6283, 3.4039, 2.4274, 3.0962, 2.7761, 4.8938, 4.0543], device='cuda:0'), covar=tensor([0.2465, 0.0479, 0.1378, 0.2416, 0.2536, 0.1745, 0.0268, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0259, 0.0287, 0.0284, 0.0283, 0.0229, 0.0272, 0.0308], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:27:45,743 INFO [train.py:904] (0/8) Epoch 13, batch 3200, loss[loss=0.1737, simple_loss=0.267, pruned_loss=0.04021, over 17037.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2632, pruned_loss=0.04774, over 3333166.50 frames. ], batch size: 50, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:27:48,507 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125004.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:27:52,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6494, 4.9588, 4.7299, 4.7080, 4.4636, 4.4268, 4.4466, 5.0523], device='cuda:0'), covar=tensor([0.1099, 0.0853, 0.1058, 0.0801, 0.0865, 0.1213, 0.1132, 0.0900], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0741, 0.0600, 0.0524, 0.0469, 0.0475, 0.0614, 0.0564], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:28:10,549 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125019.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:28:50,442 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0633, 5.0370, 4.9151, 4.4924, 4.4986, 4.9644, 4.9117, 4.6101], device='cuda:0'), covar=tensor([0.0597, 0.0552, 0.0274, 0.0312, 0.1053, 0.0472, 0.0376, 0.0764], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0371, 0.0328, 0.0307, 0.0351, 0.0354, 0.0223, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:28:56,250 INFO [train.py:904] (0/8) Epoch 13, batch 3250, loss[loss=0.1694, simple_loss=0.2593, pruned_loss=0.03977, over 17209.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2632, pruned_loss=0.04792, over 3324834.77 frames. ], batch size: 46, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:28:56,533 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125052.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:29:19,886 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-29 18:29:27,130 INFO [optim.py:368] (0/8) 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 18:29:32,647 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9107, 4.2401, 4.4924, 4.5099, 4.4281, 4.1780, 3.8549, 4.0967], device='cuda:0'), covar=tensor([0.0581, 0.0634, 0.0486, 0.0528, 0.0723, 0.0589, 0.1334, 0.0642], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0387, 0.0386, 0.0365, 0.0435, 0.0409, 0.0513, 0.0327], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 18:29:46,235 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:05,574 INFO [train.py:904] (0/8) Epoch 13, batch 3300, loss[loss=0.1871, simple_loss=0.2799, pruned_loss=0.04713, over 17059.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2643, pruned_loss=0.04807, over 3326678.43 frames. ], batch size: 53, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:30:41,585 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.93 vs. limit=5.0 2023-04-29 18:30:42,309 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125128.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:53,995 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:30:55,329 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9841, 2.5296, 2.7069, 1.8574, 2.7869, 2.8170, 2.4522, 2.3467], device='cuda:0'), covar=tensor([0.0695, 0.0232, 0.0200, 0.0934, 0.0109, 0.0259, 0.0455, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0137, 0.0071, 0.0111, 0.0121, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 18:31:15,062 INFO [train.py:904] (0/8) Epoch 13, batch 3350, loss[loss=0.1988, simple_loss=0.2801, pruned_loss=0.05872, over 15518.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2651, pruned_loss=0.04838, over 3325679.17 frames. ], batch size: 190, lr: 5.29e-03, grad_scale: 8.0 2023-04-29 18:31:44,161 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6934, 2.6024, 2.5246, 3.7263, 3.0082, 3.8080, 1.4777, 2.8035], device='cuda:0'), covar=tensor([0.1351, 0.0637, 0.1040, 0.0196, 0.0172, 0.0361, 0.1506, 0.0778], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0162, 0.0183, 0.0159, 0.0199, 0.0212, 0.0184, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 18:31:45,943 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.501e+02 2.375e+02 2.737e+02 3.421e+02 8.798e+02, threshold=5.473e+02, percent-clipped=2.0 2023-04-29 18:31:48,362 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125176.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:31:51,551 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8105, 3.0453, 2.6528, 4.6855, 3.7819, 4.2271, 1.6595, 3.1486], device='cuda:0'), covar=tensor([0.1390, 0.0681, 0.1208, 0.0179, 0.0296, 0.0430, 0.1585, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0162, 0.0183, 0.0160, 0.0200, 0.0212, 0.0185, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 18:32:06,968 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125189.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:32:24,429 INFO [train.py:904] (0/8) Epoch 13, batch 3400, loss[loss=0.1516, simple_loss=0.2395, pruned_loss=0.03181, over 17181.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2645, pruned_loss=0.04781, over 3323484.54 frames. ], batch size: 46, lr: 5.29e-03, grad_scale: 4.0 2023-04-29 18:32:59,625 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5575, 4.4024, 4.4206, 4.1556, 4.1098, 4.5094, 4.3167, 4.1840], device='cuda:0'), covar=tensor([0.0541, 0.0724, 0.0270, 0.0298, 0.0908, 0.0397, 0.0489, 0.0646], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0368, 0.0324, 0.0304, 0.0349, 0.0350, 0.0220, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:33:35,403 INFO [train.py:904] (0/8) Epoch 13, batch 3450, loss[loss=0.1831, simple_loss=0.2637, pruned_loss=0.05126, over 16724.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.264, pruned_loss=0.04729, over 3325170.64 frames. ], batch size: 134, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:33:42,792 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-29 18:34:04,770 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9982, 4.4992, 4.5906, 3.2169, 3.7993, 4.4836, 3.9985, 2.5296], device='cuda:0'), covar=tensor([0.0372, 0.0035, 0.0021, 0.0252, 0.0084, 0.0064, 0.0063, 0.0369], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0074, 0.0073, 0.0128, 0.0085, 0.0094, 0.0084, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:34:07,280 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.453e+02 2.393e+02 2.670e+02 3.290e+02 7.203e+02, threshold=5.341e+02, percent-clipped=1.0 2023-04-29 18:34:34,796 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125294.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:34:40,450 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5466, 3.4296, 3.8147, 1.9116, 3.8911, 3.9477, 3.1310, 3.0605], device='cuda:0'), covar=tensor([0.0703, 0.0211, 0.0167, 0.1060, 0.0072, 0.0153, 0.0322, 0.0377], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0137, 0.0070, 0.0112, 0.0121, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 18:34:46,249 INFO [train.py:904] (0/8) Epoch 13, batch 3500, loss[loss=0.2134, simple_loss=0.2856, pruned_loss=0.07057, over 11938.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2635, pruned_loss=0.04683, over 3320829.17 frames. ], batch size: 248, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:35:09,800 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125319.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:35:42,977 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125342.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:35:57,854 INFO [train.py:904] (0/8) Epoch 13, batch 3550, loss[loss=0.1716, simple_loss=0.2542, pruned_loss=0.04451, over 16537.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2624, pruned_loss=0.04669, over 3327578.99 frames. ], batch size: 75, lr: 5.28e-03, grad_scale: 4.0 2023-04-29 18:36:03,220 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:18,667 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125367.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:30,598 INFO [optim.py:368] (0/8) 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 18:36:32,290 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:36:39,767 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2743, 4.6192, 4.6111, 3.3424, 3.8859, 4.5043, 3.9919, 2.8740], device='cuda:0'), covar=tensor([0.0324, 0.0034, 0.0028, 0.0261, 0.0087, 0.0069, 0.0063, 0.0329], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0074, 0.0073, 0.0128, 0.0085, 0.0094, 0.0083, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:37:08,979 INFO [train.py:904] (0/8) Epoch 13, batch 3600, loss[loss=0.1666, simple_loss=0.2494, pruned_loss=0.04192, over 16831.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2608, pruned_loss=0.0463, over 3322741.78 frames. ], batch size: 42, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:37:31,370 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125417.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 18:37:35,915 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125421.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:37:40,314 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6273, 2.4044, 1.8025, 2.2310, 2.7135, 2.5226, 2.8644, 2.8637], device='cuda:0'), covar=tensor([0.0134, 0.0282, 0.0414, 0.0313, 0.0174, 0.0255, 0.0164, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0218, 0.0209, 0.0211, 0.0220, 0.0217, 0.0228, 0.0209], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:37:59,867 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:38:01,064 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8691, 3.3652, 3.0285, 5.0894, 4.3275, 4.5994, 1.7328, 3.4388], device='cuda:0'), covar=tensor([0.1243, 0.0596, 0.0969, 0.0168, 0.0249, 0.0337, 0.1459, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0162, 0.0183, 0.0159, 0.0199, 0.0210, 0.0183, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 18:38:21,362 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9378, 2.1351, 2.3742, 3.1891, 2.2014, 2.2957, 2.3147, 2.1976], device='cuda:0'), covar=tensor([0.1035, 0.2861, 0.1965, 0.0570, 0.3398, 0.2069, 0.2455, 0.2861], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0409, 0.0342, 0.0328, 0.0421, 0.0473, 0.0373, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:38:21,929 INFO [train.py:904] (0/8) Epoch 13, batch 3650, loss[loss=0.2305, simple_loss=0.2973, pruned_loss=0.08183, over 11794.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.26, pruned_loss=0.04754, over 3305297.41 frames. ], batch size: 247, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:38:28,343 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-29 18:38:57,387 INFO [optim.py:368] (0/8) 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 18:38:59,604 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125476.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:07,718 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125482.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:10,780 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125484.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:39:36,246 INFO [train.py:904] (0/8) Epoch 13, batch 3700, loss[loss=0.1939, simple_loss=0.2551, pruned_loss=0.06638, over 16892.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2587, pruned_loss=0.04932, over 3283676.53 frames. ], batch size: 109, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:39:43,536 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.07 vs. limit=5.0 2023-04-29 18:40:10,904 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125524.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:40:24,990 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6373, 4.4870, 4.6912, 4.8487, 4.9809, 4.5123, 4.8385, 4.9448], device='cuda:0'), covar=tensor([0.1349, 0.1156, 0.1324, 0.0720, 0.0522, 0.0886, 0.1085, 0.0771], device='cuda:0'), in_proj_covar=tensor([0.0571, 0.0721, 0.0869, 0.0728, 0.0545, 0.0572, 0.0577, 0.0668], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:40:51,181 INFO [train.py:904] (0/8) Epoch 13, batch 3750, loss[loss=0.1613, simple_loss=0.2287, pruned_loss=0.04696, over 16943.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2599, pruned_loss=0.05109, over 3284415.73 frames. ], batch size: 90, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:40:53,153 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.67 vs. limit=5.0 2023-04-29 18:41:16,422 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3164, 4.3400, 4.5125, 2.9393, 3.7831, 4.3757, 3.9494, 2.3450], device='cuda:0'), covar=tensor([0.0474, 0.0023, 0.0018, 0.0301, 0.0072, 0.0068, 0.0051, 0.0361], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0073, 0.0071, 0.0126, 0.0083, 0.0092, 0.0082, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:41:24,166 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.582e+02 2.222e+02 2.735e+02 3.211e+02 5.083e+02, threshold=5.471e+02, percent-clipped=0.0 2023-04-29 18:42:05,170 INFO [train.py:904] (0/8) Epoch 13, batch 3800, loss[loss=0.1812, simple_loss=0.2596, pruned_loss=0.05138, over 16565.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2607, pruned_loss=0.05223, over 3282353.69 frames. ], batch size: 68, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:42:07,385 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125603.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:42:11,226 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5039, 3.5010, 1.8224, 3.6924, 2.5094, 3.6389, 1.9171, 2.8804], device='cuda:0'), covar=tensor([0.0172, 0.0286, 0.1440, 0.0190, 0.0643, 0.0445, 0.1273, 0.0485], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0168, 0.0189, 0.0144, 0.0169, 0.0215, 0.0198, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 18:42:32,155 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4120, 1.6031, 2.1231, 2.2599, 2.4738, 2.3240, 1.6029, 2.5227], device='cuda:0'), covar=tensor([0.0148, 0.0350, 0.0215, 0.0199, 0.0192, 0.0209, 0.0373, 0.0084], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0179, 0.0164, 0.0170, 0.0178, 0.0133, 0.0178, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:43:18,387 INFO [train.py:904] (0/8) Epoch 13, batch 3850, loss[loss=0.1903, simple_loss=0.2712, pruned_loss=0.05471, over 16804.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.261, pruned_loss=0.05287, over 3290101.18 frames. ], batch size: 124, lr: 5.28e-03, grad_scale: 8.0 2023-04-29 18:43:29,193 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 18:43:35,069 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125664.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:43:50,565 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.444e+02 2.805e+02 3.444e+02 9.574e+02, threshold=5.610e+02, percent-clipped=2.0 2023-04-29 18:44:25,654 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-29 18:44:29,774 INFO [train.py:904] (0/8) Epoch 13, batch 3900, loss[loss=0.1943, simple_loss=0.2648, pruned_loss=0.06186, over 16787.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2604, pruned_loss=0.05317, over 3289716.33 frames. ], batch size: 102, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:44:45,088 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125712.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 18:45:03,284 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125725.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:45:14,754 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125732.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:45:14,894 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6084, 4.6775, 4.8563, 3.1139, 4.0914, 4.6780, 4.1308, 2.6443], device='cuda:0'), covar=tensor([0.0439, 0.0018, 0.0015, 0.0292, 0.0056, 0.0056, 0.0054, 0.0325], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0072, 0.0072, 0.0127, 0.0083, 0.0092, 0.0083, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:45:42,349 INFO [train.py:904] (0/8) Epoch 13, batch 3950, loss[loss=0.1821, simple_loss=0.2497, pruned_loss=0.0573, over 16769.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2609, pruned_loss=0.05386, over 3295730.80 frames. ], batch size: 83, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:46:16,760 INFO [optim.py:368] (0/8) 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 18:46:18,419 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2132, 2.7696, 2.1507, 2.4967, 3.1082, 2.7426, 3.2414, 3.2097], device='cuda:0'), covar=tensor([0.0146, 0.0263, 0.0407, 0.0335, 0.0164, 0.0275, 0.0181, 0.0187], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0216, 0.0207, 0.0209, 0.0216, 0.0214, 0.0226, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:46:19,423 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125777.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:29,393 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=125784.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:32,665 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=125786.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:46:45,069 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6601, 4.7753, 4.8875, 3.0932, 4.1260, 4.7098, 4.1539, 2.6445], device='cuda:0'), covar=tensor([0.0422, 0.0014, 0.0015, 0.0300, 0.0060, 0.0058, 0.0047, 0.0335], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0073, 0.0072, 0.0128, 0.0084, 0.0094, 0.0084, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:46:55,495 INFO [train.py:904] (0/8) Epoch 13, batch 4000, loss[loss=0.1839, simple_loss=0.2564, pruned_loss=0.05574, over 16884.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2606, pruned_loss=0.05423, over 3300028.45 frames. ], batch size: 109, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:47:04,986 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6821, 4.9016, 5.0596, 4.8947, 4.9325, 5.5340, 5.0651, 4.7577], device='cuda:0'), covar=tensor([0.1128, 0.1722, 0.1864, 0.2103, 0.2749, 0.0971, 0.1361, 0.2334], device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0528, 0.0571, 0.0456, 0.0614, 0.0594, 0.0453, 0.0607], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 18:47:37,208 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=125832.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:48:05,370 INFO [train.py:904] (0/8) Epoch 13, batch 4050, loss[loss=0.1822, simple_loss=0.2608, pruned_loss=0.05182, over 12100.00 frames. ], tot_loss[loss=0.1835, simple_loss=0.2611, pruned_loss=0.05298, over 3295449.87 frames. ], batch size: 246, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:48:36,497 INFO [optim.py:368] (0/8) 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 18:48:49,465 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 18:49:15,129 INFO [train.py:904] (0/8) Epoch 13, batch 4100, loss[loss=0.1872, simple_loss=0.2767, pruned_loss=0.04886, over 16709.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2628, pruned_loss=0.0524, over 3284636.95 frames. ], batch size: 68, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:50:30,158 INFO [train.py:904] (0/8) Epoch 13, batch 4150, loss[loss=0.2477, simple_loss=0.3134, pruned_loss=0.09103, over 11449.00 frames. ], tot_loss[loss=0.19, simple_loss=0.27, pruned_loss=0.05501, over 3245488.63 frames. ], batch size: 246, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:50:42,649 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=125959.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:05,851 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.529e+02 2.394e+02 2.865e+02 3.650e+02 6.775e+02, threshold=5.730e+02, percent-clipped=10.0 2023-04-29 18:51:40,612 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=125998.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:51:42,827 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-126000.pt 2023-04-29 18:51:48,710 INFO [train.py:904] (0/8) Epoch 13, batch 4200, loss[loss=0.2109, simple_loss=0.2982, pruned_loss=0.06176, over 16322.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2771, pruned_loss=0.05639, over 3222387.28 frames. ], batch size: 35, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:52:04,530 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126012.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 18:52:12,856 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8376, 3.9592, 4.2416, 4.2414, 4.2406, 3.9696, 3.9263, 3.9181], device='cuda:0'), covar=tensor([0.0346, 0.0510, 0.0407, 0.0409, 0.0461, 0.0397, 0.1056, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0363, 0.0362, 0.0347, 0.0410, 0.0385, 0.0479, 0.0308], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 18:52:34,316 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:02,532 INFO [train.py:904] (0/8) Epoch 13, batch 4250, loss[loss=0.2066, simple_loss=0.2954, pruned_loss=0.05896, over 16363.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2798, pruned_loss=0.05608, over 3202076.76 frames. ], batch size: 35, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:53:13,055 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126059.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:14,084 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126060.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:36,049 INFO [optim.py:368] (0/8) 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 18:53:39,751 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126077.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:43,857 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:53:45,079 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126081.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:54:14,715 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2792, 4.1346, 4.3042, 4.4867, 4.5836, 4.2085, 4.5392, 4.6196], device='cuda:0'), covar=tensor([0.1359, 0.1054, 0.1368, 0.0593, 0.0452, 0.1090, 0.0634, 0.0588], device='cuda:0'), in_proj_covar=tensor([0.0545, 0.0688, 0.0825, 0.0693, 0.0522, 0.0546, 0.0554, 0.0635], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:54:16,804 INFO [train.py:904] (0/8) Epoch 13, batch 4300, loss[loss=0.209, simple_loss=0.2877, pruned_loss=0.06519, over 11892.00 frames. ], tot_loss[loss=0.1952, simple_loss=0.2806, pruned_loss=0.05492, over 3184227.48 frames. ], batch size: 246, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:54:18,566 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4740, 2.2163, 1.7812, 1.9128, 2.4402, 2.1371, 2.4395, 2.6178], device='cuda:0'), covar=tensor([0.0124, 0.0264, 0.0393, 0.0341, 0.0171, 0.0277, 0.0128, 0.0178], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0212, 0.0206, 0.0206, 0.0213, 0.0211, 0.0220, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 18:54:51,668 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126125.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:55:30,691 INFO [train.py:904] (0/8) Epoch 13, batch 4350, loss[loss=0.1894, simple_loss=0.2857, pruned_loss=0.04652, over 16443.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2843, pruned_loss=0.05626, over 3179645.68 frames. ], batch size: 75, lr: 5.27e-03, grad_scale: 8.0 2023-04-29 18:56:06,024 INFO [optim.py:368] (0/8) 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 18:56:46,675 INFO [train.py:904] (0/8) Epoch 13, batch 4400, loss[loss=0.172, simple_loss=0.2669, pruned_loss=0.03853, over 16944.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2861, pruned_loss=0.05733, over 3179679.66 frames. ], batch size: 96, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:57:59,274 INFO [train.py:904] (0/8) Epoch 13, batch 4450, loss[loss=0.2193, simple_loss=0.2962, pruned_loss=0.07117, over 16367.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.29, pruned_loss=0.05867, over 3194106.38 frames. ], batch size: 35, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:58:10,188 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126259.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 18:58:33,038 INFO [optim.py:368] (0/8) 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 18:59:14,449 INFO [train.py:904] (0/8) Epoch 13, batch 4500, loss[loss=0.1925, simple_loss=0.2819, pruned_loss=0.05156, over 16713.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2904, pruned_loss=0.05892, over 3197145.32 frames. ], batch size: 57, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 18:59:21,403 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126307.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:25,573 INFO [train.py:904] (0/8) Epoch 13, batch 4550, loss[loss=0.2045, simple_loss=0.2829, pruned_loss=0.0631, over 16597.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2907, pruned_loss=0.05943, over 3199687.52 frames. ], batch size: 62, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:00:28,795 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126354.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:00:59,202 INFO [optim.py:368] (0/8) 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 19:01:06,186 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126380.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:07,336 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:01:13,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6190, 3.6063, 1.9046, 4.2086, 2.5730, 4.1210, 2.3103, 2.7474], device='cuda:0'), covar=tensor([0.0206, 0.0315, 0.1778, 0.0102, 0.0844, 0.0387, 0.1385, 0.0746], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0165, 0.0188, 0.0138, 0.0166, 0.0209, 0.0195, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:01:35,441 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4726, 4.4319, 4.2286, 3.4801, 4.3553, 1.6035, 4.0792, 3.7758], device='cuda:0'), covar=tensor([0.0058, 0.0051, 0.0127, 0.0348, 0.0061, 0.2706, 0.0096, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0126, 0.0171, 0.0161, 0.0144, 0.0182, 0.0160, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:01:36,165 INFO [train.py:904] (0/8) Epoch 13, batch 4600, loss[loss=0.2061, simple_loss=0.2937, pruned_loss=0.05919, over 16711.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2915, pruned_loss=0.05947, over 3215144.55 frames. ], batch size: 134, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:02:00,583 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.3837, 2.6719, 2.6194, 4.0947, 3.1889, 3.9718, 1.5963, 2.9179], device='cuda:0'), covar=tensor([0.1594, 0.0755, 0.1163, 0.0157, 0.0323, 0.0338, 0.1644, 0.0791], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0161, 0.0183, 0.0157, 0.0198, 0.0206, 0.0182, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 19:02:22,727 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126429.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:23,034 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3130, 2.3337, 2.2914, 4.3102, 2.1438, 2.7519, 2.3711, 2.4300], device='cuda:0'), covar=tensor([0.1066, 0.2927, 0.2222, 0.0330, 0.3679, 0.2011, 0.2724, 0.2904], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0405, 0.0336, 0.0319, 0.0419, 0.0470, 0.0369, 0.0473], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:02:26,972 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-29 19:02:38,627 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126441.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:53,236 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:02:55,173 INFO [train.py:904] (0/8) Epoch 13, batch 4650, loss[loss=0.1853, simple_loss=0.27, pruned_loss=0.05027, over 16475.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2899, pruned_loss=0.05921, over 3215706.04 frames. ], batch size: 35, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:03:27,528 INFO [optim.py:368] (0/8) 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 19:03:35,751 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 19:04:07,655 INFO [train.py:904] (0/8) Epoch 13, batch 4700, loss[loss=0.1973, simple_loss=0.2774, pruned_loss=0.05859, over 17041.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.288, pruned_loss=0.05842, over 3203092.99 frames. ], batch size: 41, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:04:21,551 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126511.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:05:13,284 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-29 19:05:21,238 INFO [train.py:904] (0/8) Epoch 13, batch 4750, loss[loss=0.2052, simple_loss=0.2829, pruned_loss=0.06375, over 11936.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2842, pruned_loss=0.05671, over 3195585.66 frames. ], batch size: 248, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:05:53,859 INFO [optim.py:368] (0/8) 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 19:06:32,271 INFO [train.py:904] (0/8) Epoch 13, batch 4800, loss[loss=0.1639, simple_loss=0.2627, pruned_loss=0.03251, over 16813.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2811, pruned_loss=0.05503, over 3198336.75 frames. ], batch size: 83, lr: 5.26e-03, grad_scale: 8.0 2023-04-29 19:07:26,289 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-29 19:07:34,127 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5569, 3.6252, 1.9665, 4.0848, 2.5679, 4.0064, 2.1395, 2.6804], device='cuda:0'), covar=tensor([0.0203, 0.0305, 0.1625, 0.0083, 0.0825, 0.0419, 0.1510, 0.0785], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0166, 0.0189, 0.0138, 0.0167, 0.0209, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:07:43,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5194, 2.3458, 2.2256, 3.4356, 2.2811, 3.7580, 1.3601, 2.7178], device='cuda:0'), covar=tensor([0.1457, 0.0801, 0.1289, 0.0138, 0.0131, 0.0340, 0.1682, 0.0783], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0185, 0.0159, 0.0201, 0.0208, 0.0184, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:07:46,139 INFO [train.py:904] (0/8) Epoch 13, batch 4850, loss[loss=0.2088, simple_loss=0.2911, pruned_loss=0.06329, over 12205.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2815, pruned_loss=0.054, over 3207461.02 frames. ], batch size: 248, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:07:50,110 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=126654.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:08:19,812 INFO [optim.py:368] (0/8) 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 19:08:59,242 INFO [train.py:904] (0/8) Epoch 13, batch 4900, loss[loss=0.2006, simple_loss=0.2742, pruned_loss=0.06347, over 16562.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2808, pruned_loss=0.05304, over 3201631.75 frames. ], batch size: 35, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:08:59,540 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=126702.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:09:22,285 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8829, 5.1803, 4.9141, 4.9713, 4.6378, 4.6463, 4.5742, 5.2497], device='cuda:0'), covar=tensor([0.1010, 0.0779, 0.0898, 0.0688, 0.0752, 0.0840, 0.1077, 0.0904], device='cuda:0'), in_proj_covar=tensor([0.0553, 0.0698, 0.0563, 0.0490, 0.0441, 0.0447, 0.0578, 0.0533], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:09:49,776 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126736.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:10:06,596 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9289, 4.9568, 4.7991, 4.4165, 4.3844, 4.8435, 4.8113, 4.5637], device='cuda:0'), covar=tensor([0.0554, 0.0418, 0.0251, 0.0254, 0.0972, 0.0407, 0.0265, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0249, 0.0338, 0.0298, 0.0279, 0.0319, 0.0322, 0.0203, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:10:12,993 INFO [train.py:904] (0/8) Epoch 13, batch 4950, loss[loss=0.1989, simple_loss=0.2899, pruned_loss=0.05394, over 16393.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2799, pruned_loss=0.05228, over 3209287.54 frames. ], batch size: 146, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:10:39,214 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4266, 4.3758, 4.3420, 3.5898, 4.3658, 1.4115, 4.1295, 4.0308], device='cuda:0'), covar=tensor([0.0075, 0.0080, 0.0127, 0.0407, 0.0093, 0.2652, 0.0121, 0.0211], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0124, 0.0168, 0.0159, 0.0141, 0.0180, 0.0158, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:10:45,733 INFO [optim.py:368] (0/8) 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 19:11:26,274 INFO [train.py:904] (0/8) Epoch 13, batch 5000, loss[loss=0.1758, simple_loss=0.2652, pruned_loss=0.04317, over 16479.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2821, pruned_loss=0.05289, over 3193289.74 frames. ], batch size: 75, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:11:32,734 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=126806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:11:47,336 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=126816.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:12:38,210 INFO [train.py:904] (0/8) Epoch 13, batch 5050, loss[loss=0.1796, simple_loss=0.2717, pruned_loss=0.04371, over 17003.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.2825, pruned_loss=0.05266, over 3184109.15 frames. ], batch size: 41, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:13:11,129 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.161e+02 2.561e+02 3.247e+02 4.827e+02, threshold=5.121e+02, percent-clipped=0.0 2023-04-29 19:13:14,621 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=126877.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:13:51,463 INFO [train.py:904] (0/8) Epoch 13, batch 5100, loss[loss=0.1879, simple_loss=0.2762, pruned_loss=0.04976, over 16430.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2805, pruned_loss=0.05201, over 3180449.87 frames. ], batch size: 146, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:14:02,120 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-29 19:15:04,900 INFO [train.py:904] (0/8) Epoch 13, batch 5150, loss[loss=0.1955, simple_loss=0.2883, pruned_loss=0.05129, over 16862.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2805, pruned_loss=0.05198, over 3164811.09 frames. ], batch size: 109, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:15:14,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9550, 3.4238, 3.5159, 2.3008, 3.1761, 3.4749, 3.2578, 2.1184], device='cuda:0'), covar=tensor([0.0488, 0.0038, 0.0038, 0.0341, 0.0080, 0.0092, 0.0072, 0.0364], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0071, 0.0072, 0.0128, 0.0085, 0.0093, 0.0083, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 19:15:37,791 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.071e+02 2.483e+02 2.841e+02 7.629e+02, threshold=4.965e+02, percent-clipped=2.0 2023-04-29 19:16:05,412 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0471, 1.9562, 2.4389, 3.0393, 2.7884, 3.3810, 1.8831, 3.3944], device='cuda:0'), covar=tensor([0.0139, 0.0432, 0.0320, 0.0211, 0.0251, 0.0116, 0.0489, 0.0097], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0177, 0.0162, 0.0165, 0.0176, 0.0131, 0.0177, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:16:17,748 INFO [train.py:904] (0/8) Epoch 13, batch 5200, loss[loss=0.1857, simple_loss=0.274, pruned_loss=0.0487, over 16332.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2796, pruned_loss=0.05149, over 3167794.05 frames. ], batch size: 165, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:16:25,722 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1005, 2.9716, 3.2282, 1.6916, 3.2898, 3.3715, 2.6822, 2.5392], device='cuda:0'), covar=tensor([0.0805, 0.0205, 0.0127, 0.1159, 0.0071, 0.0124, 0.0362, 0.0477], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0103, 0.0088, 0.0137, 0.0070, 0.0108, 0.0122, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 19:16:58,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4310, 5.7297, 5.3724, 5.4940, 5.0947, 5.0299, 5.1747, 5.7610], device='cuda:0'), covar=tensor([0.1109, 0.0769, 0.1104, 0.0732, 0.0932, 0.0656, 0.0965, 0.0954], device='cuda:0'), in_proj_covar=tensor([0.0562, 0.0705, 0.0571, 0.0498, 0.0447, 0.0451, 0.0586, 0.0543], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:17:08,374 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127036.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:17:31,268 INFO [train.py:904] (0/8) Epoch 13, batch 5250, loss[loss=0.1725, simple_loss=0.2683, pruned_loss=0.03837, over 16926.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2768, pruned_loss=0.05085, over 3183894.75 frames. ], batch size: 96, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:04,189 INFO [optim.py:368] (0/8) 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 19:18:18,855 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127084.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:18:44,213 INFO [train.py:904] (0/8) Epoch 13, batch 5300, loss[loss=0.1687, simple_loss=0.2552, pruned_loss=0.04113, over 16529.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2732, pruned_loss=0.04954, over 3182237.21 frames. ], batch size: 68, lr: 5.25e-03, grad_scale: 8.0 2023-04-29 19:18:50,615 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127106.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:18:56,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 19:19:50,733 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6074, 4.6012, 4.4885, 3.7747, 4.5185, 1.6269, 4.2699, 4.2590], device='cuda:0'), covar=tensor([0.0078, 0.0065, 0.0125, 0.0376, 0.0087, 0.2511, 0.0113, 0.0191], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0125, 0.0170, 0.0162, 0.0143, 0.0183, 0.0160, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:19:58,317 INFO [train.py:904] (0/8) Epoch 13, batch 5350, loss[loss=0.1878, simple_loss=0.2726, pruned_loss=0.05152, over 16637.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2714, pruned_loss=0.04887, over 3187814.67 frames. ], batch size: 57, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:20:01,843 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127154.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:27,745 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127172.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:20:32,663 INFO [optim.py:368] (0/8) 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 19:21:10,978 INFO [train.py:904] (0/8) Epoch 13, batch 5400, loss[loss=0.1998, simple_loss=0.2873, pruned_loss=0.05618, over 12303.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2743, pruned_loss=0.04941, over 3183941.64 frames. ], batch size: 248, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:22:20,876 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127249.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:22:26,317 INFO [train.py:904] (0/8) Epoch 13, batch 5450, loss[loss=0.195, simple_loss=0.2725, pruned_loss=0.05874, over 17186.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2769, pruned_loss=0.05058, over 3187871.04 frames. ], batch size: 44, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:02,327 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.587e+02 2.258e+02 2.675e+02 3.738e+02 1.100e+03, threshold=5.349e+02, percent-clipped=10.0 2023-04-29 19:23:28,934 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-04-29 19:23:43,879 INFO [train.py:904] (0/8) Epoch 13, batch 5500, loss[loss=0.2019, simple_loss=0.2955, pruned_loss=0.05411, over 16804.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2853, pruned_loss=0.05588, over 3168477.67 frames. ], batch size: 83, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:23:57,079 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127310.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:24:43,224 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1367, 3.2175, 1.6901, 3.4748, 2.2973, 3.4357, 2.0604, 2.5198], device='cuda:0'), covar=tensor([0.0265, 0.0352, 0.1689, 0.0163, 0.0788, 0.0550, 0.1362, 0.0711], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0166, 0.0190, 0.0137, 0.0168, 0.0207, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:24:45,404 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4443, 3.4302, 3.4292, 2.8265, 3.3855, 2.0302, 3.1957, 2.8530], device='cuda:0'), covar=tensor([0.0122, 0.0105, 0.0147, 0.0205, 0.0087, 0.1901, 0.0122, 0.0204], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0126, 0.0172, 0.0163, 0.0144, 0.0185, 0.0161, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:25:00,717 INFO [train.py:904] (0/8) Epoch 13, batch 5550, loss[loss=0.2039, simple_loss=0.2835, pruned_loss=0.0621, over 17001.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2935, pruned_loss=0.0619, over 3138542.90 frames. ], batch size: 55, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:25:24,440 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7373, 1.7599, 1.4966, 1.4916, 1.9118, 1.5309, 1.6421, 1.8903], device='cuda:0'), covar=tensor([0.0131, 0.0218, 0.0321, 0.0271, 0.0147, 0.0221, 0.0156, 0.0154], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0213, 0.0205, 0.0207, 0.0212, 0.0212, 0.0218, 0.0204], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:25:31,104 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8964, 2.6535, 2.5874, 1.9016, 2.4933, 2.7136, 2.5979, 1.8798], device='cuda:0'), covar=tensor([0.0335, 0.0053, 0.0054, 0.0302, 0.0102, 0.0091, 0.0079, 0.0326], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0070, 0.0071, 0.0126, 0.0083, 0.0091, 0.0081, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 19:25:38,574 INFO [optim.py:368] (0/8) 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 19:25:54,363 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2296, 3.2611, 3.5715, 1.7851, 3.6881, 3.7539, 2.7747, 2.6599], device='cuda:0'), covar=tensor([0.0800, 0.0224, 0.0161, 0.1137, 0.0062, 0.0153, 0.0391, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0101, 0.0087, 0.0136, 0.0069, 0.0108, 0.0120, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 19:26:20,876 INFO [train.py:904] (0/8) Epoch 13, batch 5600, loss[loss=0.198, simple_loss=0.2894, pruned_loss=0.05327, over 16663.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2987, pruned_loss=0.06643, over 3102692.32 frames. ], batch size: 83, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:27:02,215 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127427.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:27:41,575 INFO [train.py:904] (0/8) Epoch 13, batch 5650, loss[loss=0.2915, simple_loss=0.3496, pruned_loss=0.1167, over 11157.00 frames. ], tot_loss[loss=0.2223, simple_loss=0.3037, pruned_loss=0.07042, over 3070260.49 frames. ], batch size: 246, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:28:12,123 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127472.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:28:17,740 INFO [optim.py:368] (0/8) 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 19:28:36,997 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127488.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:28:58,359 INFO [train.py:904] (0/8) Epoch 13, batch 5700, loss[loss=0.1974, simple_loss=0.2871, pruned_loss=0.05386, over 16626.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3045, pruned_loss=0.07109, over 3082057.25 frames. ], batch size: 62, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:29:03,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4109, 3.4940, 3.2657, 2.9869, 3.0808, 3.3656, 3.2346, 3.1719], device='cuda:0'), covar=tensor([0.0587, 0.0456, 0.0278, 0.0263, 0.0585, 0.0427, 0.1426, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0344, 0.0303, 0.0282, 0.0323, 0.0331, 0.0205, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:29:27,679 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127520.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:29:33,331 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8345, 3.4168, 3.3106, 1.8589, 2.7584, 2.2402, 3.3175, 3.5219], device='cuda:0'), covar=tensor([0.0301, 0.0612, 0.0566, 0.1899, 0.0836, 0.0960, 0.0640, 0.0939], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0150, 0.0161, 0.0146, 0.0140, 0.0127, 0.0139, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:29:41,936 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127529.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:30:18,045 INFO [train.py:904] (0/8) Epoch 13, batch 5750, loss[loss=0.2199, simple_loss=0.302, pruned_loss=0.06893, over 16825.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3068, pruned_loss=0.0729, over 3030828.83 frames. ], batch size: 116, lr: 5.24e-03, grad_scale: 8.0 2023-04-29 19:30:56,349 INFO [optim.py:368] (0/8) 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 19:31:21,259 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127590.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:31:24,439 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-29 19:31:39,686 INFO [train.py:904] (0/8) Epoch 13, batch 5800, loss[loss=0.2382, simple_loss=0.3082, pruned_loss=0.08408, over 11664.00 frames. ], tot_loss[loss=0.2263, simple_loss=0.3071, pruned_loss=0.07276, over 3012220.15 frames. ], batch size: 247, lr: 5.24e-03, grad_scale: 4.0 2023-04-29 19:31:44,432 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127605.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:32:57,629 INFO [train.py:904] (0/8) Epoch 13, batch 5850, loss[loss=0.2172, simple_loss=0.2896, pruned_loss=0.07238, over 11711.00 frames. ], tot_loss[loss=0.2243, simple_loss=0.3056, pruned_loss=0.07152, over 3013835.59 frames. ], batch size: 247, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:33:15,140 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127662.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:33:36,797 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.121e+02 3.118e+02 3.715e+02 4.396e+02 1.301e+03, threshold=7.431e+02, percent-clipped=3.0 2023-04-29 19:34:19,532 INFO [train.py:904] (0/8) Epoch 13, batch 5900, loss[loss=0.2803, simple_loss=0.3304, pruned_loss=0.1151, over 11550.00 frames. ], tot_loss[loss=0.2238, simple_loss=0.3053, pruned_loss=0.07117, over 3026263.75 frames. ], batch size: 248, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:34:30,664 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 19:34:57,891 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127723.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:35:23,271 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-29 19:35:30,573 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.06 vs. limit=5.0 2023-04-29 19:35:42,725 INFO [train.py:904] (0/8) Epoch 13, batch 5950, loss[loss=0.2155, simple_loss=0.3123, pruned_loss=0.05937, over 16884.00 frames. ], tot_loss[loss=0.2225, simple_loss=0.306, pruned_loss=0.0695, over 3052807.21 frames. ], batch size: 96, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:36:04,601 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8533, 5.1571, 4.8808, 4.8855, 4.7028, 4.6306, 4.5123, 5.2525], device='cuda:0'), covar=tensor([0.1042, 0.0801, 0.0969, 0.0768, 0.0718, 0.0876, 0.0983, 0.0830], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0704, 0.0578, 0.0503, 0.0444, 0.0454, 0.0590, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:36:21,833 INFO [optim.py:368] (0/8) 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 19:36:26,301 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127779.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:36:32,795 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127783.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 19:37:04,082 INFO [train.py:904] (0/8) Epoch 13, batch 6000, loss[loss=0.2205, simple_loss=0.3014, pruned_loss=0.0698, over 16715.00 frames. ], tot_loss[loss=0.2217, simple_loss=0.3052, pruned_loss=0.06908, over 3059844.68 frames. ], batch size: 124, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:37:04,083 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 19:37:14,220 INFO [train.py:938] (0/8) Epoch 13, validation: loss=0.1599, simple_loss=0.2726, pruned_loss=0.02359, over 944034.00 frames. 2023-04-29 19:37:14,221 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 19:37:18,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7587, 1.3314, 1.6182, 1.6273, 1.8278, 1.9210, 1.5096, 1.7485], device='cuda:0'), covar=tensor([0.0204, 0.0297, 0.0152, 0.0214, 0.0195, 0.0131, 0.0321, 0.0090], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0177, 0.0160, 0.0163, 0.0174, 0.0130, 0.0176, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 19:38:13,268 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127840.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:32,239 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127850.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:38:35,172 INFO [train.py:904] (0/8) Epoch 13, batch 6050, loss[loss=0.2182, simple_loss=0.3015, pruned_loss=0.06742, over 15347.00 frames. ], tot_loss[loss=0.2193, simple_loss=0.3028, pruned_loss=0.06788, over 3078506.85 frames. ], batch size: 190, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:38:36,627 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5613, 2.7345, 2.3654, 3.9095, 2.8587, 4.0075, 1.3632, 2.8212], device='cuda:0'), covar=tensor([0.1402, 0.0679, 0.1214, 0.0199, 0.0241, 0.0391, 0.1664, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0184, 0.0157, 0.0202, 0.0209, 0.0184, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 19:38:57,291 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-29 19:39:07,428 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=127872.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:14,945 INFO [optim.py:368] (0/8) 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 19:39:28,479 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=127885.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:39:28,609 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4674, 4.4296, 4.3486, 3.6460, 4.3904, 1.7320, 4.1179, 4.0354], device='cuda:0'), covar=tensor([0.0090, 0.0083, 0.0157, 0.0320, 0.0098, 0.2381, 0.0133, 0.0216], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0125, 0.0171, 0.0161, 0.0143, 0.0184, 0.0159, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:39:55,366 INFO [train.py:904] (0/8) Epoch 13, batch 6100, loss[loss=0.2253, simple_loss=0.3059, pruned_loss=0.07234, over 16566.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3028, pruned_loss=0.06732, over 3086725.31 frames. ], batch size: 57, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:40:01,529 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=127905.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:12,556 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127911.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:40:47,067 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=127933.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:15,208 INFO [train.py:904] (0/8) Epoch 13, batch 6150, loss[loss=0.199, simple_loss=0.2854, pruned_loss=0.05624, over 16880.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.3004, pruned_loss=0.06627, over 3090167.18 frames. ], batch size: 116, lr: 5.23e-03, grad_scale: 8.0 2023-04-29 19:41:17,428 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=127953.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:41:25,036 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7128, 1.8190, 2.2621, 2.6131, 2.6335, 3.0140, 1.8827, 2.9855], device='cuda:0'), covar=tensor([0.0165, 0.0378, 0.0261, 0.0235, 0.0215, 0.0129, 0.0404, 0.0097], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0163, 0.0173, 0.0130, 0.0176, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 19:41:48,300 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9526, 2.7456, 2.7479, 2.0420, 2.6527, 2.1666, 2.8363, 2.9271], device='cuda:0'), covar=tensor([0.0273, 0.0714, 0.0489, 0.1635, 0.0726, 0.0853, 0.0557, 0.0749], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0149, 0.0159, 0.0145, 0.0139, 0.0126, 0.0139, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:41:56,818 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.069e+02 3.037e+02 3.583e+02 4.334e+02 1.010e+03, threshold=7.167e+02, percent-clipped=2.0 2023-04-29 19:42:33,255 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-128000.pt 2023-04-29 19:42:39,415 INFO [train.py:904] (0/8) Epoch 13, batch 6200, loss[loss=0.1863, simple_loss=0.2708, pruned_loss=0.05094, over 17247.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.298, pruned_loss=0.06525, over 3104783.04 frames. ], batch size: 52, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:43:05,160 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128018.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:43:44,143 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7512, 4.1372, 3.2344, 2.3200, 2.8826, 2.6111, 4.4693, 3.8875], device='cuda:0'), covar=tensor([0.2734, 0.0712, 0.1520, 0.2348, 0.2242, 0.1716, 0.0434, 0.0970], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0259, 0.0288, 0.0285, 0.0283, 0.0227, 0.0272, 0.0305], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:43:57,675 INFO [train.py:904] (0/8) Epoch 13, batch 6250, loss[loss=0.2002, simple_loss=0.2942, pruned_loss=0.05315, over 16716.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2978, pruned_loss=0.0653, over 3108081.22 frames. ], batch size: 76, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:44:37,188 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.989e+02 2.817e+02 3.684e+02 4.282e+02 1.153e+03, threshold=7.367e+02, percent-clipped=4.0 2023-04-29 19:44:44,568 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128083.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 19:45:13,045 INFO [train.py:904] (0/8) Epoch 13, batch 6300, loss[loss=0.261, simple_loss=0.3236, pruned_loss=0.09924, over 11146.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2974, pruned_loss=0.0645, over 3114611.33 frames. ], batch size: 247, lr: 5.23e-03, grad_scale: 4.0 2023-04-29 19:45:50,774 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.35 vs. limit=5.0 2023-04-29 19:45:53,612 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6821, 4.9254, 4.7205, 4.6811, 4.4215, 4.4255, 4.4385, 5.0046], device='cuda:0'), covar=tensor([0.1023, 0.0874, 0.1005, 0.0793, 0.0794, 0.1098, 0.1006, 0.0899], device='cuda:0'), in_proj_covar=tensor([0.0562, 0.0696, 0.0571, 0.0500, 0.0439, 0.0449, 0.0585, 0.0539], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:46:02,062 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128131.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 19:46:08,188 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128135.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:46:14,627 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0322, 2.4912, 2.6089, 1.8994, 2.7172, 2.8017, 2.4027, 2.3338], device='cuda:0'), covar=tensor([0.0684, 0.0208, 0.0194, 0.0941, 0.0088, 0.0206, 0.0403, 0.0435], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0101, 0.0088, 0.0138, 0.0069, 0.0109, 0.0121, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 19:46:33,881 INFO [train.py:904] (0/8) Epoch 13, batch 6350, loss[loss=0.2175, simple_loss=0.3054, pruned_loss=0.06481, over 16738.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2988, pruned_loss=0.06582, over 3105554.35 frames. ], batch size: 83, lr: 5.22e-03, grad_scale: 4.0 2023-04-29 19:46:59,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1309, 3.7942, 3.7391, 2.3993, 3.3994, 3.6935, 3.4231, 2.2086], device='cuda:0'), covar=tensor([0.0457, 0.0031, 0.0038, 0.0339, 0.0071, 0.0108, 0.0070, 0.0355], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0069, 0.0072, 0.0126, 0.0083, 0.0093, 0.0082, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 19:47:13,677 INFO [optim.py:368] (0/8) 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 19:47:15,498 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0961, 1.9075, 2.5347, 2.9621, 2.9358, 3.3952, 1.8185, 3.3362], device='cuda:0'), covar=tensor([0.0154, 0.0402, 0.0261, 0.0189, 0.0202, 0.0122, 0.0468, 0.0096], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0176, 0.0161, 0.0162, 0.0174, 0.0131, 0.0176, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 19:47:24,819 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128185.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:47:49,937 INFO [train.py:904] (0/8) Epoch 13, batch 6400, loss[loss=0.2068, simple_loss=0.2897, pruned_loss=0.06197, over 15461.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2991, pruned_loss=0.06681, over 3102961.50 frames. ], batch size: 191, lr: 5.22e-03, grad_scale: 8.0 2023-04-29 19:47:56,282 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128206.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:47:59,288 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.84 vs. limit=5.0 2023-04-29 19:48:11,459 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1712, 3.1849, 2.7484, 5.1350, 3.6949, 4.4058, 2.0723, 3.2464], device='cuda:0'), covar=tensor([0.1216, 0.0731, 0.1210, 0.0124, 0.0416, 0.0390, 0.1407, 0.0858], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0162, 0.0181, 0.0155, 0.0200, 0.0207, 0.0183, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 19:48:29,391 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128228.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:35,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4453, 3.2605, 2.6372, 2.0683, 2.2853, 2.1439, 3.3375, 3.0712], device='cuda:0'), covar=tensor([0.2583, 0.0869, 0.1678, 0.2642, 0.2584, 0.1953, 0.0540, 0.1276], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0260, 0.0288, 0.0286, 0.0283, 0.0228, 0.0273, 0.0306], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:48:36,343 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128233.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:48:47,748 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1639, 3.4803, 3.5839, 1.9188, 3.0715, 2.3668, 3.6570, 3.7539], device='cuda:0'), covar=tensor([0.0259, 0.0733, 0.0525, 0.1957, 0.0745, 0.0931, 0.0616, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0149, 0.0160, 0.0145, 0.0139, 0.0126, 0.0139, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:49:02,046 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5625, 3.6005, 2.1549, 4.1013, 2.7156, 4.0224, 2.1947, 2.7883], device='cuda:0'), covar=tensor([0.0226, 0.0340, 0.1511, 0.0145, 0.0698, 0.0508, 0.1497, 0.0753], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0167, 0.0190, 0.0137, 0.0167, 0.0207, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:49:04,035 INFO [train.py:904] (0/8) Epoch 13, batch 6450, loss[loss=0.2319, simple_loss=0.3159, pruned_loss=0.07393, over 15302.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2988, pruned_loss=0.0661, over 3107732.46 frames. ], batch size: 190, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:49:48,885 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.913e+02 2.912e+02 3.353e+02 4.052e+02 7.402e+02, threshold=6.705e+02, percent-clipped=0.0 2023-04-29 19:49:51,455 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.60 vs. limit=5.0 2023-04-29 19:49:59,233 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7866, 1.3701, 1.6669, 1.6275, 1.8548, 1.9369, 1.5299, 1.7884], device='cuda:0'), covar=tensor([0.0186, 0.0304, 0.0152, 0.0218, 0.0184, 0.0133, 0.0315, 0.0093], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0176, 0.0160, 0.0162, 0.0174, 0.0131, 0.0177, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 19:50:21,899 INFO [train.py:904] (0/8) Epoch 13, batch 6500, loss[loss=0.2054, simple_loss=0.2907, pruned_loss=0.06008, over 16772.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2968, pruned_loss=0.06564, over 3107248.88 frames. ], batch size: 83, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:50:27,169 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2006, 3.6665, 3.6112, 2.3917, 3.2893, 3.6011, 3.3861, 2.1921], device='cuda:0'), covar=tensor([0.0467, 0.0037, 0.0047, 0.0336, 0.0085, 0.0106, 0.0074, 0.0372], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0070, 0.0072, 0.0127, 0.0083, 0.0094, 0.0083, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 19:50:42,285 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4410, 3.4928, 1.7247, 3.9877, 2.4119, 3.8887, 1.8956, 2.6985], device='cuda:0'), covar=tensor([0.0245, 0.0352, 0.1939, 0.0171, 0.0862, 0.0469, 0.1837, 0.0809], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0167, 0.0190, 0.0137, 0.0168, 0.0208, 0.0197, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 19:50:45,334 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128318.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:50:47,662 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128320.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:51:03,376 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-29 19:51:39,460 INFO [train.py:904] (0/8) Epoch 13, batch 6550, loss[loss=0.2265, simple_loss=0.3231, pruned_loss=0.06498, over 15496.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2996, pruned_loss=0.06649, over 3107199.98 frames. ], batch size: 190, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:52:01,589 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128366.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:22,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3906, 4.6650, 4.4367, 4.4394, 4.1742, 4.1531, 4.1807, 4.6849], device='cuda:0'), covar=tensor([0.1019, 0.0794, 0.0959, 0.0781, 0.0838, 0.1407, 0.0983, 0.0926], device='cuda:0'), in_proj_covar=tensor([0.0564, 0.0697, 0.0572, 0.0501, 0.0439, 0.0448, 0.0585, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:52:22,963 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.909e+02 2.847e+02 3.449e+02 4.250e+02 8.173e+02, threshold=6.898e+02, percent-clipped=4.0 2023-04-29 19:52:25,047 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128381.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:52:36,105 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8729, 4.9235, 4.7448, 4.4764, 4.3559, 4.8331, 4.7375, 4.5362], device='cuda:0'), covar=tensor([0.0716, 0.0689, 0.0268, 0.0263, 0.0955, 0.0467, 0.0401, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0340, 0.0297, 0.0277, 0.0317, 0.0321, 0.0202, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:52:56,151 INFO [train.py:904] (0/8) Epoch 13, batch 6600, loss[loss=0.2768, simple_loss=0.3286, pruned_loss=0.1125, over 11808.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3021, pruned_loss=0.06696, over 3109312.81 frames. ], batch size: 247, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:53:13,433 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128413.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:53:31,171 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-29 19:53:48,848 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128435.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:14,514 INFO [train.py:904] (0/8) Epoch 13, batch 6650, loss[loss=0.2142, simple_loss=0.3028, pruned_loss=0.06282, over 16767.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.3021, pruned_loss=0.0676, over 3107158.18 frames. ], batch size: 83, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:54:47,971 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:54:56,957 INFO [optim.py:368] (0/8) 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 19:55:02,270 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128483.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:30,393 INFO [train.py:904] (0/8) Epoch 13, batch 6700, loss[loss=0.2012, simple_loss=0.2878, pruned_loss=0.0573, over 16817.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.2995, pruned_loss=0.06662, over 3122228.08 frames. ], batch size: 102, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:55:36,470 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128506.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:55:56,519 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0962, 5.0511, 4.8821, 4.2445, 4.9455, 1.8476, 4.6721, 4.7109], device='cuda:0'), covar=tensor([0.0064, 0.0055, 0.0126, 0.0316, 0.0069, 0.2338, 0.0104, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0124, 0.0170, 0.0160, 0.0142, 0.0184, 0.0159, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:56:10,016 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128528.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:13,743 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-29 19:56:46,051 INFO [train.py:904] (0/8) Epoch 13, batch 6750, loss[loss=0.2021, simple_loss=0.2813, pruned_loss=0.06146, over 17046.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2992, pruned_loss=0.06717, over 3120902.37 frames. ], batch size: 53, lr: 5.22e-03, grad_scale: 2.0 2023-04-29 19:56:49,433 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128554.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:56:55,863 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-29 19:57:22,424 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=128576.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 19:57:28,179 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.231e+02 3.101e+02 3.923e+02 4.752e+02 1.055e+03, threshold=7.847e+02, percent-clipped=2.0 2023-04-29 19:58:01,819 INFO [train.py:904] (0/8) Epoch 13, batch 6800, loss[loss=0.2268, simple_loss=0.302, pruned_loss=0.07585, over 11526.00 frames. ], tot_loss[loss=0.2168, simple_loss=0.2992, pruned_loss=0.06722, over 3103215.85 frames. ], batch size: 246, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:58:51,555 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5191, 3.4835, 3.4899, 2.9036, 3.3899, 2.0161, 3.1484, 2.8212], device='cuda:0'), covar=tensor([0.0123, 0.0108, 0.0154, 0.0244, 0.0089, 0.2096, 0.0133, 0.0204], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0125, 0.0171, 0.0161, 0.0143, 0.0185, 0.0159, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 19:59:19,101 INFO [train.py:904] (0/8) Epoch 13, batch 6850, loss[loss=0.197, simple_loss=0.2985, pruned_loss=0.04773, over 16435.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3007, pruned_loss=0.06759, over 3100525.76 frames. ], batch size: 146, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 19:59:54,929 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128676.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:00:00,832 INFO [optim.py:368] (0/8) 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 20:00:34,564 INFO [train.py:904] (0/8) Epoch 13, batch 6900, loss[loss=0.2246, simple_loss=0.3122, pruned_loss=0.06854, over 16199.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3028, pruned_loss=0.06681, over 3121892.65 frames. ], batch size: 165, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:01:01,448 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-29 20:01:52,397 INFO [train.py:904] (0/8) Epoch 13, batch 6950, loss[loss=0.2174, simple_loss=0.3011, pruned_loss=0.06689, over 16379.00 frames. ], tot_loss[loss=0.2209, simple_loss=0.3046, pruned_loss=0.06855, over 3115062.00 frames. ], batch size: 146, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:02:18,933 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=128769.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:02:36,170 INFO [optim.py:368] (0/8) 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 20:02:38,371 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128781.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:03:08,756 INFO [train.py:904] (0/8) Epoch 13, batch 7000, loss[loss=0.2467, simple_loss=0.3346, pruned_loss=0.07937, over 16277.00 frames. ], tot_loss[loss=0.2213, simple_loss=0.3051, pruned_loss=0.0688, over 3097932.00 frames. ], batch size: 165, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:03:43,270 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2296, 2.0669, 2.0895, 3.9648, 2.0427, 2.5246, 2.1937, 2.2527], device='cuda:0'), covar=tensor([0.1093, 0.3475, 0.2668, 0.0432, 0.3939, 0.2294, 0.3314, 0.3236], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0401, 0.0334, 0.0314, 0.0415, 0.0461, 0.0367, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:04:07,305 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128842.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:04:20,579 INFO [train.py:904] (0/8) Epoch 13, batch 7050, loss[loss=0.2068, simple_loss=0.2979, pruned_loss=0.05782, over 16836.00 frames. ], tot_loss[loss=0.2214, simple_loss=0.3058, pruned_loss=0.06845, over 3096289.58 frames. ], batch size: 116, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:05:02,585 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.996e+02 2.845e+02 3.584e+02 4.333e+02 9.135e+02, threshold=7.167e+02, percent-clipped=3.0 2023-04-29 20:05:19,189 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=128890.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:05:38,304 INFO [train.py:904] (0/8) Epoch 13, batch 7100, loss[loss=0.204, simple_loss=0.2892, pruned_loss=0.05939, over 16599.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3041, pruned_loss=0.06763, over 3116069.86 frames. ], batch size: 62, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:05:47,063 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6900, 4.6646, 5.1430, 5.1176, 5.1034, 4.7723, 4.7599, 4.4987], device='cuda:0'), covar=tensor([0.0246, 0.0453, 0.0303, 0.0327, 0.0402, 0.0306, 0.0802, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0348, 0.0368, 0.0369, 0.0350, 0.0416, 0.0390, 0.0488, 0.0312], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 20:06:56,801 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=128951.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:06:57,455 INFO [train.py:904] (0/8) Epoch 13, batch 7150, loss[loss=0.2236, simple_loss=0.3113, pruned_loss=0.06795, over 16464.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.3017, pruned_loss=0.06704, over 3107143.22 frames. ], batch size: 146, lr: 5.21e-03, grad_scale: 4.0 2023-04-29 20:07:34,037 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=128976.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:07:39,437 INFO [optim.py:368] (0/8) 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 20:08:12,092 INFO [train.py:904] (0/8) Epoch 13, batch 7200, loss[loss=0.1927, simple_loss=0.2752, pruned_loss=0.05511, over 11296.00 frames. ], tot_loss[loss=0.2162, simple_loss=0.2997, pruned_loss=0.06637, over 3066823.33 frames. ], batch size: 247, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:08:44,529 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129024.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:09:29,939 INFO [train.py:904] (0/8) Epoch 13, batch 7250, loss[loss=0.1811, simple_loss=0.2628, pruned_loss=0.0497, over 16990.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2968, pruned_loss=0.06489, over 3067626.29 frames. ], batch size: 55, lr: 5.21e-03, grad_scale: 8.0 2023-04-29 20:09:56,905 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129069.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:10:12,184 INFO [optim.py:368] (0/8) 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 20:10:45,724 INFO [train.py:904] (0/8) Epoch 13, batch 7300, loss[loss=0.2117, simple_loss=0.2979, pruned_loss=0.0628, over 16697.00 frames. ], tot_loss[loss=0.213, simple_loss=0.2965, pruned_loss=0.06474, over 3070393.67 frames. ], batch size: 62, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:11:09,075 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129117.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:11:29,129 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2921, 3.3328, 1.9433, 3.7623, 2.4946, 3.7100, 1.9945, 2.5828], device='cuda:0'), covar=tensor([0.0249, 0.0327, 0.1630, 0.0130, 0.0841, 0.0366, 0.1579, 0.0807], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0165, 0.0190, 0.0136, 0.0169, 0.0207, 0.0197, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 20:11:40,208 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129137.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:11:52,974 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-29 20:12:02,353 INFO [train.py:904] (0/8) Epoch 13, batch 7350, loss[loss=0.1984, simple_loss=0.2862, pruned_loss=0.05533, over 16384.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2975, pruned_loss=0.06555, over 3057546.41 frames. ], batch size: 68, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:12:27,875 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-29 20:12:46,228 INFO [optim.py:368] (0/8) 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 20:13:21,038 INFO [train.py:904] (0/8) Epoch 13, batch 7400, loss[loss=0.2121, simple_loss=0.3084, pruned_loss=0.05788, over 16448.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2991, pruned_loss=0.06601, over 3065738.07 frames. ], batch size: 68, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:14:15,032 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9087, 4.8788, 4.7398, 3.4509, 4.7393, 1.5428, 4.4334, 4.3962], device='cuda:0'), covar=tensor([0.0107, 0.0091, 0.0210, 0.0635, 0.0114, 0.3228, 0.0170, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0122, 0.0167, 0.0159, 0.0140, 0.0182, 0.0157, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:14:32,929 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129246.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:14:41,695 INFO [train.py:904] (0/8) Epoch 13, batch 7450, loss[loss=0.2137, simple_loss=0.3058, pruned_loss=0.06082, over 16369.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3, pruned_loss=0.06714, over 3069270.70 frames. ], batch size: 146, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:15:30,920 INFO [optim.py:368] (0/8) 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 20:16:03,362 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-29 20:16:05,626 INFO [train.py:904] (0/8) Epoch 13, batch 7500, loss[loss=0.2027, simple_loss=0.2903, pruned_loss=0.05755, over 16261.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.301, pruned_loss=0.06658, over 3070216.34 frames. ], batch size: 165, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:16:19,385 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-29 20:17:24,562 INFO [train.py:904] (0/8) Epoch 13, batch 7550, loss[loss=0.1988, simple_loss=0.2908, pruned_loss=0.05341, over 16829.00 frames. ], tot_loss[loss=0.217, simple_loss=0.2999, pruned_loss=0.06704, over 3059697.89 frames. ], batch size: 102, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:17:43,647 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5557, 5.8852, 5.5532, 5.6051, 5.2958, 5.1833, 5.2994, 5.9755], device='cuda:0'), covar=tensor([0.0991, 0.0766, 0.1042, 0.0755, 0.0720, 0.0696, 0.1029, 0.0767], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0702, 0.0576, 0.0500, 0.0442, 0.0455, 0.0587, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:17:58,157 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-04-29 20:18:07,708 INFO [optim.py:368] (0/8) 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 20:18:41,444 INFO [train.py:904] (0/8) Epoch 13, batch 7600, loss[loss=0.2019, simple_loss=0.289, pruned_loss=0.05738, over 16304.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2992, pruned_loss=0.06711, over 3069715.32 frames. ], batch size: 165, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:18:59,719 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129413.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:19:37,560 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129437.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:19:41,481 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-29 20:19:43,600 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.37 vs. limit=5.0 2023-04-29 20:20:00,027 INFO [train.py:904] (0/8) Epoch 13, batch 7650, loss[loss=0.1947, simple_loss=0.2909, pruned_loss=0.04922, over 16835.00 frames. ], tot_loss[loss=0.2179, simple_loss=0.2997, pruned_loss=0.06806, over 3054930.98 frames. ], batch size: 96, lr: 5.20e-03, grad_scale: 8.0 2023-04-29 20:20:12,413 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 20:20:35,827 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129474.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:20:44,166 INFO [optim.py:368] (0/8) 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 20:20:52,540 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129485.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:21:06,953 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3850, 4.6661, 4.4529, 4.4345, 4.1918, 4.1980, 4.2110, 4.6913], device='cuda:0'), covar=tensor([0.1020, 0.0794, 0.0973, 0.0764, 0.0762, 0.1243, 0.0994, 0.0873], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0702, 0.0576, 0.0499, 0.0441, 0.0455, 0.0587, 0.0536], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:21:18,583 INFO [train.py:904] (0/8) Epoch 13, batch 7700, loss[loss=0.2442, simple_loss=0.3076, pruned_loss=0.09036, over 11797.00 frames. ], tot_loss[loss=0.2189, simple_loss=0.3, pruned_loss=0.06887, over 3045448.35 frames. ], batch size: 248, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:21:53,774 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1780, 1.4770, 1.9276, 2.1434, 2.2476, 2.4330, 1.6522, 2.3502], device='cuda:0'), covar=tensor([0.0194, 0.0431, 0.0239, 0.0242, 0.0238, 0.0159, 0.0404, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0174, 0.0158, 0.0162, 0.0173, 0.0128, 0.0175, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-29 20:22:26,987 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=129546.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:22:35,879 INFO [train.py:904] (0/8) Epoch 13, batch 7750, loss[loss=0.2004, simple_loss=0.2986, pruned_loss=0.05116, over 16707.00 frames. ], tot_loss[loss=0.2175, simple_loss=0.2994, pruned_loss=0.06776, over 3064312.14 frames. ], batch size: 89, lr: 5.20e-03, grad_scale: 4.0 2023-04-29 20:23:20,367 INFO [optim.py:368] (0/8) 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 20:23:39,723 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=129594.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:23:52,182 INFO [train.py:904] (0/8) Epoch 13, batch 7800, loss[loss=0.2334, simple_loss=0.3, pruned_loss=0.08341, over 11534.00 frames. ], tot_loss[loss=0.2191, simple_loss=0.3004, pruned_loss=0.06892, over 3040537.86 frames. ], batch size: 247, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:24:09,268 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=129612.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:24:48,712 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7922, 3.6462, 3.8658, 3.5773, 3.7367, 4.2296, 3.9512, 3.6206], device='cuda:0'), covar=tensor([0.1951, 0.2598, 0.2742, 0.2793, 0.3558, 0.1679, 0.1548, 0.2634], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0519, 0.0567, 0.0441, 0.0597, 0.0584, 0.0450, 0.0593], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 20:25:09,862 INFO [train.py:904] (0/8) Epoch 13, batch 7850, loss[loss=0.1928, simple_loss=0.2821, pruned_loss=0.05178, over 16648.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3008, pruned_loss=0.0676, over 3058776.84 frames. ], batch size: 62, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:25:26,126 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2444, 4.2136, 4.4351, 4.2440, 4.3131, 4.8133, 4.3763, 4.1054], device='cuda:0'), covar=tensor([0.1520, 0.2160, 0.2203, 0.2095, 0.2728, 0.1026, 0.1622, 0.2552], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0518, 0.0566, 0.0441, 0.0597, 0.0584, 0.0450, 0.0593], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 20:25:40,344 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129673.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:25:52,271 INFO [optim.py:368] (0/8) 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 20:26:02,574 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3684, 4.6484, 4.4606, 4.4515, 4.1896, 4.1862, 4.2396, 4.6995], device='cuda:0'), covar=tensor([0.1029, 0.0831, 0.0940, 0.0775, 0.0766, 0.1277, 0.0938, 0.0860], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0703, 0.0580, 0.0502, 0.0442, 0.0456, 0.0588, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:26:22,715 INFO [train.py:904] (0/8) Epoch 13, batch 7900, loss[loss=0.2431, simple_loss=0.304, pruned_loss=0.09111, over 11627.00 frames. ], tot_loss[loss=0.2163, simple_loss=0.2994, pruned_loss=0.06663, over 3076325.83 frames. ], batch size: 248, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:27:36,782 INFO [train.py:904] (0/8) Epoch 13, batch 7950, loss[loss=0.2616, simple_loss=0.3258, pruned_loss=0.09868, over 11582.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2997, pruned_loss=0.06674, over 3092317.27 frames. ], batch size: 246, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:28:01,704 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129769.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:28:18,218 INFO [optim.py:368] (0/8) 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 20:28:33,203 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 13, batch 8000, loss[loss=0.2345, simple_loss=0.3118, pruned_loss=0.07856, over 15273.00 frames. ], tot_loss[loss=0.2181, simple_loss=0.3005, pruned_loss=0.06784, over 3079629.09 frames. ], batch size: 190, lr: 5.19e-03, grad_scale: 8.0 2023-04-29 20:30:02,345 INFO [train.py:904] (0/8) Epoch 13, batch 8050, loss[loss=0.2325, simple_loss=0.3067, pruned_loss=0.07914, over 11779.00 frames. ], tot_loss[loss=0.2184, simple_loss=0.3006, pruned_loss=0.06813, over 3068234.60 frames. ], batch size: 248, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:30:02,929 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=129852.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:30:25,238 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 20:30:37,142 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-29 20:30:45,835 INFO [optim.py:368] (0/8) 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 20:30:53,178 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6704, 3.4972, 3.9377, 1.7486, 4.1503, 4.1783, 2.9331, 2.9961], device='cuda:0'), covar=tensor([0.0698, 0.0225, 0.0154, 0.1265, 0.0049, 0.0112, 0.0422, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0101, 0.0089, 0.0138, 0.0069, 0.0109, 0.0121, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 20:31:15,204 INFO [train.py:904] (0/8) Epoch 13, batch 8100, loss[loss=0.2587, simple_loss=0.3154, pruned_loss=0.101, over 11462.00 frames. ], tot_loss[loss=0.2169, simple_loss=0.2994, pruned_loss=0.06714, over 3070067.36 frames. ], batch size: 248, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:29,511 INFO [train.py:904] (0/8) Epoch 13, batch 8150, loss[loss=0.175, simple_loss=0.2652, pruned_loss=0.04237, over 16555.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2961, pruned_loss=0.06489, over 3107702.47 frames. ], batch size: 75, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:32:53,015 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=129968.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:14,454 INFO [optim.py:368] (0/8) 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 20:33:42,725 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-130000.pt 2023-04-29 20:33:46,577 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130000.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:33:48,628 INFO [train.py:904] (0/8) Epoch 13, batch 8200, loss[loss=0.2401, simple_loss=0.2988, pruned_loss=0.09063, over 11477.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2934, pruned_loss=0.06415, over 3112527.18 frames. ], batch size: 247, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:33:49,166 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130002.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:34:58,304 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7939, 4.6935, 4.5295, 3.2660, 4.5123, 1.5898, 4.2241, 4.2517], device='cuda:0'), covar=tensor([0.0119, 0.0112, 0.0223, 0.0688, 0.0140, 0.3149, 0.0189, 0.0377], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0123, 0.0169, 0.0159, 0.0141, 0.0184, 0.0157, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:35:09,272 INFO [train.py:904] (0/8) Epoch 13, batch 8250, loss[loss=0.2054, simple_loss=0.2817, pruned_loss=0.06457, over 11813.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2926, pruned_loss=0.06194, over 3098145.00 frames. ], batch size: 247, lr: 5.19e-03, grad_scale: 4.0 2023-04-29 20:35:24,028 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130061.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:35:27,871 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130063.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 20:35:37,096 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130069.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:35:57,083 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.490e+02 2.705e+02 3.349e+02 4.037e+02 8.257e+02, threshold=6.697e+02, percent-clipped=3.0 2023-04-29 20:36:29,990 INFO [train.py:904] (0/8) Epoch 13, batch 8300, loss[loss=0.2008, simple_loss=0.297, pruned_loss=0.05229, over 16899.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2903, pruned_loss=0.05944, over 3083116.97 frames. ], batch size: 96, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:36:55,413 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130117.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:37:43,814 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130147.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:37:50,800 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9975, 2.0960, 2.2965, 3.2246, 2.1914, 2.3621, 2.2941, 2.1813], device='cuda:0'), covar=tensor([0.0944, 0.3345, 0.2226, 0.0587, 0.4133, 0.2255, 0.3152, 0.3554], device='cuda:0'), in_proj_covar=tensor([0.0362, 0.0398, 0.0332, 0.0312, 0.0411, 0.0454, 0.0361, 0.0462], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:37:51,998 INFO [train.py:904] (0/8) Epoch 13, batch 8350, loss[loss=0.2226, simple_loss=0.2958, pruned_loss=0.07474, over 11896.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2896, pruned_loss=0.05693, over 3092519.86 frames. ], batch size: 246, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:38:39,966 INFO [optim.py:368] (0/8) 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 20:39:12,422 INFO [train.py:904] (0/8) Epoch 13, batch 8400, loss[loss=0.1652, simple_loss=0.258, pruned_loss=0.03615, over 16158.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2867, pruned_loss=0.05494, over 3071300.92 frames. ], batch size: 165, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:29,227 INFO [train.py:904] (0/8) Epoch 13, batch 8450, loss[loss=0.1754, simple_loss=0.2724, pruned_loss=0.03922, over 16377.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2851, pruned_loss=0.05341, over 3073043.39 frames. ], batch size: 146, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:40:32,070 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8120, 4.1762, 3.2615, 2.2743, 2.7850, 2.4902, 4.4958, 3.6180], device='cuda:0'), covar=tensor([0.2559, 0.0561, 0.1397, 0.2537, 0.2667, 0.1826, 0.0325, 0.1060], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0252, 0.0283, 0.0280, 0.0274, 0.0225, 0.0267, 0.0297], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:40:40,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9575, 5.3515, 5.5922, 5.3131, 5.4300, 5.9646, 5.3792, 5.1526], device='cuda:0'), covar=tensor([0.0879, 0.2052, 0.1683, 0.2087, 0.2603, 0.0952, 0.1560, 0.2314], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0499, 0.0545, 0.0428, 0.0576, 0.0570, 0.0438, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 20:40:55,928 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130268.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:41:17,435 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.422e+02 2.811e+02 3.625e+02 1.309e+03, threshold=5.622e+02, percent-clipped=6.0 2023-04-29 20:41:49,391 INFO [train.py:904] (0/8) Epoch 13, batch 8500, loss[loss=0.1744, simple_loss=0.261, pruned_loss=0.04391, over 15311.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2808, pruned_loss=0.05115, over 3046191.95 frames. ], batch size: 190, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:41:59,796 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1728, 4.1724, 4.5815, 4.5680, 4.5528, 4.2737, 4.2506, 4.2023], device='cuda:0'), covar=tensor([0.0336, 0.0656, 0.0403, 0.0393, 0.0504, 0.0409, 0.1059, 0.0484], device='cuda:0'), in_proj_covar=tensor([0.0344, 0.0364, 0.0363, 0.0349, 0.0408, 0.0387, 0.0478, 0.0309], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 20:42:12,544 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:42:45,253 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9860, 4.4236, 4.3697, 3.2320, 3.8521, 4.3720, 3.9970, 2.5774], device='cuda:0'), covar=tensor([0.0333, 0.0024, 0.0024, 0.0248, 0.0061, 0.0051, 0.0043, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0068, 0.0070, 0.0125, 0.0082, 0.0090, 0.0080, 0.0119], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 20:42:47,084 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7375, 3.5408, 3.9109, 1.9520, 4.1197, 4.1971, 3.1430, 3.2096], device='cuda:0'), covar=tensor([0.0648, 0.0220, 0.0172, 0.1125, 0.0042, 0.0094, 0.0323, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0097, 0.0085, 0.0133, 0.0066, 0.0104, 0.0116, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 20:43:10,849 INFO [train.py:904] (0/8) Epoch 13, batch 8550, loss[loss=0.21, simple_loss=0.3031, pruned_loss=0.05844, over 16462.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2787, pruned_loss=0.05012, over 3031476.44 frames. ], batch size: 146, lr: 5.18e-03, grad_scale: 8.0 2023-04-29 20:43:19,683 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:43:23,219 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130358.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 20:43:29,270 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 20:44:07,716 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.521e+02 2.204e+02 2.928e+02 3.716e+02 7.366e+02, threshold=5.857e+02, percent-clipped=6.0 2023-04-29 20:44:12,660 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130384.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:44:50,595 INFO [train.py:904] (0/8) Epoch 13, batch 8600, loss[loss=0.186, simple_loss=0.2842, pruned_loss=0.04388, over 16958.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2792, pruned_loss=0.0492, over 3033687.27 frames. ], batch size: 109, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:45:50,969 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130432.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:03,606 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130438.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:16,907 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130445.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:20,514 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130447.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:46:30,007 INFO [train.py:904] (0/8) Epoch 13, batch 8650, loss[loss=0.2012, simple_loss=0.2786, pruned_loss=0.06189, over 12035.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2767, pruned_loss=0.04756, over 3012905.58 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:47:40,992 INFO [optim.py:368] (0/8) 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 20:47:46,619 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7390, 4.4254, 4.3656, 2.8214, 3.8212, 4.3683, 3.9916, 2.6744], device='cuda:0'), covar=tensor([0.0410, 0.0025, 0.0022, 0.0331, 0.0064, 0.0042, 0.0044, 0.0379], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0068, 0.0070, 0.0126, 0.0082, 0.0090, 0.0080, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 20:48:01,249 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130493.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:04,696 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130495.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:12,937 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130499.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:48:17,495 INFO [train.py:904] (0/8) Epoch 13, batch 8700, loss[loss=0.177, simple_loss=0.2686, pruned_loss=0.04266, over 15454.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2729, pruned_loss=0.04567, over 3020138.42 frames. ], batch size: 190, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:49:02,935 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9985, 2.0108, 2.2927, 3.2202, 2.1563, 2.2705, 2.2237, 2.1478], device='cuda:0'), covar=tensor([0.1040, 0.3826, 0.2421, 0.0603, 0.4209, 0.2688, 0.3384, 0.3587], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0396, 0.0330, 0.0308, 0.0408, 0.0450, 0.0358, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:49:41,405 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8900, 2.1984, 1.8331, 2.0793, 2.5212, 2.2756, 2.6736, 2.7378], device='cuda:0'), covar=tensor([0.0122, 0.0330, 0.0399, 0.0369, 0.0229, 0.0282, 0.0159, 0.0214], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0203, 0.0197, 0.0197, 0.0201, 0.0201, 0.0201, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:49:46,679 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130548.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:49:54,313 INFO [train.py:904] (0/8) Epoch 13, batch 8750, loss[loss=0.1667, simple_loss=0.2554, pruned_loss=0.03904, over 12528.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2723, pruned_loss=0.04522, over 3015095.84 frames. ], batch size: 248, lr: 5.18e-03, grad_scale: 4.0 2023-04-29 20:50:30,135 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=130565.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:51:07,897 INFO [optim.py:368] (0/8) 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 20:51:48,216 INFO [train.py:904] (0/8) Epoch 13, batch 8800, loss[loss=0.1843, simple_loss=0.2696, pruned_loss=0.04951, over 12431.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2715, pruned_loss=0.04419, over 3028829.04 frames. ], batch size: 250, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:52:02,487 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130609.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 20:52:39,171 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=130626.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:32,461 INFO [train.py:904] (0/8) Epoch 13, batch 8850, loss[loss=0.1619, simple_loss=0.2515, pruned_loss=0.03612, over 12426.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2743, pruned_loss=0.04389, over 3029190.18 frames. ], batch size: 246, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:53:41,320 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130656.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:53:45,295 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=130658.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 20:54:37,532 INFO [optim.py:368] (0/8) 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 20:55:17,126 INFO [train.py:904] (0/8) Epoch 13, batch 8900, loss[loss=0.1798, simple_loss=0.273, pruned_loss=0.0433, over 12772.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2753, pruned_loss=0.04359, over 3029274.59 frames. ], batch size: 247, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:55:22,706 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:55:26,422 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=130706.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:55:28,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3018, 2.0888, 2.1878, 4.0027, 2.0641, 2.5466, 2.1803, 2.2604], device='cuda:0'), covar=tensor([0.1023, 0.3384, 0.2538, 0.0396, 0.3974, 0.2253, 0.3378, 0.3269], device='cuda:0'), in_proj_covar=tensor([0.0361, 0.0396, 0.0330, 0.0309, 0.0411, 0.0451, 0.0361, 0.0460], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 20:55:50,496 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5964, 3.5247, 3.8020, 2.1325, 4.0577, 4.1043, 3.0726, 3.0100], device='cuda:0'), covar=tensor([0.0693, 0.0195, 0.0198, 0.0987, 0.0037, 0.0074, 0.0314, 0.0413], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0096, 0.0084, 0.0132, 0.0065, 0.0103, 0.0115, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 20:56:54,229 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130740.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:57:21,726 INFO [train.py:904] (0/8) Epoch 13, batch 8950, loss[loss=0.1613, simple_loss=0.2594, pruned_loss=0.03162, over 16816.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2753, pruned_loss=0.04423, over 3050230.59 frames. ], batch size: 90, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:57:38,157 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0715, 3.3863, 3.3688, 2.3280, 3.1839, 3.4777, 3.3114, 2.0019], device='cuda:0'), covar=tensor([0.0470, 0.0030, 0.0036, 0.0314, 0.0070, 0.0047, 0.0052, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0067, 0.0069, 0.0125, 0.0082, 0.0089, 0.0079, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 20:58:29,464 INFO [optim.py:368] (0/8) 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 20:58:41,749 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130788.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:58:54,563 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130794.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 20:58:54,975 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-04-29 20:59:11,373 INFO [train.py:904] (0/8) Epoch 13, batch 9000, loss[loss=0.1585, simple_loss=0.2489, pruned_loss=0.03409, over 16747.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2722, pruned_loss=0.04282, over 3049014.27 frames. ], batch size: 76, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 20:59:11,374 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 20:59:22,055 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 21:00:41,462 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1823, 4.2475, 4.0461, 3.7804, 3.7479, 4.1773, 3.8586, 3.8856], device='cuda:0'), covar=tensor([0.0568, 0.0566, 0.0284, 0.0272, 0.0760, 0.0444, 0.0739, 0.0600], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0326, 0.0288, 0.0266, 0.0298, 0.0309, 0.0195, 0.0331], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:01:06,020 INFO [train.py:904] (0/8) Epoch 13, batch 9050, loss[loss=0.1786, simple_loss=0.266, pruned_loss=0.04563, over 16311.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2732, pruned_loss=0.04333, over 3061171.23 frames. ], batch size: 146, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:06,316 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5988, 2.0034, 1.6784, 1.8711, 2.3473, 2.0817, 2.2777, 2.5324], device='cuda:0'), covar=tensor([0.0123, 0.0378, 0.0404, 0.0393, 0.0220, 0.0315, 0.0159, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0206, 0.0199, 0.0200, 0.0204, 0.0204, 0.0202, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:02:07,095 INFO [optim.py:368] (0/8) 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 21:02:52,483 INFO [train.py:904] (0/8) Epoch 13, batch 9100, loss[loss=0.1602, simple_loss=0.2543, pruned_loss=0.03304, over 12928.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2727, pruned_loss=0.04369, over 3070856.82 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:02:58,112 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130904.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:03:33,381 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=130921.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:04:49,587 INFO [train.py:904] (0/8) Epoch 13, batch 9150, loss[loss=0.1662, simple_loss=0.2606, pruned_loss=0.03596, over 16832.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2727, pruned_loss=0.04312, over 3071390.57 frames. ], batch size: 124, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:05:52,922 INFO [optim.py:368] (0/8) 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 21:06:24,999 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2214, 2.9862, 3.0735, 1.8130, 3.2937, 3.3778, 2.7387, 2.5711], device='cuda:0'), covar=tensor([0.0717, 0.0235, 0.0215, 0.1035, 0.0070, 0.0133, 0.0396, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0097, 0.0083, 0.0132, 0.0065, 0.0102, 0.0115, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 21:06:31,924 INFO [train.py:904] (0/8) Epoch 13, batch 9200, loss[loss=0.154, simple_loss=0.2396, pruned_loss=0.03419, over 11840.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2686, pruned_loss=0.04226, over 3067744.54 frames. ], batch size: 248, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:06:58,363 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-29 21:07:43,548 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131040.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:08:07,061 INFO [train.py:904] (0/8) Epoch 13, batch 9250, loss[loss=0.1744, simple_loss=0.272, pruned_loss=0.03841, over 16456.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.268, pruned_loss=0.04228, over 3060059.31 frames. ], batch size: 68, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:08:53,924 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-04-29 21:09:01,863 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1096, 2.4778, 2.5814, 1.9131, 2.7461, 2.8423, 2.4417, 2.4110], device='cuda:0'), covar=tensor([0.0678, 0.0201, 0.0175, 0.0946, 0.0084, 0.0155, 0.0441, 0.0411], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0097, 0.0083, 0.0132, 0.0065, 0.0103, 0.0116, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 21:09:11,496 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7503, 2.6337, 2.4678, 3.8586, 2.4997, 3.8995, 1.3472, 3.0070], device='cuda:0'), covar=tensor([0.1411, 0.0692, 0.1164, 0.0127, 0.0124, 0.0400, 0.1701, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0163, 0.0184, 0.0154, 0.0194, 0.0207, 0.0188, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 21:09:12,853 INFO [optim.py:368] (0/8) 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 21:09:24,458 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:24,528 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131088.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:39,366 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131094.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:09:55,044 INFO [train.py:904] (0/8) Epoch 13, batch 9300, loss[loss=0.1589, simple_loss=0.252, pruned_loss=0.03294, over 16704.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2667, pruned_loss=0.0417, over 3068939.88 frames. ], batch size: 134, lr: 5.17e-03, grad_scale: 8.0 2023-04-29 21:10:41,591 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5954, 3.5666, 3.5339, 3.0102, 3.4960, 1.9441, 3.3119, 3.0031], device='cuda:0'), covar=tensor([0.0114, 0.0126, 0.0149, 0.0212, 0.0097, 0.2242, 0.0126, 0.0195], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0119, 0.0162, 0.0149, 0.0135, 0.0181, 0.0152, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:11:11,079 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131136.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:23,563 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:11:40,561 INFO [train.py:904] (0/8) Epoch 13, batch 9350, loss[loss=0.1794, simple_loss=0.2705, pruned_loss=0.04417, over 12251.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2663, pruned_loss=0.04128, over 3073433.20 frames. ], batch size: 248, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:12:25,009 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.118e+02 2.389e+02 2.842e+02 4.349e+02, threshold=4.777e+02, percent-clipped=0.0 2023-04-29 21:13:20,705 INFO [train.py:904] (0/8) Epoch 13, batch 9400, loss[loss=0.1724, simple_loss=0.2512, pruned_loss=0.04683, over 12209.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2662, pruned_loss=0.04096, over 3068717.91 frames. ], batch size: 248, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:13:25,723 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131204.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:13:59,479 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131221.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:14:17,496 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 21:14:27,979 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131235.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 21:14:37,973 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3943, 4.6881, 4.5043, 4.5160, 4.1266, 4.1660, 4.2237, 4.7106], device='cuda:0'), covar=tensor([0.0922, 0.0869, 0.0900, 0.0663, 0.0732, 0.1210, 0.0924, 0.0879], device='cuda:0'), in_proj_covar=tensor([0.0541, 0.0676, 0.0546, 0.0479, 0.0423, 0.0435, 0.0561, 0.0513], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:14:59,716 INFO [train.py:904] (0/8) Epoch 13, batch 9450, loss[loss=0.1641, simple_loss=0.2511, pruned_loss=0.03857, over 16651.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2681, pruned_loss=0.04104, over 3078975.04 frames. ], batch size: 62, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:15:00,217 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131252.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:15:12,773 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-29 21:15:34,985 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131269.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:16:03,433 INFO [optim.py:368] (0/8) 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 21:16:40,131 INFO [train.py:904] (0/8) Epoch 13, batch 9500, loss[loss=0.16, simple_loss=0.2586, pruned_loss=0.0307, over 16859.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2675, pruned_loss=0.0409, over 3080827.89 frames. ], batch size: 102, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:17:01,858 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8394, 3.3513, 3.4414, 2.0352, 2.9658, 2.2318, 3.4127, 3.2806], device='cuda:0'), covar=tensor([0.0291, 0.0609, 0.0462, 0.1711, 0.0655, 0.0911, 0.0586, 0.0881], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0138, 0.0153, 0.0140, 0.0132, 0.0122, 0.0131, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 21:18:25,261 INFO [train.py:904] (0/8) Epoch 13, batch 9550, loss[loss=0.1936, simple_loss=0.2937, pruned_loss=0.0467, over 16262.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2677, pruned_loss=0.04126, over 3085330.11 frames. ], batch size: 165, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:18:50,533 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 21:19:29,581 INFO [optim.py:368] (0/8) 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 21:20:03,687 INFO [train.py:904] (0/8) Epoch 13, batch 9600, loss[loss=0.1696, simple_loss=0.2727, pruned_loss=0.03328, over 16876.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2699, pruned_loss=0.04242, over 3073193.37 frames. ], batch size: 90, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:20:17,929 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131409.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:20:29,750 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131416.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:20:40,939 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-29 21:21:49,842 INFO [train.py:904] (0/8) Epoch 13, batch 9650, loss[loss=0.1844, simple_loss=0.279, pruned_loss=0.04488, over 16691.00 frames. ], tot_loss[loss=0.178, simple_loss=0.271, pruned_loss=0.04246, over 3065725.05 frames. ], batch size: 134, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:22:33,764 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131470.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:22:47,910 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.434e+02 2.934e+02 3.737e+02 7.668e+02, threshold=5.867e+02, percent-clipped=4.0 2023-04-29 21:23:18,254 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131492.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:23:38,324 INFO [train.py:904] (0/8) Epoch 13, batch 9700, loss[loss=0.1783, simple_loss=0.2751, pruned_loss=0.04078, over 16206.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2694, pruned_loss=0.04234, over 3048708.50 frames. ], batch size: 165, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:24:36,258 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131530.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 21:25:20,242 INFO [train.py:904] (0/8) Epoch 13, batch 9750, loss[loss=0.1812, simple_loss=0.2769, pruned_loss=0.04277, over 16313.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2689, pruned_loss=0.04278, over 3032942.05 frames. ], batch size: 146, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:25:22,343 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131553.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:26:22,853 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.333e+02 2.960e+02 3.674e+02 6.318e+02, threshold=5.921e+02, percent-clipped=2.0 2023-04-29 21:26:31,008 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6774, 2.6911, 1.8558, 2.8677, 2.1685, 2.8316, 2.0906, 2.4059], device='cuda:0'), covar=tensor([0.0258, 0.0339, 0.1173, 0.0207, 0.0661, 0.0491, 0.1134, 0.0585], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0160, 0.0186, 0.0130, 0.0165, 0.0197, 0.0194, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-29 21:26:57,825 INFO [train.py:904] (0/8) Epoch 13, batch 9800, loss[loss=0.1674, simple_loss=0.2611, pruned_loss=0.03688, over 16771.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2688, pruned_loss=0.04168, over 3047790.77 frames. ], batch size: 39, lr: 5.16e-03, grad_scale: 8.0 2023-04-29 21:27:30,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3202, 3.4235, 3.6772, 3.6430, 3.6626, 3.4546, 3.5031, 3.5310], device='cuda:0'), covar=tensor([0.0362, 0.0586, 0.0402, 0.0453, 0.0428, 0.0525, 0.0714, 0.0420], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0345, 0.0345, 0.0336, 0.0391, 0.0373, 0.0453, 0.0296], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 21:28:12,823 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-29 21:28:39,807 INFO [train.py:904] (0/8) Epoch 13, batch 9850, loss[loss=0.1753, simple_loss=0.259, pruned_loss=0.04585, over 12735.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2694, pruned_loss=0.04143, over 3044117.92 frames. ], batch size: 248, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:29:46,605 INFO [optim.py:368] (0/8) 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 21:30:32,577 INFO [train.py:904] (0/8) Epoch 13, batch 9900, loss[loss=0.1722, simple_loss=0.2719, pruned_loss=0.03627, over 16312.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2695, pruned_loss=0.04118, over 3040833.04 frames. ], batch size: 146, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:32:30,680 INFO [train.py:904] (0/8) Epoch 13, batch 9950, loss[loss=0.1647, simple_loss=0.264, pruned_loss=0.03268, over 16595.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.272, pruned_loss=0.04163, over 3063566.08 frames. ], batch size: 83, lr: 5.15e-03, grad_scale: 4.0 2023-04-29 21:33:02,013 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131765.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:20,079 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131772.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:33:47,205 INFO [optim.py:368] (0/8) 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 21:34:05,382 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 21:34:31,159 INFO [train.py:904] (0/8) Epoch 13, batch 10000, loss[loss=0.1745, simple_loss=0.2756, pruned_loss=0.03667, over 16742.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2703, pruned_loss=0.04094, over 3082558.26 frames. ], batch size: 76, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:35:26,943 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=131830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:06,302 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=131848.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:11,640 INFO [train.py:904] (0/8) Epoch 13, batch 10050, loss[loss=0.1817, simple_loss=0.2725, pruned_loss=0.04542, over 12311.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2703, pruned_loss=0.04083, over 3064828.97 frames. ], batch size: 250, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:36:22,755 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=131857.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:36:29,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0589, 1.9249, 2.0905, 3.4977, 1.9812, 2.2483, 2.1175, 2.0467], device='cuda:0'), covar=tensor([0.1048, 0.3585, 0.2497, 0.0499, 0.4142, 0.2482, 0.3219, 0.3499], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0389, 0.0328, 0.0306, 0.0405, 0.0441, 0.0354, 0.0450], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:36:44,254 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9618, 3.0385, 3.1154, 2.2038, 2.9110, 3.2421, 3.0761, 1.8587], device='cuda:0'), covar=tensor([0.0442, 0.0047, 0.0039, 0.0301, 0.0099, 0.0066, 0.0071, 0.0429], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0068, 0.0070, 0.0127, 0.0082, 0.0089, 0.0080, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 21:37:04,201 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=131878.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:37:14,102 INFO [optim.py:368] (0/8) 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 21:37:20,964 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0314, 4.0991, 4.4755, 4.4230, 4.4156, 4.1747, 4.1266, 4.1298], device='cuda:0'), covar=tensor([0.0338, 0.0699, 0.0314, 0.0368, 0.0415, 0.0403, 0.0759, 0.0403], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0340, 0.0340, 0.0330, 0.0386, 0.0366, 0.0445, 0.0293], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-29 21:37:45,831 INFO [train.py:904] (0/8) Epoch 13, batch 10100, loss[loss=0.1699, simple_loss=0.2624, pruned_loss=0.03873, over 16781.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2709, pruned_loss=0.04121, over 3065744.28 frames. ], batch size: 76, lr: 5.15e-03, grad_scale: 8.0 2023-04-29 21:38:16,839 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=131918.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:38:55,486 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4792, 3.7870, 3.9042, 2.7897, 3.4281, 3.9184, 3.6905, 2.1463], device='cuda:0'), covar=tensor([0.0414, 0.0032, 0.0028, 0.0290, 0.0086, 0.0068, 0.0054, 0.0422], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0068, 0.0069, 0.0126, 0.0081, 0.0088, 0.0079, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 21:39:04,498 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-13.pt 2023-04-29 21:39:29,405 INFO [train.py:904] (0/8) Epoch 14, batch 0, loss[loss=0.2346, simple_loss=0.2946, pruned_loss=0.08733, over 16832.00 frames. ], tot_loss[loss=0.2346, simple_loss=0.2946, pruned_loss=0.08733, over 16832.00 frames. ], batch size: 102, lr: 4.96e-03, grad_scale: 8.0 2023-04-29 21:39:29,406 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 21:39:36,899 INFO [train.py:938] (0/8) Epoch 14, validation: loss=0.1515, simple_loss=0.2551, pruned_loss=0.024, over 944034.00 frames. 2023-04-29 21:39:36,900 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 21:40:22,374 INFO [optim.py:368] (0/8) 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 21:40:40,375 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-132000.pt 2023-04-29 21:40:46,692 INFO [train.py:904] (0/8) Epoch 14, batch 50, loss[loss=0.1761, simple_loss=0.2586, pruned_loss=0.04677, over 17059.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2805, pruned_loss=0.05855, over 753881.15 frames. ], batch size: 41, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:27,145 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132031.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:41:51,545 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-29 21:41:55,090 INFO [train.py:904] (0/8) Epoch 14, batch 100, loss[loss=0.1859, simple_loss=0.2822, pruned_loss=0.04485, over 17249.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2758, pruned_loss=0.05411, over 1329382.37 frames. ], batch size: 52, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:41:58,464 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132054.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:13,241 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132065.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:23,811 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132072.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:42:44,832 INFO [optim.py:368] (0/8) 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 21:42:45,592 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 2023-04-29 21:42:51,088 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132092.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:04,250 INFO [train.py:904] (0/8) Epoch 14, batch 150, loss[loss=0.1931, simple_loss=0.2645, pruned_loss=0.06083, over 16411.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.274, pruned_loss=0.05272, over 1771716.97 frames. ], batch size: 146, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:43:20,795 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132113.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:23,943 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132115.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:43:30,551 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132120.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:43:46,459 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.05 vs. limit=2.0 2023-04-29 21:44:00,868 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3022, 5.3281, 5.1055, 4.6732, 5.1181, 2.0750, 4.8500, 5.0577], device='cuda:0'), covar=tensor([0.0079, 0.0067, 0.0154, 0.0302, 0.0092, 0.2193, 0.0132, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0124, 0.0166, 0.0152, 0.0139, 0.0187, 0.0157, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:44:08,567 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132148.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:44:15,129 INFO [train.py:904] (0/8) Epoch 14, batch 200, loss[loss=0.2255, simple_loss=0.2853, pruned_loss=0.08284, over 16887.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.274, pruned_loss=0.05416, over 2117285.98 frames. ], batch size: 116, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:44:37,485 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8074, 2.9304, 3.0741, 1.9178, 2.7339, 2.2037, 3.2175, 3.1953], device='cuda:0'), covar=tensor([0.0238, 0.0862, 0.0551, 0.1829, 0.0772, 0.0949, 0.0585, 0.0847], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0143, 0.0157, 0.0145, 0.0136, 0.0125, 0.0135, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 21:45:02,151 INFO [optim.py:368] (0/8) 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 21:45:15,517 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132196.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:45:23,106 INFO [train.py:904] (0/8) Epoch 14, batch 250, loss[loss=0.1784, simple_loss=0.2666, pruned_loss=0.04509, over 16777.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.2723, pruned_loss=0.05339, over 2389910.19 frames. ], batch size: 57, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:45:38,833 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132213.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:46:33,341 INFO [train.py:904] (0/8) Epoch 14, batch 300, loss[loss=0.1829, simple_loss=0.2584, pruned_loss=0.05372, over 16697.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.269, pruned_loss=0.052, over 2596468.48 frames. ], batch size: 89, lr: 4.96e-03, grad_scale: 1.0 2023-04-29 21:47:22,690 INFO [optim.py:368] (0/8) 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 21:47:43,470 INFO [train.py:904] (0/8) Epoch 14, batch 350, loss[loss=0.2113, simple_loss=0.278, pruned_loss=0.07232, over 16683.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2663, pruned_loss=0.05058, over 2766235.38 frames. ], batch size: 76, lr: 4.95e-03, grad_scale: 1.0 2023-04-29 21:48:51,155 INFO [train.py:904] (0/8) Epoch 14, batch 400, loss[loss=0.1537, simple_loss=0.239, pruned_loss=0.03423, over 17235.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2647, pruned_loss=0.05031, over 2899933.26 frames. ], batch size: 44, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:49:38,427 INFO [optim.py:368] (0/8) 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 21:49:39,503 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132387.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:50:00,136 INFO [train.py:904] (0/8) Epoch 14, batch 450, loss[loss=0.1819, simple_loss=0.2488, pruned_loss=0.05755, over 16726.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2626, pruned_loss=0.04893, over 2989289.92 frames. ], batch size: 134, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:50:11,966 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=132410.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 21:50:54,570 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1686, 5.7064, 5.8711, 5.5979, 5.6483, 6.2138, 5.7768, 5.4927], device='cuda:0'), covar=tensor([0.0829, 0.1927, 0.2041, 0.1996, 0.2705, 0.0869, 0.1293, 0.2245], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0519, 0.0568, 0.0439, 0.0597, 0.0591, 0.0449, 0.0585], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 21:51:11,604 INFO [train.py:904] (0/8) Epoch 14, batch 500, loss[loss=0.1764, simple_loss=0.256, pruned_loss=0.04835, over 16842.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.262, pruned_loss=0.04861, over 3069061.02 frames. ], batch size: 96, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:51:22,620 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-29 21:51:28,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9018, 2.9000, 2.6355, 4.2816, 3.5315, 4.1368, 1.7891, 3.0229], device='cuda:0'), covar=tensor([0.1286, 0.0631, 0.1009, 0.0161, 0.0159, 0.0360, 0.1348, 0.0722], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0162, 0.0182, 0.0157, 0.0193, 0.0206, 0.0186, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 21:51:43,084 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0593, 2.5187, 2.6643, 1.8393, 2.7910, 2.8193, 2.4365, 2.3877], device='cuda:0'), covar=tensor([0.0745, 0.0234, 0.0232, 0.0994, 0.0103, 0.0226, 0.0489, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0103, 0.0088, 0.0139, 0.0070, 0.0112, 0.0123, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 21:51:58,584 INFO [optim.py:368] (0/8) 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 21:52:15,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3158, 2.9191, 2.5263, 2.2053, 2.2300, 2.0736, 2.9433, 2.7937], device='cuda:0'), covar=tensor([0.2476, 0.0853, 0.1852, 0.2272, 0.2469, 0.2418, 0.0596, 0.1351], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0258, 0.0289, 0.0284, 0.0274, 0.0229, 0.0271, 0.0304], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:52:19,298 INFO [train.py:904] (0/8) Epoch 14, batch 550, loss[loss=0.2029, simple_loss=0.269, pruned_loss=0.06841, over 16496.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2611, pruned_loss=0.04816, over 3120931.50 frames. ], batch size: 75, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:52:34,276 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132513.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:53:28,147 INFO [train.py:904] (0/8) Epoch 14, batch 600, loss[loss=0.1719, simple_loss=0.247, pruned_loss=0.04834, over 16445.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2611, pruned_loss=0.04823, over 3163405.70 frames. ], batch size: 146, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:53:38,344 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7472, 5.1015, 4.8440, 4.8751, 4.5942, 4.6786, 4.5886, 5.1988], device='cuda:0'), covar=tensor([0.1169, 0.1030, 0.1283, 0.0701, 0.0864, 0.1019, 0.1093, 0.0998], device='cuda:0'), in_proj_covar=tensor([0.0589, 0.0738, 0.0598, 0.0523, 0.0470, 0.0469, 0.0616, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 21:53:41,208 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132561.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:54:17,821 INFO [optim.py:368] (0/8) 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 21:54:24,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7736, 2.4528, 2.2877, 3.4145, 2.7361, 3.6595, 1.4754, 2.7751], device='cuda:0'), covar=tensor([0.1302, 0.0670, 0.1152, 0.0194, 0.0169, 0.0381, 0.1512, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0161, 0.0181, 0.0157, 0.0193, 0.0206, 0.0185, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 21:54:39,153 INFO [train.py:904] (0/8) Epoch 14, batch 650, loss[loss=0.1809, simple_loss=0.2618, pruned_loss=0.05005, over 16222.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2597, pruned_loss=0.04773, over 3195395.69 frames. ], batch size: 164, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:54:39,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4810, 3.5825, 2.0965, 3.7318, 2.7576, 3.7334, 2.1548, 2.8346], device='cuda:0'), covar=tensor([0.0217, 0.0318, 0.1394, 0.0272, 0.0700, 0.0615, 0.1395, 0.0600], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0166, 0.0190, 0.0141, 0.0169, 0.0207, 0.0199, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 21:54:46,533 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6336, 4.6047, 5.0890, 5.0940, 5.0903, 4.7421, 4.7072, 4.5202], device='cuda:0'), covar=tensor([0.0339, 0.0639, 0.0393, 0.0398, 0.0444, 0.0430, 0.0912, 0.0498], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0376, 0.0377, 0.0360, 0.0420, 0.0403, 0.0495, 0.0320], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 21:55:49,517 INFO [train.py:904] (0/8) Epoch 14, batch 700, loss[loss=0.1906, simple_loss=0.2627, pruned_loss=0.05925, over 16854.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2594, pruned_loss=0.04722, over 3220149.36 frames. ], batch size: 90, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:56:01,408 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.258e+02 2.690e+02 3.168e+02 6.172e+02, threshold=5.381e+02, percent-clipped=1.0 2023-04-29 21:56:38,396 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132687.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:56:58,536 INFO [train.py:904] (0/8) Epoch 14, batch 750, loss[loss=0.1756, simple_loss=0.2642, pruned_loss=0.04348, over 17045.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2604, pruned_loss=0.04768, over 3250806.56 frames. ], batch size: 55, lr: 4.95e-03, grad_scale: 2.0 2023-04-29 21:57:05,545 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-29 21:57:09,306 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=132710.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 21:57:21,892 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-29 21:57:24,384 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132721.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:43,140 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132735.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:57:59,124 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=132746.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:07,430 INFO [train.py:904] (0/8) Epoch 14, batch 800, loss[loss=0.1886, simple_loss=0.2567, pruned_loss=0.06024, over 15548.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2593, pruned_loss=0.0469, over 3273910.49 frames. ], batch size: 191, lr: 4.95e-03, grad_scale: 4.0 2023-04-29 21:58:16,057 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=132758.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 21:58:45,559 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 21:58:54,915 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.382e+02 2.716e+02 3.142e+02 6.467e+02, threshold=5.432e+02, percent-clipped=2.0 2023-04-29 21:59:15,854 INFO [train.py:904] (0/8) Epoch 14, batch 850, loss[loss=0.1763, simple_loss=0.2489, pruned_loss=0.05181, over 16848.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2592, pruned_loss=0.04705, over 3287476.81 frames. ], batch size: 109, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 21:59:23,503 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=132807.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:00:24,478 INFO [train.py:904] (0/8) Epoch 14, batch 900, loss[loss=0.1556, simple_loss=0.239, pruned_loss=0.03609, over 17209.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2582, pruned_loss=0.04631, over 3299634.43 frames. ], batch size: 45, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:00:28,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5172, 4.4935, 4.4593, 3.9275, 4.4627, 1.7833, 4.1881, 4.0557], device='cuda:0'), covar=tensor([0.0102, 0.0086, 0.0145, 0.0298, 0.0092, 0.2512, 0.0132, 0.0193], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0130, 0.0174, 0.0161, 0.0145, 0.0191, 0.0164, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:01:14,374 INFO [optim.py:368] (0/8) 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 22:01:15,245 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-04-29 22:01:34,141 INFO [train.py:904] (0/8) Epoch 14, batch 950, loss[loss=0.1576, simple_loss=0.2412, pruned_loss=0.03703, over 16772.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2576, pruned_loss=0.04589, over 3300025.32 frames. ], batch size: 39, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:02:13,918 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-04-29 22:02:36,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9818, 4.3453, 4.3607, 3.2363, 3.6642, 4.2768, 3.9799, 2.5323], device='cuda:0'), covar=tensor([0.0365, 0.0050, 0.0030, 0.0286, 0.0091, 0.0077, 0.0068, 0.0365], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0073, 0.0074, 0.0129, 0.0085, 0.0094, 0.0083, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:02:41,762 INFO [train.py:904] (0/8) Epoch 14, batch 1000, loss[loss=0.1848, simple_loss=0.2569, pruned_loss=0.05639, over 16924.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2571, pruned_loss=0.04596, over 3301227.63 frames. ], batch size: 96, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:03:29,372 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.118e+02 2.524e+02 3.195e+02 6.689e+02, threshold=5.047e+02, percent-clipped=1.0 2023-04-29 22:03:50,009 INFO [train.py:904] (0/8) Epoch 14, batch 1050, loss[loss=0.1762, simple_loss=0.2525, pruned_loss=0.04998, over 16871.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2568, pruned_loss=0.046, over 3307663.01 frames. ], batch size: 90, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:04:10,803 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133016.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:04:59,798 INFO [train.py:904] (0/8) Epoch 14, batch 1100, loss[loss=0.1628, simple_loss=0.2376, pruned_loss=0.04402, over 15618.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2563, pruned_loss=0.04563, over 3319945.63 frames. ], batch size: 190, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:05:23,001 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7460, 2.9589, 2.7829, 4.8954, 3.8716, 4.4257, 1.6916, 3.2417], device='cuda:0'), covar=tensor([0.1488, 0.0790, 0.1215, 0.0194, 0.0293, 0.0399, 0.1658, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0162, 0.0182, 0.0159, 0.0195, 0.0207, 0.0185, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 22:05:29,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7140, 2.8888, 2.8610, 4.9034, 3.9431, 4.3919, 1.7391, 3.2657], device='cuda:0'), covar=tensor([0.1378, 0.0764, 0.1053, 0.0175, 0.0240, 0.0379, 0.1482, 0.0697], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0162, 0.0182, 0.0159, 0.0195, 0.0207, 0.0185, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 22:05:47,410 INFO [optim.py:368] (0/8) 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 22:06:05,195 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1364, 5.6358, 5.7983, 5.5572, 5.6892, 6.2355, 5.8053, 5.5160], device='cuda:0'), covar=tensor([0.0862, 0.1868, 0.1898, 0.2042, 0.2593, 0.0897, 0.1393, 0.2273], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0528, 0.0581, 0.0451, 0.0612, 0.0602, 0.0460, 0.0600], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:06:08,351 INFO [train.py:904] (0/8) Epoch 14, batch 1150, loss[loss=0.1751, simple_loss=0.2583, pruned_loss=0.04601, over 17027.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2553, pruned_loss=0.04536, over 3322934.21 frames. ], batch size: 41, lr: 4.94e-03, grad_scale: 4.0 2023-04-29 22:06:08,679 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133102.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:07:16,496 INFO [train.py:904] (0/8) Epoch 14, batch 1200, loss[loss=0.1487, simple_loss=0.2301, pruned_loss=0.03367, over 16780.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2547, pruned_loss=0.04504, over 3312877.85 frames. ], batch size: 39, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:08:05,979 INFO [optim.py:368] (0/8) 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 22:08:28,188 INFO [train.py:904] (0/8) Epoch 14, batch 1250, loss[loss=0.194, simple_loss=0.2601, pruned_loss=0.06397, over 16843.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2551, pruned_loss=0.04527, over 3320402.98 frames. ], batch size: 116, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:09:32,828 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1352, 2.1000, 2.5908, 2.9697, 2.8210, 3.4004, 2.3995, 3.3528], device='cuda:0'), covar=tensor([0.0189, 0.0397, 0.0268, 0.0260, 0.0282, 0.0170, 0.0362, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0163, 0.0166, 0.0175, 0.0133, 0.0178, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:09:37,986 INFO [train.py:904] (0/8) Epoch 14, batch 1300, loss[loss=0.1725, simple_loss=0.2489, pruned_loss=0.04807, over 15477.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2548, pruned_loss=0.0453, over 3313945.37 frames. ], batch size: 191, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:10:27,127 INFO [optim.py:368] (0/8) 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 22:10:48,355 INFO [train.py:904] (0/8) Epoch 14, batch 1350, loss[loss=0.1741, simple_loss=0.2499, pruned_loss=0.04917, over 16901.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2556, pruned_loss=0.04527, over 3320750.38 frames. ], batch size: 96, lr: 4.94e-03, grad_scale: 8.0 2023-04-29 22:11:06,812 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133316.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:11:28,716 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1076, 4.7986, 5.0736, 5.3125, 5.5142, 4.7787, 5.4544, 5.4270], device='cuda:0'), covar=tensor([0.1526, 0.1222, 0.1627, 0.0634, 0.0464, 0.0769, 0.0443, 0.0497], device='cuda:0'), in_proj_covar=tensor([0.0581, 0.0717, 0.0861, 0.0735, 0.0552, 0.0561, 0.0577, 0.0676], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:11:33,344 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6654, 4.7015, 5.1087, 5.1089, 5.1024, 4.7721, 4.7422, 4.5158], device='cuda:0'), covar=tensor([0.0339, 0.0794, 0.0411, 0.0444, 0.0514, 0.0470, 0.0951, 0.0547], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0389, 0.0388, 0.0370, 0.0431, 0.0415, 0.0506, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 22:11:56,459 INFO [train.py:904] (0/8) Epoch 14, batch 1400, loss[loss=0.1465, simple_loss=0.2281, pruned_loss=0.03242, over 16783.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2554, pruned_loss=0.04499, over 3322322.17 frames. ], batch size: 39, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:12:12,800 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133364.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:26,510 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133374.0, num_to_drop=1, layers_to_drop={1} 2023-04-29 22:12:29,460 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133376.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:12:44,805 INFO [optim.py:368] (0/8) 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 22:13:06,253 INFO [train.py:904] (0/8) Epoch 14, batch 1450, loss[loss=0.191, simple_loss=0.2666, pruned_loss=0.05773, over 16511.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2542, pruned_loss=0.04455, over 3325155.02 frames. ], batch size: 68, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:13:06,588 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=133402.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:13:52,407 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133435.0, num_to_drop=1, layers_to_drop={2} 2023-04-29 22:13:54,594 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=133450.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:14:14,819 INFO [train.py:904] (0/8) Epoch 14, batch 1500, loss[loss=0.1903, simple_loss=0.2824, pruned_loss=0.04911, over 17107.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2545, pruned_loss=0.04444, over 3319761.50 frames. ], batch size: 48, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:14:46,586 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133474.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:15:03,758 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.705e+02 2.267e+02 2.734e+02 3.162e+02 6.055e+02, threshold=5.467e+02, percent-clipped=3.0 2023-04-29 22:15:23,313 INFO [train.py:904] (0/8) Epoch 14, batch 1550, loss[loss=0.2032, simple_loss=0.281, pruned_loss=0.06271, over 16273.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2567, pruned_loss=0.04598, over 3322759.01 frames. ], batch size: 165, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:15:58,531 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-29 22:16:10,824 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133535.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:16:33,561 INFO [train.py:904] (0/8) Epoch 14, batch 1600, loss[loss=0.2026, simple_loss=0.2974, pruned_loss=0.05387, over 16606.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2582, pruned_loss=0.0463, over 3326743.07 frames. ], batch size: 62, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:17:21,878 INFO [optim.py:368] (0/8) 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 22:17:43,028 INFO [train.py:904] (0/8) Epoch 14, batch 1650, loss[loss=0.189, simple_loss=0.2624, pruned_loss=0.05783, over 16721.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2597, pruned_loss=0.04695, over 3321123.35 frames. ], batch size: 124, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:28,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5184, 3.6939, 3.8669, 2.1891, 3.1492, 2.5817, 3.8052, 3.8671], device='cuda:0'), covar=tensor([0.0273, 0.0779, 0.0560, 0.1890, 0.0793, 0.0981, 0.0642, 0.1021], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0149, 0.0158, 0.0145, 0.0137, 0.0125, 0.0136, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 22:18:29,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8923, 2.3017, 2.2547, 2.8169, 2.2382, 3.2122, 1.7666, 2.6611], device='cuda:0'), covar=tensor([0.1129, 0.0598, 0.1032, 0.0144, 0.0126, 0.0359, 0.1320, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0162, 0.0181, 0.0160, 0.0196, 0.0208, 0.0185, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 22:18:31,259 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 22:18:52,512 INFO [train.py:904] (0/8) Epoch 14, batch 1700, loss[loss=0.1807, simple_loss=0.2553, pruned_loss=0.05302, over 16868.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2606, pruned_loss=0.04734, over 3329182.03 frames. ], batch size: 116, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:18:53,007 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133652.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:19:42,413 INFO [optim.py:368] (0/8) 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 22:20:03,032 INFO [train.py:904] (0/8) Epoch 14, batch 1750, loss[loss=0.1773, simple_loss=0.2732, pruned_loss=0.04067, over 16655.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2618, pruned_loss=0.04748, over 3322887.98 frames. ], batch size: 62, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:20:12,016 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8011, 5.0103, 5.1859, 4.9420, 4.9946, 5.5990, 5.1949, 4.8718], device='cuda:0'), covar=tensor([0.1301, 0.1998, 0.1988, 0.2073, 0.2716, 0.1051, 0.1652, 0.2843], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0537, 0.0587, 0.0454, 0.0615, 0.0609, 0.0464, 0.0610], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:20:18,790 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=133713.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:20:42,448 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133730.0, num_to_drop=1, layers_to_drop={3} 2023-04-29 22:20:44,855 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133732.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:21:12,614 INFO [train.py:904] (0/8) Epoch 14, batch 1800, loss[loss=0.1893, simple_loss=0.2659, pruned_loss=0.05631, over 16829.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2615, pruned_loss=0.04705, over 3316575.76 frames. ], batch size: 102, lr: 4.93e-03, grad_scale: 8.0 2023-04-29 22:21:50,717 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-29 22:22:00,853 INFO [optim.py:368] (0/8) 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 22:22:04,673 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-29 22:22:22,390 INFO [train.py:904] (0/8) Epoch 14, batch 1850, loss[loss=0.1845, simple_loss=0.281, pruned_loss=0.04399, over 16750.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.262, pruned_loss=0.04685, over 3316687.95 frames. ], batch size: 62, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:23:03,001 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=133830.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:23:32,965 INFO [train.py:904] (0/8) Epoch 14, batch 1900, loss[loss=0.1526, simple_loss=0.2313, pruned_loss=0.03699, over 17000.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2616, pruned_loss=0.04666, over 3314446.39 frames. ], batch size: 41, lr: 4.93e-03, grad_scale: 4.0 2023-04-29 22:24:23,481 INFO [optim.py:368] (0/8) 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 22:24:25,217 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2230, 2.0615, 2.2850, 3.8951, 2.0757, 2.4624, 2.1377, 2.2687], device='cuda:0'), covar=tensor([0.1175, 0.3436, 0.2439, 0.0540, 0.3615, 0.2355, 0.3475, 0.2910], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0412, 0.0345, 0.0327, 0.0421, 0.0475, 0.0378, 0.0483], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:24:42,163 INFO [train.py:904] (0/8) Epoch 14, batch 1950, loss[loss=0.1942, simple_loss=0.2726, pruned_loss=0.05788, over 16752.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2622, pruned_loss=0.04642, over 3315703.11 frames. ], batch size: 124, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:24:54,620 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8132, 4.7912, 5.2674, 5.2694, 5.2830, 4.9220, 4.8857, 4.6904], device='cuda:0'), covar=tensor([0.0267, 0.0459, 0.0395, 0.0391, 0.0407, 0.0349, 0.0870, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0394, 0.0396, 0.0378, 0.0438, 0.0421, 0.0515, 0.0333], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 22:25:40,134 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1263, 3.9528, 4.3214, 2.2478, 4.5751, 4.6098, 3.2260, 3.6445], device='cuda:0'), covar=tensor([0.0621, 0.0204, 0.0228, 0.1047, 0.0061, 0.0146, 0.0402, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0137, 0.0070, 0.0112, 0.0121, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 22:25:52,917 INFO [train.py:904] (0/8) Epoch 14, batch 2000, loss[loss=0.1571, simple_loss=0.245, pruned_loss=0.03454, over 17103.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2621, pruned_loss=0.04641, over 3319986.85 frames. ], batch size: 47, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:26:43,622 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.687e+02 2.429e+02 2.777e+02 3.533e+02 6.577e+02, threshold=5.554e+02, percent-clipped=5.0 2023-04-29 22:26:57,048 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=133997.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:01,080 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-134000.pt 2023-04-29 22:27:06,910 INFO [train.py:904] (0/8) Epoch 14, batch 2050, loss[loss=0.1717, simple_loss=0.2598, pruned_loss=0.04176, over 16526.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2621, pruned_loss=0.04656, over 3308933.78 frames. ], batch size: 68, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:27:15,549 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:27:24,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0309, 5.5495, 5.7086, 5.4731, 5.4956, 6.0943, 5.6680, 5.4350], device='cuda:0'), covar=tensor([0.0944, 0.1837, 0.2018, 0.1989, 0.2679, 0.0912, 0.1366, 0.2256], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0538, 0.0587, 0.0456, 0.0615, 0.0608, 0.0463, 0.0612], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:27:46,431 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134030.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:27:49,373 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134032.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:28:16,829 INFO [train.py:904] (0/8) Epoch 14, batch 2100, loss[loss=0.1525, simple_loss=0.2359, pruned_loss=0.03457, over 16787.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2623, pruned_loss=0.04669, over 3314308.03 frames. ], batch size: 39, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:28:25,605 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134078.0, num_to_drop=1, layers_to_drop={0} 2023-04-29 22:28:56,312 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134080.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:29:09,013 INFO [optim.py:368] (0/8) 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 22:29:15,402 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 22:29:26,683 INFO [train.py:904] (0/8) Epoch 14, batch 2150, loss[loss=0.2275, simple_loss=0.2973, pruned_loss=0.07882, over 12193.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2637, pruned_loss=0.048, over 3295574.56 frames. ], batch size: 246, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:29:28,104 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8888, 4.8551, 4.7187, 4.2097, 4.8114, 1.9220, 4.5460, 4.5018], device='cuda:0'), covar=tensor([0.0120, 0.0082, 0.0180, 0.0341, 0.0091, 0.2506, 0.0142, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0135, 0.0182, 0.0169, 0.0153, 0.0195, 0.0172, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:30:00,801 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7118, 2.6406, 2.1522, 2.5301, 2.9810, 2.8465, 3.4445, 3.2167], device='cuda:0'), covar=tensor([0.0096, 0.0351, 0.0468, 0.0371, 0.0240, 0.0289, 0.0207, 0.0210], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0221, 0.0213, 0.0212, 0.0219, 0.0220, 0.0227, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:30:06,052 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134130.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:30:34,472 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-04-29 22:30:35,966 INFO [train.py:904] (0/8) Epoch 14, batch 2200, loss[loss=0.1887, simple_loss=0.2717, pruned_loss=0.05281, over 16414.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2648, pruned_loss=0.04847, over 3305105.62 frames. ], batch size: 68, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:31:12,756 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134178.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:31:26,216 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9544, 5.2508, 5.4825, 5.2392, 5.2952, 5.9020, 5.4497, 5.1438], device='cuda:0'), covar=tensor([0.0945, 0.1843, 0.2127, 0.1774, 0.2644, 0.0960, 0.1260, 0.2251], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0539, 0.0590, 0.0456, 0.0619, 0.0613, 0.0467, 0.0610], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:31:27,738 INFO [optim.py:368] (0/8) 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 22:31:28,459 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-04-29 22:31:45,991 INFO [train.py:904] (0/8) Epoch 14, batch 2250, loss[loss=0.1686, simple_loss=0.2485, pruned_loss=0.04432, over 16764.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2664, pruned_loss=0.04916, over 3297285.97 frames. ], batch size: 83, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:32:10,077 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 22:32:56,454 INFO [train.py:904] (0/8) Epoch 14, batch 2300, loss[loss=0.1522, simple_loss=0.2361, pruned_loss=0.03415, over 16813.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2659, pruned_loss=0.0489, over 3286303.48 frames. ], batch size: 39, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:33:48,102 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.417e+02 2.438e+02 2.937e+02 3.292e+02 7.337e+02, threshold=5.875e+02, percent-clipped=1.0 2023-04-29 22:33:48,506 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5962, 4.1390, 4.2205, 2.8922, 3.6226, 4.2342, 3.8572, 2.4871], device='cuda:0'), covar=tensor([0.0446, 0.0061, 0.0033, 0.0329, 0.0089, 0.0073, 0.0065, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0073, 0.0074, 0.0128, 0.0085, 0.0094, 0.0083, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:33:54,817 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4789, 3.9968, 4.1162, 2.7580, 3.5575, 4.0888, 3.7432, 2.2862], device='cuda:0'), covar=tensor([0.0459, 0.0106, 0.0047, 0.0337, 0.0107, 0.0087, 0.0079, 0.0410], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0073, 0.0073, 0.0128, 0.0085, 0.0094, 0.0083, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:34:06,357 INFO [train.py:904] (0/8) Epoch 14, batch 2350, loss[loss=0.1909, simple_loss=0.2788, pruned_loss=0.05149, over 17083.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2665, pruned_loss=0.04902, over 3293594.52 frames. ], batch size: 55, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:34:09,697 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134304.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:15,157 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134308.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:16,995 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134309.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:34:29,526 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5500, 3.4059, 3.6936, 1.8713, 3.7964, 3.7740, 3.1173, 2.9004], device='cuda:0'), covar=tensor([0.0666, 0.0176, 0.0126, 0.1065, 0.0063, 0.0163, 0.0345, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0102, 0.0089, 0.0136, 0.0070, 0.0112, 0.0121, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-29 22:34:40,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8396, 3.0759, 2.6090, 4.4220, 3.6378, 4.2348, 1.6631, 2.9780], device='cuda:0'), covar=tensor([0.1296, 0.0572, 0.1039, 0.0163, 0.0197, 0.0381, 0.1383, 0.0780], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0181, 0.0161, 0.0197, 0.0210, 0.0185, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-29 22:35:17,013 INFO [train.py:904] (0/8) Epoch 14, batch 2400, loss[loss=0.1883, simple_loss=0.2779, pruned_loss=0.04933, over 16544.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2672, pruned_loss=0.04939, over 3300776.13 frames. ], batch size: 68, lr: 4.92e-03, grad_scale: 8.0 2023-04-29 22:35:19,214 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134353.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:23,355 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134356.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:35,613 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134365.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:35:42,610 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134370.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:36:08,841 INFO [optim.py:368] (0/8) 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 22:36:25,727 INFO [train.py:904] (0/8) Epoch 14, batch 2450, loss[loss=0.1901, simple_loss=0.2699, pruned_loss=0.0552, over 16763.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2679, pruned_loss=0.04904, over 3314903.75 frames. ], batch size: 83, lr: 4.92e-03, grad_scale: 4.0 2023-04-29 22:36:46,322 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7269, 4.3136, 3.9006, 4.7760, 4.8792, 4.4743, 4.9635, 4.9540], device='cuda:0'), covar=tensor([0.1409, 0.1663, 0.3819, 0.1751, 0.1363, 0.1443, 0.1450, 0.1620], device='cuda:0'), in_proj_covar=tensor([0.0605, 0.0747, 0.0899, 0.0767, 0.0575, 0.0593, 0.0606, 0.0705], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:37:09,678 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-29 22:37:16,820 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6658, 4.6193, 4.5280, 4.2648, 4.2413, 4.6183, 4.4102, 4.3468], device='cuda:0'), covar=tensor([0.0570, 0.0593, 0.0246, 0.0259, 0.0749, 0.0408, 0.0448, 0.0577], device='cuda:0'), in_proj_covar=tensor([0.0271, 0.0371, 0.0328, 0.0308, 0.0343, 0.0355, 0.0223, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:37:35,062 INFO [train.py:904] (0/8) Epoch 14, batch 2500, loss[loss=0.1987, simple_loss=0.2719, pruned_loss=0.06274, over 16763.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2675, pruned_loss=0.04871, over 3325608.69 frames. ], batch size: 124, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:38:28,208 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.634e+02 2.356e+02 2.745e+02 3.374e+02 4.761e+02, threshold=5.489e+02, percent-clipped=0.0 2023-04-29 22:38:45,448 INFO [train.py:904] (0/8) Epoch 14, batch 2550, loss[loss=0.2102, simple_loss=0.2858, pruned_loss=0.06731, over 16361.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2678, pruned_loss=0.04935, over 3320200.85 frames. ], batch size: 146, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:39:55,159 INFO [train.py:904] (0/8) Epoch 14, batch 2600, loss[loss=0.1864, simple_loss=0.2697, pruned_loss=0.0516, over 16910.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2675, pruned_loss=0.04897, over 3316927.45 frames. ], batch size: 90, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:40:46,904 INFO [optim.py:368] (0/8) 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 22:41:03,703 INFO [train.py:904] (0/8) Epoch 14, batch 2650, loss[loss=0.1756, simple_loss=0.2763, pruned_loss=0.03749, over 17029.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2676, pruned_loss=0.04781, over 3326927.76 frames. ], batch size: 55, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:41:22,123 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7020, 3.8338, 2.2816, 4.1582, 2.9403, 4.1546, 2.3148, 3.0859], device='cuda:0'), covar=tensor([0.0258, 0.0322, 0.1478, 0.0294, 0.0700, 0.0574, 0.1410, 0.0612], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0173, 0.0192, 0.0149, 0.0171, 0.0215, 0.0201, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 22:42:06,658 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8914, 4.0432, 2.4094, 4.6833, 3.2122, 4.5582, 2.4163, 3.3171], device='cuda:0'), covar=tensor([0.0259, 0.0341, 0.1521, 0.0195, 0.0718, 0.0482, 0.1483, 0.0647], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0173, 0.0192, 0.0149, 0.0171, 0.0215, 0.0201, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 22:42:06,905 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-29 22:42:12,178 INFO [train.py:904] (0/8) Epoch 14, batch 2700, loss[loss=0.162, simple_loss=0.2511, pruned_loss=0.03641, over 16502.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2672, pruned_loss=0.04733, over 3330404.79 frames. ], batch size: 68, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:42:13,631 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134653.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:23,629 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134660.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:42:31,232 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=134665.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:43:04,248 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.644e+02 2.219e+02 2.559e+02 3.067e+02 4.567e+02, threshold=5.119e+02, percent-clipped=0.0 2023-04-29 22:43:20,429 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=134701.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:43:21,291 INFO [train.py:904] (0/8) Epoch 14, batch 2750, loss[loss=0.1717, simple_loss=0.2638, pruned_loss=0.03981, over 17238.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2674, pruned_loss=0.04636, over 3333495.50 frames. ], batch size: 45, lr: 4.91e-03, grad_scale: 4.0 2023-04-29 22:44:29,019 INFO [train.py:904] (0/8) Epoch 14, batch 2800, loss[loss=0.2004, simple_loss=0.2923, pruned_loss=0.05429, over 16774.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2679, pruned_loss=0.04709, over 3328208.76 frames. ], batch size: 57, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:45:11,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9471, 5.3423, 5.0549, 5.0794, 4.8397, 4.7347, 4.7844, 5.4190], device='cuda:0'), covar=tensor([0.1190, 0.0865, 0.1111, 0.0745, 0.0834, 0.0979, 0.1075, 0.0980], device='cuda:0'), in_proj_covar=tensor([0.0611, 0.0761, 0.0620, 0.0547, 0.0485, 0.0486, 0.0635, 0.0588], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:45:20,159 INFO [optim.py:368] (0/8) 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 22:45:37,185 INFO [train.py:904] (0/8) Epoch 14, batch 2850, loss[loss=0.1435, simple_loss=0.2302, pruned_loss=0.02844, over 16967.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2677, pruned_loss=0.04722, over 3329088.86 frames. ], batch size: 41, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:18,534 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5460, 4.5452, 4.7218, 4.4936, 4.5580, 5.1759, 4.7390, 4.4298], device='cuda:0'), covar=tensor([0.1613, 0.2264, 0.2444, 0.2214, 0.3009, 0.1241, 0.1471, 0.2672], device='cuda:0'), in_proj_covar=tensor([0.0382, 0.0541, 0.0593, 0.0462, 0.0622, 0.0619, 0.0469, 0.0616], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 22:46:45,308 INFO [train.py:904] (0/8) Epoch 14, batch 2900, loss[loss=0.1713, simple_loss=0.248, pruned_loss=0.04731, over 16853.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2666, pruned_loss=0.04783, over 3317317.20 frames. ], batch size: 90, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:46:58,784 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=134862.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:47:36,338 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.583e+02 2.493e+02 2.912e+02 3.442e+02 5.810e+02, threshold=5.824e+02, percent-clipped=2.0 2023-04-29 22:47:54,153 INFO [train.py:904] (0/8) Epoch 14, batch 2950, loss[loss=0.1787, simple_loss=0.2732, pruned_loss=0.04211, over 17051.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2657, pruned_loss=0.04797, over 3326133.46 frames. ], batch size: 53, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:48:24,020 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=134923.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:48:48,789 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3291, 4.2779, 4.2408, 3.7548, 4.2728, 1.7224, 4.0432, 3.8900], device='cuda:0'), covar=tensor([0.0096, 0.0088, 0.0160, 0.0285, 0.0085, 0.2487, 0.0119, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0136, 0.0184, 0.0172, 0.0154, 0.0195, 0.0173, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:49:02,694 INFO [train.py:904] (0/8) Epoch 14, batch 3000, loss[loss=0.1772, simple_loss=0.2544, pruned_loss=0.05003, over 16778.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2662, pruned_loss=0.04841, over 3331759.94 frames. ], batch size: 102, lr: 4.91e-03, grad_scale: 8.0 2023-04-29 22:49:02,695 INFO [train.py:929] (0/8) Computing validation loss 2023-04-29 22:49:12,425 INFO [train.py:938] (0/8) Epoch 14, validation: loss=0.1382, simple_loss=0.244, pruned_loss=0.01621, over 944034.00 frames. 2023-04-29 22:49:12,426 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-29 22:49:25,014 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134960.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:49:31,415 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=134965.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:06,982 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.394e+02 2.911e+02 3.370e+02 7.406e+02, threshold=5.821e+02, percent-clipped=1.0 2023-04-29 22:50:24,227 INFO [train.py:904] (0/8) Epoch 14, batch 3050, loss[loss=0.1993, simple_loss=0.2707, pruned_loss=0.06398, over 16819.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2657, pruned_loss=0.04838, over 3321343.33 frames. ], batch size: 83, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:50:33,224 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135008.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:50:40,115 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135013.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:51:18,806 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0209, 4.9900, 4.8017, 4.1992, 4.9137, 1.9173, 4.6526, 4.7195], device='cuda:0'), covar=tensor([0.0093, 0.0087, 0.0183, 0.0404, 0.0091, 0.2530, 0.0135, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0136, 0.0184, 0.0172, 0.0154, 0.0195, 0.0174, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:51:32,228 INFO [train.py:904] (0/8) Epoch 14, batch 3100, loss[loss=0.2119, simple_loss=0.2827, pruned_loss=0.07056, over 16857.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.265, pruned_loss=0.04817, over 3317256.77 frames. ], batch size: 116, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:51:42,676 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-29 22:52:24,972 INFO [optim.py:368] (0/8) 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 22:52:40,988 INFO [train.py:904] (0/8) Epoch 14, batch 3150, loss[loss=0.1894, simple_loss=0.2767, pruned_loss=0.05112, over 17027.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2645, pruned_loss=0.04853, over 3317424.88 frames. ], batch size: 55, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:53:50,740 INFO [train.py:904] (0/8) Epoch 14, batch 3200, loss[loss=0.1704, simple_loss=0.2506, pruned_loss=0.04512, over 16870.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2635, pruned_loss=0.04791, over 3319145.63 frames. ], batch size: 116, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:54:10,568 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-04-29 22:54:17,324 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135171.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:54:41,152 INFO [optim.py:368] (0/8) 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 22:54:59,023 INFO [train.py:904] (0/8) Epoch 14, batch 3250, loss[loss=0.1659, simple_loss=0.2608, pruned_loss=0.03549, over 17112.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2629, pruned_loss=0.04715, over 3320365.26 frames. ], batch size: 47, lr: 4.90e-03, grad_scale: 8.0 2023-04-29 22:55:18,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4486, 5.8712, 5.6543, 5.6349, 5.2924, 5.1101, 5.2492, 5.9875], device='cuda:0'), covar=tensor([0.1179, 0.0948, 0.0974, 0.0793, 0.0831, 0.0741, 0.1187, 0.0976], device='cuda:0'), in_proj_covar=tensor([0.0608, 0.0758, 0.0620, 0.0546, 0.0482, 0.0482, 0.0632, 0.0584], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 22:55:22,128 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135218.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:55:41,206 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135232.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 22:55:57,061 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-29 22:56:07,855 INFO [train.py:904] (0/8) Epoch 14, batch 3300, loss[loss=0.1992, simple_loss=0.2823, pruned_loss=0.05807, over 16399.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2644, pruned_loss=0.04785, over 3322620.94 frames. ], batch size: 146, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:57:01,615 INFO [optim.py:368] (0/8) 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 22:57:16,997 INFO [train.py:904] (0/8) Epoch 14, batch 3350, loss[loss=0.186, simple_loss=0.2658, pruned_loss=0.05306, over 16771.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2648, pruned_loss=0.04772, over 3323659.96 frames. ], batch size: 83, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:57:42,738 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1468, 4.2026, 4.5457, 4.5518, 4.5575, 4.2372, 4.2690, 4.0932], device='cuda:0'), covar=tensor([0.0342, 0.0552, 0.0368, 0.0391, 0.0467, 0.0411, 0.0804, 0.0637], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0394, 0.0393, 0.0372, 0.0438, 0.0418, 0.0514, 0.0335], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 22:58:24,251 INFO [train.py:904] (0/8) Epoch 14, batch 3400, loss[loss=0.1585, simple_loss=0.2411, pruned_loss=0.03796, over 15708.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2637, pruned_loss=0.04675, over 3324682.36 frames. ], batch size: 35, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 22:58:32,594 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-29 22:59:16,985 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.205e+02 2.551e+02 3.042e+02 5.815e+02, threshold=5.103e+02, percent-clipped=2.0 2023-04-29 22:59:32,662 INFO [train.py:904] (0/8) Epoch 14, batch 3450, loss[loss=0.1764, simple_loss=0.2666, pruned_loss=0.04308, over 17077.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2634, pruned_loss=0.0469, over 3322705.61 frames. ], batch size: 53, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:00:41,091 INFO [train.py:904] (0/8) Epoch 14, batch 3500, loss[loss=0.1447, simple_loss=0.2289, pruned_loss=0.03022, over 16840.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2622, pruned_loss=0.0471, over 3324580.47 frames. ], batch size: 42, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:01:32,706 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-29 23:01:37,085 INFO [optim.py:368] (0/8) 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 23:01:51,754 INFO [train.py:904] (0/8) Epoch 14, batch 3550, loss[loss=0.1682, simple_loss=0.263, pruned_loss=0.03672, over 17118.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2605, pruned_loss=0.04637, over 3329281.26 frames. ], batch size: 47, lr: 4.90e-03, grad_scale: 4.0 2023-04-29 23:02:03,405 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2795, 3.4141, 3.6897, 2.2562, 3.0759, 2.6264, 3.6930, 3.6658], device='cuda:0'), covar=tensor([0.0236, 0.0915, 0.0493, 0.1684, 0.0749, 0.0870, 0.0560, 0.0897], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0155, 0.0161, 0.0147, 0.0139, 0.0126, 0.0140, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:02:15,145 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135518.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:02:27,276 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135527.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:03:02,253 INFO [train.py:904] (0/8) Epoch 14, batch 3600, loss[loss=0.1901, simple_loss=0.2788, pruned_loss=0.05071, over 17031.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2597, pruned_loss=0.04606, over 3314186.64 frames. ], batch size: 55, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:03:22,565 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 23:03:59,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7702, 2.9402, 2.6322, 4.8079, 3.9740, 4.4547, 1.5674, 3.2841], device='cuda:0'), covar=tensor([0.1335, 0.0716, 0.1105, 0.0209, 0.0262, 0.0358, 0.1545, 0.0653], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0182, 0.0163, 0.0200, 0.0210, 0.0186, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:04:14,257 INFO [train.py:904] (0/8) Epoch 14, batch 3650, loss[loss=0.1778, simple_loss=0.2553, pruned_loss=0.0501, over 16293.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2583, pruned_loss=0.04681, over 3310720.35 frames. ], batch size: 165, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:04:27,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9005, 3.7482, 3.8981, 4.0177, 4.0923, 3.6869, 3.9386, 4.1152], device='cuda:0'), covar=tensor([0.1154, 0.0928, 0.1042, 0.0570, 0.0563, 0.1938, 0.1638, 0.0639], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0745, 0.0899, 0.0765, 0.0576, 0.0589, 0.0601, 0.0704], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:04:45,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4805, 2.2517, 2.4305, 4.3477, 2.3024, 2.6908, 2.2969, 2.4773], device='cuda:0'), covar=tensor([0.1110, 0.3523, 0.2498, 0.0436, 0.3565, 0.2196, 0.3567, 0.2732], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0417, 0.0347, 0.0331, 0.0423, 0.0480, 0.0379, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:04:57,283 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 14, batch 3700, loss[loss=0.1945, simple_loss=0.2588, pruned_loss=0.06508, over 11396.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2567, pruned_loss=0.04818, over 3293674.67 frames. ], batch size: 248, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:06:30,623 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=135691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:06:32,154 INFO [optim.py:368] (0/8) 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 23:06:46,323 INFO [train.py:904] (0/8) Epoch 14, batch 3750, loss[loss=0.183, simple_loss=0.2553, pruned_loss=0.05534, over 16849.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2571, pruned_loss=0.04938, over 3291506.28 frames. ], batch size: 116, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:06:47,961 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0045, 5.0170, 4.8555, 4.2815, 4.9741, 1.9220, 4.7074, 4.5597], device='cuda:0'), covar=tensor([0.0092, 0.0069, 0.0168, 0.0312, 0.0078, 0.2373, 0.0122, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0135, 0.0180, 0.0168, 0.0153, 0.0192, 0.0171, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:08:01,587 INFO [train.py:904] (0/8) Epoch 14, batch 3800, loss[loss=0.185, simple_loss=0.2675, pruned_loss=0.05127, over 17110.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2586, pruned_loss=0.05069, over 3276359.48 frames. ], batch size: 49, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:00,855 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.772e+02 2.313e+02 2.548e+02 3.011e+02 5.523e+02, threshold=5.096e+02, percent-clipped=2.0 2023-04-29 23:09:16,026 INFO [train.py:904] (0/8) Epoch 14, batch 3850, loss[loss=0.1689, simple_loss=0.243, pruned_loss=0.04744, over 16757.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2591, pruned_loss=0.05146, over 3278499.53 frames. ], batch size: 83, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:09:25,288 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-29 23:09:53,698 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=135827.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:10:30,029 INFO [train.py:904] (0/8) Epoch 14, batch 3900, loss[loss=0.1764, simple_loss=0.2547, pruned_loss=0.04903, over 15388.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2587, pruned_loss=0.05163, over 3281565.63 frames. ], batch size: 190, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:11:05,162 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=135875.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:11:08,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3187, 2.6977, 2.0388, 2.3572, 3.0635, 2.7399, 3.2128, 3.1590], device='cuda:0'), covar=tensor([0.0127, 0.0272, 0.0455, 0.0391, 0.0185, 0.0304, 0.0190, 0.0203], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0216, 0.0211, 0.0211, 0.0218, 0.0218, 0.0227, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:11:30,221 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.144e+02 2.448e+02 3.272e+02 6.091e+02, threshold=4.896e+02, percent-clipped=2.0 2023-04-29 23:11:45,243 INFO [train.py:904] (0/8) Epoch 14, batch 3950, loss[loss=0.17, simple_loss=0.2572, pruned_loss=0.04138, over 17121.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2586, pruned_loss=0.05204, over 3280379.68 frames. ], batch size: 48, lr: 4.89e-03, grad_scale: 4.0 2023-04-29 23:11:57,908 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6977, 2.4579, 1.9640, 2.1709, 2.8453, 2.5982, 2.8549, 2.9252], device='cuda:0'), covar=tensor([0.0140, 0.0263, 0.0384, 0.0344, 0.0157, 0.0240, 0.0166, 0.0186], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0215, 0.0210, 0.0210, 0.0216, 0.0217, 0.0226, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:12:58,434 INFO [train.py:904] (0/8) Epoch 14, batch 4000, loss[loss=0.2068, simple_loss=0.2917, pruned_loss=0.06094, over 16465.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2592, pruned_loss=0.05273, over 3280272.27 frames. ], batch size: 75, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:13:41,784 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135982.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:47,611 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=135986.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:13:55,966 INFO [optim.py:368] (0/8) 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 23:13:58,940 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=135994.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:14:07,924 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-136000.pt 2023-04-29 23:14:13,530 INFO [train.py:904] (0/8) Epoch 14, batch 4050, loss[loss=0.1723, simple_loss=0.2583, pruned_loss=0.04312, over 16304.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.26, pruned_loss=0.05181, over 3279583.69 frames. ], batch size: 165, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:12,428 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136043.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:15:24,951 INFO [train.py:904] (0/8) Epoch 14, batch 4100, loss[loss=0.2018, simple_loss=0.292, pruned_loss=0.05583, over 16680.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2611, pruned_loss=0.05086, over 3276081.53 frames. ], batch size: 83, lr: 4.89e-03, grad_scale: 8.0 2023-04-29 23:15:29,877 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136055.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:16:07,775 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7136, 3.8480, 2.1643, 4.5317, 2.9969, 4.4480, 2.3301, 2.9467], device='cuda:0'), covar=tensor([0.0234, 0.0320, 0.1671, 0.0113, 0.0689, 0.0284, 0.1514, 0.0693], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0188, 0.0146, 0.0169, 0.0213, 0.0198, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:16:24,202 INFO [optim.py:368] (0/8) 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 23:16:39,711 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8909, 4.1917, 3.9842, 4.0747, 3.6905, 3.7873, 3.8587, 4.1374], device='cuda:0'), covar=tensor([0.1127, 0.0820, 0.0949, 0.0649, 0.0721, 0.1461, 0.0806, 0.1017], device='cuda:0'), in_proj_covar=tensor([0.0596, 0.0738, 0.0602, 0.0534, 0.0474, 0.0474, 0.0619, 0.0572], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:16:40,535 INFO [train.py:904] (0/8) Epoch 14, batch 4150, loss[loss=0.267, simple_loss=0.3454, pruned_loss=0.09431, over 11458.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2688, pruned_loss=0.05374, over 3226823.81 frames. ], batch size: 246, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:17:41,252 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136142.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:17:55,765 INFO [train.py:904] (0/8) Epoch 14, batch 4200, loss[loss=0.2172, simple_loss=0.3035, pruned_loss=0.06543, over 16487.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2763, pruned_loss=0.05603, over 3194828.32 frames. ], batch size: 68, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:18:54,053 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-04-29 23:18:55,505 INFO [optim.py:368] (0/8) 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 23:19:03,429 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-29 23:19:10,111 INFO [train.py:904] (0/8) Epoch 14, batch 4250, loss[loss=0.1675, simple_loss=0.2633, pruned_loss=0.03588, over 16957.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2798, pruned_loss=0.05614, over 3169129.38 frames. ], batch size: 109, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:19:12,626 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136203.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:19:16,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7356, 3.8915, 4.0731, 4.0765, 4.1034, 3.8404, 3.7991, 3.8140], device='cuda:0'), covar=tensor([0.0331, 0.0577, 0.0495, 0.0474, 0.0426, 0.0446, 0.1032, 0.0567], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0377, 0.0374, 0.0356, 0.0422, 0.0399, 0.0491, 0.0318], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-29 23:19:24,057 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.73 vs. limit=5.0 2023-04-29 23:19:27,069 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3078, 5.2889, 5.0030, 4.4401, 5.1664, 1.9603, 4.8968, 4.8537], device='cuda:0'), covar=tensor([0.0060, 0.0049, 0.0130, 0.0260, 0.0064, 0.2236, 0.0093, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0133, 0.0179, 0.0167, 0.0151, 0.0191, 0.0170, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:20:23,655 INFO [train.py:904] (0/8) Epoch 14, batch 4300, loss[loss=0.1982, simple_loss=0.2879, pruned_loss=0.05427, over 17006.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2805, pruned_loss=0.05534, over 3162128.76 frames. ], batch size: 55, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:20:31,573 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2423, 3.9724, 3.8148, 2.4902, 3.5246, 3.9230, 3.5718, 2.1166], device='cuda:0'), covar=tensor([0.0467, 0.0025, 0.0040, 0.0349, 0.0070, 0.0062, 0.0067, 0.0389], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0073, 0.0074, 0.0130, 0.0086, 0.0095, 0.0084, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 23:21:14,097 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136286.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:21:21,588 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 23:21:23,664 INFO [optim.py:368] (0/8) 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 23:21:38,141 INFO [train.py:904] (0/8) Epoch 14, batch 4350, loss[loss=0.2043, simple_loss=0.2894, pruned_loss=0.05961, over 16881.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2836, pruned_loss=0.05589, over 3181009.95 frames. ], batch size: 109, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:22:00,712 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7139, 4.7421, 4.5994, 4.3117, 4.2652, 4.6888, 4.4443, 4.3600], device='cuda:0'), covar=tensor([0.0506, 0.0324, 0.0229, 0.0235, 0.0820, 0.0322, 0.0385, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0355, 0.0312, 0.0291, 0.0328, 0.0339, 0.0212, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:22:26,307 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136334.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:28,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136335.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:33,166 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136338.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:50,534 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136350.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:22:53,927 INFO [train.py:904] (0/8) Epoch 14, batch 4400, loss[loss=0.2038, simple_loss=0.2912, pruned_loss=0.05825, over 16673.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2857, pruned_loss=0.05703, over 3165143.17 frames. ], batch size: 57, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:23:03,531 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0801, 2.2217, 1.6385, 1.9849, 2.6387, 2.3140, 2.7390, 2.8825], device='cuda:0'), covar=tensor([0.0095, 0.0378, 0.0590, 0.0439, 0.0216, 0.0327, 0.0199, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0214, 0.0210, 0.0208, 0.0216, 0.0216, 0.0222, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:23:51,149 INFO [optim.py:368] (0/8) 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 23:23:57,351 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136396.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:23:59,387 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5149, 1.7181, 2.1568, 2.4882, 2.4456, 2.8454, 1.7108, 2.6079], device='cuda:0'), covar=tensor([0.0162, 0.0410, 0.0236, 0.0244, 0.0234, 0.0149, 0.0428, 0.0127], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0179, 0.0163, 0.0172, 0.0179, 0.0135, 0.0180, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:24:06,500 INFO [train.py:904] (0/8) Epoch 14, batch 4450, loss[loss=0.2127, simple_loss=0.3075, pruned_loss=0.059, over 16876.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2885, pruned_loss=0.05728, over 3194023.61 frames. ], batch size: 96, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:18,616 INFO [train.py:904] (0/8) Epoch 14, batch 4500, loss[loss=0.2028, simple_loss=0.283, pruned_loss=0.06128, over 17106.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2889, pruned_loss=0.05783, over 3216665.16 frames. ], batch size: 47, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:25:19,399 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-29 23:26:18,143 INFO [optim.py:368] (0/8) 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 23:26:27,084 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136498.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:26:31,834 INFO [train.py:904] (0/8) Epoch 14, batch 4550, loss[loss=0.2423, simple_loss=0.3205, pruned_loss=0.08203, over 16388.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2898, pruned_loss=0.05888, over 3212074.65 frames. ], batch size: 35, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:27:27,147 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9973, 3.0657, 1.7130, 3.2666, 2.3080, 3.2900, 1.9847, 2.5422], device='cuda:0'), covar=tensor([0.0249, 0.0329, 0.1749, 0.0135, 0.0774, 0.0420, 0.1516, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0168, 0.0188, 0.0141, 0.0168, 0.0209, 0.0195, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:27:45,696 INFO [train.py:904] (0/8) Epoch 14, batch 4600, loss[loss=0.1941, simple_loss=0.2799, pruned_loss=0.05415, over 16689.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2909, pruned_loss=0.05927, over 3216561.48 frames. ], batch size: 124, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:28:29,755 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5843, 2.6231, 1.7847, 2.7334, 2.1318, 2.7538, 2.0446, 2.3457], device='cuda:0'), covar=tensor([0.0250, 0.0336, 0.1296, 0.0164, 0.0638, 0.0423, 0.1151, 0.0594], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0168, 0.0188, 0.0141, 0.0168, 0.0210, 0.0196, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:28:42,913 INFO [optim.py:368] (0/8) 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 23:28:57,103 INFO [train.py:904] (0/8) Epoch 14, batch 4650, loss[loss=0.1961, simple_loss=0.2924, pruned_loss=0.0499, over 16789.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2906, pruned_loss=0.05957, over 3206297.93 frames. ], batch size: 83, lr: 4.88e-03, grad_scale: 8.0 2023-04-29 23:29:31,230 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-29 23:29:51,591 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136638.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:29:58,855 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136643.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:03,452 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.33 vs. limit=5.0 2023-04-29 23:30:07,930 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136650.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:30:10,586 INFO [train.py:904] (0/8) Epoch 14, batch 4700, loss[loss=0.1856, simple_loss=0.2732, pruned_loss=0.04897, over 16430.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2872, pruned_loss=0.05802, over 3200427.21 frames. ], batch size: 68, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:30:23,030 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5630, 2.6607, 2.2256, 4.3548, 2.8595, 3.8708, 1.5628, 2.6445], device='cuda:0'), covar=tensor([0.1544, 0.0857, 0.1502, 0.0157, 0.0270, 0.0427, 0.1797, 0.1015], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0184, 0.0162, 0.0201, 0.0210, 0.0187, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:31:01,308 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136686.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:08,484 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136691.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:08,702 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6604, 3.0726, 3.0782, 1.8238, 2.6574, 2.1908, 3.0662, 3.2915], device='cuda:0'), covar=tensor([0.0406, 0.0811, 0.0672, 0.2062, 0.0958, 0.0960, 0.0989, 0.0987], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0153, 0.0161, 0.0147, 0.0140, 0.0126, 0.0139, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:31:09,912 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.115e+02 2.410e+02 3.015e+02 5.394e+02, threshold=4.821e+02, percent-clipped=2.0 2023-04-29 23:31:18,050 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136698.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:31:23,585 INFO [train.py:904] (0/8) Epoch 14, batch 4750, loss[loss=0.184, simple_loss=0.2668, pruned_loss=0.05062, over 16647.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2838, pruned_loss=0.05602, over 3202877.75 frames. ], batch size: 62, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:31:27,880 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136704.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:32:00,493 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-29 23:32:36,924 INFO [train.py:904] (0/8) Epoch 14, batch 4800, loss[loss=0.1799, simple_loss=0.274, pruned_loss=0.04285, over 16898.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2795, pruned_loss=0.0536, over 3206017.35 frames. ], batch size: 116, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:32:39,417 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9900, 4.1050, 3.8888, 3.5944, 3.5451, 3.9910, 3.6859, 3.7381], device='cuda:0'), covar=tensor([0.0566, 0.0437, 0.0270, 0.0273, 0.0791, 0.0425, 0.0968, 0.0544], device='cuda:0'), in_proj_covar=tensor([0.0253, 0.0351, 0.0308, 0.0287, 0.0325, 0.0334, 0.0210, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:33:01,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5544, 3.6429, 1.9060, 4.2055, 2.6441, 4.0388, 2.2638, 2.8164], device='cuda:0'), covar=tensor([0.0258, 0.0313, 0.1721, 0.0118, 0.0837, 0.0443, 0.1432, 0.0758], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0190, 0.0141, 0.0170, 0.0212, 0.0198, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:33:36,439 INFO [optim.py:368] (0/8) 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 23:33:46,310 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136798.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:33:52,185 INFO [train.py:904] (0/8) Epoch 14, batch 4850, loss[loss=0.1941, simple_loss=0.2849, pruned_loss=0.05162, over 15384.00 frames. ], tot_loss[loss=0.1935, simple_loss=0.2805, pruned_loss=0.05328, over 3191793.81 frames. ], batch size: 191, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:33:58,173 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=136806.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:34:56,353 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=136846.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:35:05,048 INFO [train.py:904] (0/8) Epoch 14, batch 4900, loss[loss=0.2225, simple_loss=0.2913, pruned_loss=0.07682, over 16611.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2795, pruned_loss=0.0522, over 3190258.60 frames. ], batch size: 57, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:35:28,880 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=136867.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:36:02,818 INFO [optim.py:368] (0/8) 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 23:36:18,508 INFO [train.py:904] (0/8) Epoch 14, batch 4950, loss[loss=0.1871, simple_loss=0.2815, pruned_loss=0.04637, over 17158.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.279, pruned_loss=0.05133, over 3199704.77 frames. ], batch size: 47, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:36:56,550 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1460, 4.2444, 3.4327, 2.7131, 3.3206, 2.8692, 4.7496, 4.0386], device='cuda:0'), covar=tensor([0.2294, 0.0687, 0.1492, 0.2016, 0.2146, 0.1410, 0.0451, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0259, 0.0289, 0.0288, 0.0284, 0.0230, 0.0275, 0.0309], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:37:30,154 INFO [train.py:904] (0/8) Epoch 14, batch 5000, loss[loss=0.1989, simple_loss=0.2842, pruned_loss=0.05679, over 16515.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2812, pruned_loss=0.05169, over 3187636.70 frames. ], batch size: 62, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:38:08,580 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6736, 2.4542, 2.4748, 4.3140, 2.2332, 2.9046, 2.4880, 2.6748], device='cuda:0'), covar=tensor([0.0999, 0.3095, 0.2300, 0.0419, 0.3559, 0.2202, 0.2868, 0.2910], device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0413, 0.0342, 0.0321, 0.0418, 0.0475, 0.0374, 0.0482], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:38:24,334 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=136991.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:38:25,067 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.688e+02 2.271e+02 2.523e+02 3.075e+02 5.093e+02, threshold=5.046e+02, percent-clipped=1.0 2023-04-29 23:38:36,211 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=136999.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:38:39,697 INFO [train.py:904] (0/8) Epoch 14, batch 5050, loss[loss=0.1822, simple_loss=0.2792, pruned_loss=0.04264, over 15455.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2817, pruned_loss=0.05168, over 3192914.52 frames. ], batch size: 190, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:39:31,244 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137039.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:39:49,989 INFO [train.py:904] (0/8) Epoch 14, batch 5100, loss[loss=0.1615, simple_loss=0.2456, pruned_loss=0.03871, over 16855.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2804, pruned_loss=0.05141, over 3180281.54 frames. ], batch size: 42, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:40:45,961 INFO [optim.py:368] (0/8) 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 23:41:00,266 INFO [train.py:904] (0/8) Epoch 14, batch 5150, loss[loss=0.1803, simple_loss=0.2836, pruned_loss=0.03848, over 16849.00 frames. ], tot_loss[loss=0.1898, simple_loss=0.2797, pruned_loss=0.05002, over 3180774.34 frames. ], batch size: 102, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:41:26,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9002, 5.0086, 4.7635, 4.4263, 4.1702, 4.9215, 4.8097, 4.4897], device='cuda:0'), covar=tensor([0.0698, 0.0488, 0.0365, 0.0339, 0.1399, 0.0424, 0.0316, 0.0745], device='cuda:0'), in_proj_covar=tensor([0.0250, 0.0347, 0.0304, 0.0285, 0.0322, 0.0332, 0.0206, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:42:08,570 INFO [train.py:904] (0/8) Epoch 14, batch 5200, loss[loss=0.2016, simple_loss=0.2893, pruned_loss=0.05693, over 15450.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2787, pruned_loss=0.05001, over 3177676.67 frames. ], batch size: 190, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:42:21,667 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=137162.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:43:03,331 INFO [optim.py:368] (0/8) 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 23:43:13,254 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.36 vs. limit=5.0 2023-04-29 23:43:18,148 INFO [train.py:904] (0/8) Epoch 14, batch 5250, loss[loss=0.2283, simple_loss=0.2998, pruned_loss=0.07841, over 12387.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2763, pruned_loss=0.04973, over 3187297.30 frames. ], batch size: 246, lr: 4.87e-03, grad_scale: 8.0 2023-04-29 23:43:57,276 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-04-29 23:44:25,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5660, 4.1706, 4.1065, 2.6973, 3.6094, 4.1044, 3.7183, 2.3307], device='cuda:0'), covar=tensor([0.0414, 0.0027, 0.0032, 0.0331, 0.0080, 0.0080, 0.0071, 0.0401], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0073, 0.0073, 0.0130, 0.0086, 0.0096, 0.0084, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-29 23:44:26,552 INFO [train.py:904] (0/8) Epoch 14, batch 5300, loss[loss=0.1754, simple_loss=0.2642, pruned_loss=0.04336, over 16569.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2731, pruned_loss=0.04881, over 3189399.00 frames. ], batch size: 75, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:45:23,606 INFO [optim.py:368] (0/8) 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 23:45:33,700 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137299.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:45:36,785 INFO [train.py:904] (0/8) Epoch 14, batch 5350, loss[loss=0.2038, simple_loss=0.2907, pruned_loss=0.0585, over 15296.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2708, pruned_loss=0.04799, over 3184374.23 frames. ], batch size: 190, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:46:40,460 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137347.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:46:46,693 INFO [train.py:904] (0/8) Epoch 14, batch 5400, loss[loss=0.1662, simple_loss=0.2648, pruned_loss=0.03377, over 16837.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2738, pruned_loss=0.04899, over 3186795.42 frames. ], batch size: 96, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:47:46,872 INFO [optim.py:368] (0/8) 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 23:48:02,473 INFO [train.py:904] (0/8) Epoch 14, batch 5450, loss[loss=0.2688, simple_loss=0.3473, pruned_loss=0.09515, over 15272.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2775, pruned_loss=0.05085, over 3191630.39 frames. ], batch size: 190, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:48:51,032 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4197, 2.0739, 2.2260, 4.1261, 1.9459, 2.3890, 2.2718, 2.2365], device='cuda:0'), covar=tensor([0.1214, 0.3829, 0.2672, 0.0502, 0.4812, 0.2867, 0.3353, 0.3795], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0407, 0.0339, 0.0317, 0.0414, 0.0468, 0.0371, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:49:15,266 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-29 23:49:19,967 INFO [train.py:904] (0/8) Epoch 14, batch 5500, loss[loss=0.2378, simple_loss=0.312, pruned_loss=0.08181, over 16684.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2855, pruned_loss=0.05561, over 3169962.28 frames. ], batch size: 62, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:49:35,268 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=137462.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:49:38,734 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9752, 2.1189, 2.3570, 3.1412, 2.2033, 2.3407, 2.3150, 2.2251], device='cuda:0'), covar=tensor([0.1019, 0.2725, 0.1927, 0.0570, 0.3363, 0.1862, 0.2520, 0.2759], device='cuda:0'), in_proj_covar=tensor([0.0371, 0.0409, 0.0341, 0.0319, 0.0416, 0.0470, 0.0373, 0.0477], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:50:16,520 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-29 23:50:23,841 INFO [optim.py:368] (0/8) 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 23:50:37,003 INFO [train.py:904] (0/8) Epoch 14, batch 5550, loss[loss=0.2734, simple_loss=0.3446, pruned_loss=0.1012, over 16202.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.293, pruned_loss=0.06136, over 3126115.41 frames. ], batch size: 165, lr: 4.86e-03, grad_scale: 4.0 2023-04-29 23:50:51,280 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=137510.0, num_to_drop=0, layers_to_drop=set() 2023-04-29 23:51:55,249 INFO [train.py:904] (0/8) Epoch 14, batch 5600, loss[loss=0.2143, simple_loss=0.2976, pruned_loss=0.06549, over 16614.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2985, pruned_loss=0.06636, over 3069970.39 frames. ], batch size: 62, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:51:57,840 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1530, 5.4590, 5.1779, 5.1867, 4.8967, 4.7822, 4.8349, 5.5350], device='cuda:0'), covar=tensor([0.1018, 0.0707, 0.0914, 0.0739, 0.0755, 0.0689, 0.1014, 0.0816], device='cuda:0'), in_proj_covar=tensor([0.0572, 0.0710, 0.0585, 0.0516, 0.0453, 0.0458, 0.0595, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:52:07,697 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-29 23:52:58,601 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-29 23:53:03,065 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.471e+02 3.513e+02 4.095e+02 5.435e+02 1.324e+03, threshold=8.190e+02, percent-clipped=8.0 2023-04-29 23:53:17,475 INFO [train.py:904] (0/8) Epoch 14, batch 5650, loss[loss=0.2075, simple_loss=0.2995, pruned_loss=0.05772, over 16302.00 frames. ], tot_loss[loss=0.2224, simple_loss=0.3037, pruned_loss=0.0705, over 3064336.23 frames. ], batch size: 146, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:54:05,363 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3793, 2.2127, 2.2162, 4.1044, 2.1564, 2.6873, 2.2760, 2.4452], device='cuda:0'), covar=tensor([0.1034, 0.3249, 0.2470, 0.0433, 0.3703, 0.2142, 0.3066, 0.2967], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0411, 0.0341, 0.0319, 0.0417, 0.0470, 0.0374, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:54:38,722 INFO [train.py:904] (0/8) Epoch 14, batch 5700, loss[loss=0.206, simple_loss=0.2985, pruned_loss=0.05676, over 16788.00 frames. ], tot_loss[loss=0.2233, simple_loss=0.3042, pruned_loss=0.07117, over 3056003.58 frames. ], batch size: 39, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:55:08,145 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9891, 2.3214, 2.3380, 2.8033, 2.0686, 3.2295, 1.6834, 2.6989], device='cuda:0'), covar=tensor([0.1155, 0.0578, 0.1046, 0.0164, 0.0123, 0.0378, 0.1423, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0161, 0.0201, 0.0209, 0.0188, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-29 23:55:45,167 INFO [optim.py:368] (0/8) 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 23:55:59,934 INFO [train.py:904] (0/8) Epoch 14, batch 5750, loss[loss=0.2556, simple_loss=0.3165, pruned_loss=0.09734, over 11560.00 frames. ], tot_loss[loss=0.2268, simple_loss=0.3073, pruned_loss=0.07316, over 3018941.72 frames. ], batch size: 248, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:56:14,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7796, 3.6899, 3.8354, 3.9863, 4.0594, 3.6408, 3.9455, 4.0779], device='cuda:0'), covar=tensor([0.1466, 0.1064, 0.1220, 0.0591, 0.0557, 0.1847, 0.0856, 0.0633], device='cuda:0'), in_proj_covar=tensor([0.0554, 0.0685, 0.0817, 0.0697, 0.0527, 0.0545, 0.0545, 0.0644], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-29 23:57:21,164 INFO [train.py:904] (0/8) Epoch 14, batch 5800, loss[loss=0.204, simple_loss=0.2964, pruned_loss=0.05576, over 16431.00 frames. ], tot_loss[loss=0.2252, simple_loss=0.3065, pruned_loss=0.07194, over 3019059.02 frames. ], batch size: 146, lr: 4.86e-03, grad_scale: 8.0 2023-04-29 23:58:14,277 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-29 23:58:26,269 INFO [optim.py:368] (0/8) 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 23:58:41,174 INFO [train.py:904] (0/8) Epoch 14, batch 5850, loss[loss=0.2157, simple_loss=0.3028, pruned_loss=0.06429, over 16936.00 frames. ], tot_loss[loss=0.222, simple_loss=0.3043, pruned_loss=0.06988, over 3033977.20 frames. ], batch size: 109, lr: 4.85e-03, grad_scale: 8.0 2023-04-29 23:59:00,730 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-04-30 00:00:03,519 INFO [train.py:904] (0/8) Epoch 14, batch 5900, loss[loss=0.2024, simple_loss=0.2831, pruned_loss=0.06079, over 15337.00 frames. ], tot_loss[loss=0.2207, simple_loss=0.3031, pruned_loss=0.06913, over 3045820.68 frames. ], batch size: 190, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:01:10,296 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137892.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:10,920 INFO [optim.py:368] (0/8) 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 00:01:15,023 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=137895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:01:24,300 INFO [train.py:904] (0/8) Epoch 14, batch 5950, loss[loss=0.2037, simple_loss=0.2902, pruned_loss=0.05866, over 15426.00 frames. ], tot_loss[loss=0.2198, simple_loss=0.3039, pruned_loss=0.06787, over 3055689.23 frames. ], batch size: 190, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:01:53,466 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6979, 4.0390, 4.0321, 2.2166, 3.4314, 2.7173, 4.1101, 4.1950], device='cuda:0'), covar=tensor([0.0202, 0.0594, 0.0473, 0.1812, 0.0659, 0.0856, 0.0482, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0150, 0.0161, 0.0146, 0.0138, 0.0126, 0.0139, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:02:44,561 INFO [train.py:904] (0/8) Epoch 14, batch 6000, loss[loss=0.2079, simple_loss=0.2979, pruned_loss=0.05897, over 16911.00 frames. ], tot_loss[loss=0.2188, simple_loss=0.3028, pruned_loss=0.06745, over 3048698.76 frames. ], batch size: 109, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:02:44,562 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 00:02:55,349 INFO [train.py:938] (0/8) Epoch 14, validation: loss=0.1574, simple_loss=0.2706, pruned_loss=0.0221, over 944034.00 frames. 2023-04-30 00:02:55,350 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 00:02:56,921 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:01,620 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=137956.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:03:26,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1135, 2.3528, 1.8224, 2.2080, 2.7802, 2.4880, 2.9268, 3.0479], device='cuda:0'), covar=tensor([0.0101, 0.0338, 0.0453, 0.0337, 0.0179, 0.0266, 0.0170, 0.0173], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0211, 0.0207, 0.0207, 0.0212, 0.0213, 0.0218, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 00:03:31,969 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 00:03:43,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7378, 2.3036, 2.2407, 3.2081, 2.2443, 3.5768, 1.3915, 2.6788], device='cuda:0'), covar=tensor([0.1364, 0.0822, 0.1206, 0.0161, 0.0213, 0.0382, 0.1683, 0.0816], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0184, 0.0160, 0.0201, 0.0208, 0.0187, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:03:59,084 INFO [optim.py:368] (0/8) 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 00:04:10,597 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-138000.pt 2023-04-30 00:04:18,781 INFO [train.py:904] (0/8) Epoch 14, batch 6050, loss[loss=0.2154, simple_loss=0.3069, pruned_loss=0.06194, over 16724.00 frames. ], tot_loss[loss=0.2172, simple_loss=0.3011, pruned_loss=0.0666, over 3074533.92 frames. ], batch size: 57, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:05:22,048 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0127, 2.5150, 2.6692, 1.9001, 2.7382, 2.8167, 2.4344, 2.3711], device='cuda:0'), covar=tensor([0.0704, 0.0202, 0.0196, 0.0889, 0.0090, 0.0209, 0.0434, 0.0422], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0104, 0.0089, 0.0137, 0.0072, 0.0113, 0.0122, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 00:05:40,172 INFO [train.py:904] (0/8) Epoch 14, batch 6100, loss[loss=0.2428, simple_loss=0.3104, pruned_loss=0.08763, over 11691.00 frames. ], tot_loss[loss=0.216, simple_loss=0.3007, pruned_loss=0.0657, over 3084012.22 frames. ], batch size: 246, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:06:44,481 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2646, 2.0608, 2.7358, 3.0871, 3.0020, 3.7415, 2.1106, 3.4492], device='cuda:0'), covar=tensor([0.0161, 0.0407, 0.0261, 0.0224, 0.0214, 0.0095, 0.0430, 0.0100], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0176, 0.0161, 0.0167, 0.0176, 0.0133, 0.0179, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 00:06:45,103 INFO [optim.py:368] (0/8) 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 00:06:59,058 INFO [train.py:904] (0/8) Epoch 14, batch 6150, loss[loss=0.2216, simple_loss=0.29, pruned_loss=0.07666, over 11652.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2981, pruned_loss=0.06471, over 3095768.82 frames. ], batch size: 248, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:08:16,180 INFO [train.py:904] (0/8) Epoch 14, batch 6200, loss[loss=0.2252, simple_loss=0.3056, pruned_loss=0.0724, over 15463.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2956, pruned_loss=0.06379, over 3104146.48 frames. ], batch size: 191, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:08:43,236 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9755, 3.0818, 1.6104, 3.2363, 2.2347, 3.2751, 1.8748, 2.4600], device='cuda:0'), covar=tensor([0.0245, 0.0349, 0.1828, 0.0180, 0.0826, 0.0509, 0.1598, 0.0731], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0167, 0.0187, 0.0138, 0.0166, 0.0206, 0.0195, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:09:17,892 INFO [optim.py:368] (0/8) 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 00:09:32,468 INFO [train.py:904] (0/8) Epoch 14, batch 6250, loss[loss=0.209, simple_loss=0.3008, pruned_loss=0.05862, over 16554.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2958, pruned_loss=0.0639, over 3097038.08 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:10:42,277 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:10:46,683 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 14, batch 6300, loss[loss=0.2266, simple_loss=0.2939, pruned_loss=0.07966, over 11537.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2951, pruned_loss=0.06299, over 3111151.06 frames. ], batch size: 247, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:11:17,691 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:11:52,440 INFO [optim.py:368] (0/8) 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 00:12:05,888 INFO [train.py:904] (0/8) Epoch 14, batch 6350, loss[loss=0.2021, simple_loss=0.2792, pruned_loss=0.06247, over 16371.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2959, pruned_loss=0.0638, over 3116654.18 frames. ], batch size: 68, lr: 4.85e-03, grad_scale: 8.0 2023-04-30 00:12:36,943 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138322.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:12:50,094 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9931, 4.1160, 2.4402, 4.8520, 3.0191, 4.7040, 2.8148, 3.1047], device='cuda:0'), covar=tensor([0.0213, 0.0281, 0.1571, 0.0128, 0.0718, 0.0448, 0.1199, 0.0677], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0168, 0.0188, 0.0138, 0.0166, 0.0206, 0.0195, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:12:50,122 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:13:22,163 INFO [train.py:904] (0/8) Epoch 14, batch 6400, loss[loss=0.1947, simple_loss=0.2848, pruned_loss=0.0523, over 16592.00 frames. ], tot_loss[loss=0.2131, simple_loss=0.2962, pruned_loss=0.06501, over 3113189.54 frames. ], batch size: 62, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:14:09,373 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:14:23,620 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.151e+02 3.125e+02 3.740e+02 4.766e+02 1.292e+03, threshold=7.481e+02, percent-clipped=6.0 2023-04-30 00:14:37,182 INFO [train.py:904] (0/8) Epoch 14, batch 6450, loss[loss=0.204, simple_loss=0.2865, pruned_loss=0.0608, over 16395.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2964, pruned_loss=0.06371, over 3136411.28 frames. ], batch size: 146, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:14:58,921 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 00:15:54,751 INFO [train.py:904] (0/8) Epoch 14, batch 6500, loss[loss=0.1806, simple_loss=0.2592, pruned_loss=0.051, over 16791.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2944, pruned_loss=0.06364, over 3111130.68 frames. ], batch size: 39, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:16:28,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7613, 3.8217, 2.0658, 4.4581, 2.7207, 4.3604, 2.4007, 2.9818], device='cuda:0'), covar=tensor([0.0224, 0.0299, 0.1670, 0.0150, 0.0790, 0.0432, 0.1428, 0.0681], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0168, 0.0188, 0.0138, 0.0166, 0.0207, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:16:59,451 INFO [optim.py:368] (0/8) 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 00:17:12,096 INFO [train.py:904] (0/8) Epoch 14, batch 6550, loss[loss=0.2871, simple_loss=0.3421, pruned_loss=0.1161, over 11556.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.297, pruned_loss=0.06381, over 3130753.01 frames. ], batch size: 248, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:17:32,632 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 00:18:22,253 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:25,798 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:18:26,460 INFO [train.py:904] (0/8) Epoch 14, batch 6600, loss[loss=0.2372, simple_loss=0.3061, pruned_loss=0.08413, over 11695.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2995, pruned_loss=0.06435, over 3129340.84 frames. ], batch size: 246, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:18:38,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4454, 2.9077, 2.9531, 1.8771, 2.6901, 2.1107, 3.0493, 3.1035], device='cuda:0'), covar=tensor([0.0301, 0.0690, 0.0596, 0.1906, 0.0807, 0.0968, 0.0661, 0.0777], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0149, 0.0160, 0.0144, 0.0136, 0.0125, 0.0137, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:19:13,952 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1798, 4.2121, 4.5582, 4.5069, 4.5305, 4.2234, 4.2219, 4.1481], device='cuda:0'), covar=tensor([0.0312, 0.0503, 0.0391, 0.0458, 0.0541, 0.0418, 0.1027, 0.0511], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0380, 0.0377, 0.0360, 0.0425, 0.0404, 0.0498, 0.0319], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 00:19:29,572 INFO [optim.py:368] (0/8) 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 00:19:30,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5202, 3.6262, 2.8035, 2.0988, 2.4724, 2.2569, 3.8974, 3.4020], device='cuda:0'), covar=tensor([0.2846, 0.0790, 0.1702, 0.2322, 0.2455, 0.1931, 0.0475, 0.1111], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0261, 0.0289, 0.0290, 0.0285, 0.0232, 0.0278, 0.0309], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 00:19:33,987 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:38,210 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138599.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:19:42,710 INFO [train.py:904] (0/8) Epoch 14, batch 6650, loss[loss=0.2099, simple_loss=0.2976, pruned_loss=0.06108, over 16759.00 frames. ], tot_loss[loss=0.2155, simple_loss=0.2999, pruned_loss=0.06549, over 3124896.49 frames. ], batch size: 83, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:19:46,744 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 00:20:17,067 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 00:20:19,659 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:20:51,079 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5293, 3.7505, 4.2513, 2.2119, 4.3999, 4.4740, 3.0971, 3.2027], device='cuda:0'), covar=tensor([0.0845, 0.0230, 0.0175, 0.1055, 0.0054, 0.0097, 0.0385, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0104, 0.0089, 0.0139, 0.0073, 0.0113, 0.0124, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 00:20:58,175 INFO [train.py:904] (0/8) Epoch 14, batch 6700, loss[loss=0.1923, simple_loss=0.2869, pruned_loss=0.04881, over 16792.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2987, pruned_loss=0.06552, over 3123675.56 frames. ], batch size: 89, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:21:39,963 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=138678.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:44,943 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138681.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:21:53,095 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:03,785 INFO [optim.py:368] (0/8) 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 00:22:08,565 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:22:15,573 INFO [train.py:904] (0/8) Epoch 14, batch 6750, loss[loss=0.1881, simple_loss=0.277, pruned_loss=0.0496, over 16803.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2979, pruned_loss=0.06588, over 3099451.30 frames. ], batch size: 102, lr: 4.84e-03, grad_scale: 4.0 2023-04-30 00:22:46,552 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-30 00:23:14,913 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:22,353 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:29,047 INFO [train.py:904] (0/8) Epoch 14, batch 6800, loss[loss=0.2322, simple_loss=0.3155, pruned_loss=0.07442, over 16631.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.298, pruned_loss=0.06554, over 3106141.61 frames. ], batch size: 62, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:23:38,394 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138758.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:23:53,487 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-04-30 00:23:57,739 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4885, 4.4606, 4.8589, 4.8110, 4.8639, 4.5371, 4.5308, 4.3004], device='cuda:0'), covar=tensor([0.0296, 0.0530, 0.0344, 0.0423, 0.0446, 0.0358, 0.0922, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0376, 0.0375, 0.0357, 0.0422, 0.0400, 0.0491, 0.0317], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 00:24:34,694 INFO [optim.py:368] (0/8) 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 00:24:46,921 INFO [train.py:904] (0/8) Epoch 14, batch 6850, loss[loss=0.2086, simple_loss=0.3156, pruned_loss=0.05075, over 16587.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.299, pruned_loss=0.06606, over 3106167.47 frames. ], batch size: 35, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:25:15,882 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8135, 1.6023, 2.1952, 2.5897, 2.5932, 2.7816, 1.6369, 2.9250], device='cuda:0'), covar=tensor([0.0124, 0.0448, 0.0265, 0.0231, 0.0222, 0.0155, 0.0500, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0174, 0.0158, 0.0165, 0.0172, 0.0131, 0.0176, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-30 00:25:33,757 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4426, 3.6888, 3.6993, 2.0838, 3.1546, 2.5768, 3.9037, 3.9175], device='cuda:0'), covar=tensor([0.0198, 0.0670, 0.0567, 0.1737, 0.0705, 0.0874, 0.0427, 0.0616], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0150, 0.0160, 0.0145, 0.0137, 0.0125, 0.0138, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:26:02,487 INFO [train.py:904] (0/8) Epoch 14, batch 6900, loss[loss=0.2993, simple_loss=0.3523, pruned_loss=0.1231, over 11253.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.3014, pruned_loss=0.06584, over 3106810.89 frames. ], batch size: 248, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:26:21,485 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.52 vs. limit=2.0 2023-04-30 00:26:33,062 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9504, 2.7690, 2.7985, 2.1065, 2.6416, 2.1810, 2.7027, 2.8818], device='cuda:0'), covar=tensor([0.0284, 0.0653, 0.0447, 0.1510, 0.0680, 0.0818, 0.0529, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0150, 0.0160, 0.0146, 0.0137, 0.0126, 0.0137, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:26:33,280 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-30 00:26:51,083 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1242, 3.4894, 3.7065, 1.5424, 3.7830, 4.0309, 3.0046, 2.7189], device='cuda:0'), covar=tensor([0.1112, 0.0165, 0.0189, 0.1452, 0.0073, 0.0113, 0.0404, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0104, 0.0089, 0.0138, 0.0072, 0.0113, 0.0124, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 00:27:04,436 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-30 00:27:10,349 INFO [optim.py:368] (0/8) 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 00:27:22,857 INFO [train.py:904] (0/8) Epoch 14, batch 6950, loss[loss=0.2254, simple_loss=0.3073, pruned_loss=0.07174, over 15506.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3034, pruned_loss=0.06752, over 3101643.14 frames. ], batch size: 191, lr: 4.84e-03, grad_scale: 8.0 2023-04-30 00:27:33,756 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=138909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:00,044 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138926.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:28:03,019 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8690, 5.4040, 5.6262, 5.3171, 5.3048, 5.9281, 5.4050, 5.2016], device='cuda:0'), covar=tensor([0.0994, 0.1402, 0.1731, 0.1768, 0.2304, 0.0817, 0.1421, 0.2072], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0529, 0.0582, 0.0449, 0.0602, 0.0602, 0.0458, 0.0608], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 00:28:38,217 INFO [train.py:904] (0/8) Epoch 14, batch 7000, loss[loss=0.2016, simple_loss=0.3094, pruned_loss=0.04689, over 16777.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3031, pruned_loss=0.06693, over 3092841.47 frames. ], batch size: 83, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:29:06,820 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=138970.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 00:29:12,265 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=138974.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:18,578 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=138978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:29:37,962 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 00:29:42,507 INFO [optim.py:368] (0/8) 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 00:29:46,137 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1960, 3.3398, 3.5428, 3.5203, 3.5124, 3.3043, 3.3538, 3.4196], device='cuda:0'), covar=tensor([0.0400, 0.0682, 0.0435, 0.0437, 0.0509, 0.0538, 0.0802, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0354, 0.0374, 0.0373, 0.0356, 0.0420, 0.0400, 0.0490, 0.0315], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 00:29:55,164 INFO [train.py:904] (0/8) Epoch 14, batch 7050, loss[loss=0.2125, simple_loss=0.2991, pruned_loss=0.06292, over 16613.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.304, pruned_loss=0.06741, over 3080350.20 frames. ], batch size: 62, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:30:08,329 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-30 00:30:33,409 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:42,634 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4577, 1.6344, 1.9776, 2.3069, 2.4186, 2.6638, 1.6960, 2.5132], device='cuda:0'), covar=tensor([0.0163, 0.0393, 0.0290, 0.0273, 0.0249, 0.0161, 0.0410, 0.0120], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0173, 0.0159, 0.0164, 0.0172, 0.0131, 0.0176, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-30 00:30:49,919 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:30:57,653 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139042.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:31:13,411 INFO [train.py:904] (0/8) Epoch 14, batch 7100, loss[loss=0.1728, simple_loss=0.2642, pruned_loss=0.04071, over 16789.00 frames. ], tot_loss[loss=0.2192, simple_loss=0.3027, pruned_loss=0.06785, over 3055140.27 frames. ], batch size: 83, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:31:15,012 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139053.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:32:18,348 INFO [optim.py:368] (0/8) 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 00:32:31,537 INFO [train.py:904] (0/8) Epoch 14, batch 7150, loss[loss=0.221, simple_loss=0.2992, pruned_loss=0.07139, over 15417.00 frames. ], tot_loss[loss=0.218, simple_loss=0.3009, pruned_loss=0.06751, over 3053025.42 frames. ], batch size: 191, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:33:47,884 INFO [train.py:904] (0/8) Epoch 14, batch 7200, loss[loss=0.2, simple_loss=0.2912, pruned_loss=0.0544, over 16366.00 frames. ], tot_loss[loss=0.2148, simple_loss=0.2987, pruned_loss=0.06549, over 3054854.75 frames. ], batch size: 35, lr: 4.83e-03, grad_scale: 8.0 2023-04-30 00:33:48,910 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 00:34:55,360 INFO [optim.py:368] (0/8) 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 00:35:07,123 INFO [train.py:904] (0/8) Epoch 14, batch 7250, loss[loss=0.1769, simple_loss=0.2584, pruned_loss=0.04771, over 16452.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.296, pruned_loss=0.06428, over 3051806.14 frames. ], batch size: 68, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:36:01,919 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:36:21,529 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2108, 1.3813, 1.8466, 2.0130, 2.1209, 2.3422, 1.5906, 2.1557], device='cuda:0'), covar=tensor([0.0162, 0.0399, 0.0224, 0.0256, 0.0228, 0.0159, 0.0386, 0.0097], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0174, 0.0158, 0.0164, 0.0172, 0.0130, 0.0176, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-30 00:36:22,188 INFO [train.py:904] (0/8) Epoch 14, batch 7300, loss[loss=0.2069, simple_loss=0.2978, pruned_loss=0.05799, over 15338.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2951, pruned_loss=0.06399, over 3048303.93 frames. ], batch size: 191, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:36:42,740 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139265.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:37:00,526 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139276.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:19,405 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 00:37:20,391 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139289.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:29,630 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.078e+02 2.984e+02 3.521e+02 4.495e+02 8.579e+02, threshold=7.042e+02, percent-clipped=5.0 2023-04-30 00:37:37,012 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:37:40,544 INFO [train.py:904] (0/8) Epoch 14, batch 7350, loss[loss=0.1978, simple_loss=0.2845, pruned_loss=0.05554, over 16532.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2958, pruned_loss=0.06449, over 3044056.95 frames. ], batch size: 68, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:37:57,881 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5579, 4.5870, 5.0269, 4.9899, 4.9869, 4.6370, 4.6397, 4.4270], device='cuda:0'), covar=tensor([0.0303, 0.0459, 0.0321, 0.0431, 0.0496, 0.0353, 0.0949, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0352, 0.0371, 0.0370, 0.0355, 0.0418, 0.0397, 0.0484, 0.0315], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 00:38:36,759 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139337.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:36,864 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139337.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 00:38:44,982 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:56,636 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:38:59,360 INFO [train.py:904] (0/8) Epoch 14, batch 7400, loss[loss=0.2034, simple_loss=0.2954, pruned_loss=0.05567, over 16892.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2967, pruned_loss=0.06499, over 3044442.35 frames. ], batch size: 109, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:39:01,650 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139353.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:39:01,848 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8170, 3.1982, 3.1522, 1.9685, 2.7790, 2.1093, 3.4136, 3.4470], device='cuda:0'), covar=tensor([0.0261, 0.0721, 0.0641, 0.2041, 0.0887, 0.1012, 0.0615, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0152, 0.0162, 0.0147, 0.0139, 0.0127, 0.0139, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:39:31,322 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5003, 3.0479, 2.6112, 2.2660, 2.4247, 2.2035, 2.9864, 2.9374], device='cuda:0'), covar=tensor([0.2528, 0.0726, 0.1580, 0.2176, 0.2279, 0.2039, 0.0538, 0.1203], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0262, 0.0292, 0.0292, 0.0286, 0.0233, 0.0277, 0.0310], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 00:39:52,862 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:00,671 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:08,331 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.794e+02 3.378e+02 4.039e+02 6.978e+02, threshold=6.757e+02, percent-clipped=0.0 2023-04-30 00:40:18,578 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:40:19,234 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-04-30 00:40:19,491 INFO [train.py:904] (0/8) Epoch 14, batch 7450, loss[loss=0.1947, simple_loss=0.2928, pruned_loss=0.04826, over 16369.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.298, pruned_loss=0.06562, over 3083605.65 frames. ], batch size: 146, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:41:43,387 INFO [train.py:904] (0/8) Epoch 14, batch 7500, loss[loss=0.2469, simple_loss=0.3105, pruned_loss=0.09167, over 11345.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2983, pruned_loss=0.06485, over 3074683.72 frames. ], batch size: 248, lr: 4.83e-03, grad_scale: 4.0 2023-04-30 00:42:53,287 INFO [optim.py:368] (0/8) 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 00:43:02,976 INFO [train.py:904] (0/8) Epoch 14, batch 7550, loss[loss=0.2125, simple_loss=0.2907, pruned_loss=0.06714, over 15192.00 frames. ], tot_loss[loss=0.2149, simple_loss=0.2984, pruned_loss=0.06565, over 3060571.56 frames. ], batch size: 190, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:43:51,639 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6182, 2.5276, 2.2762, 4.1859, 3.1426, 4.0360, 1.4815, 2.9343], device='cuda:0'), covar=tensor([0.1354, 0.0794, 0.1309, 0.0142, 0.0283, 0.0353, 0.1581, 0.0775], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0162, 0.0203, 0.0209, 0.0187, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:43:53,251 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 00:44:19,050 INFO [train.py:904] (0/8) Epoch 14, batch 7600, loss[loss=0.2798, simple_loss=0.3384, pruned_loss=0.1106, over 11452.00 frames. ], tot_loss[loss=0.2139, simple_loss=0.2971, pruned_loss=0.06541, over 3072616.95 frames. ], batch size: 248, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:44:39,705 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139565.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:44:49,028 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139571.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:05,716 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:45:19,210 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 00:45:23,449 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.790e+02 2.745e+02 3.435e+02 4.599e+02 7.932e+02, threshold=6.869e+02, percent-clipped=2.0 2023-04-30 00:45:29,391 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-04-30 00:45:34,165 INFO [train.py:904] (0/8) Epoch 14, batch 7650, loss[loss=0.2601, simple_loss=0.315, pruned_loss=0.1026, over 11273.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2976, pruned_loss=0.06578, over 3086951.16 frames. ], batch size: 250, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:45:50,720 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:19,137 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139632.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 00:46:19,318 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:36,550 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:39,510 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139645.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:46:49,448 INFO [train.py:904] (0/8) Epoch 14, batch 7700, loss[loss=0.1839, simple_loss=0.2711, pruned_loss=0.04829, over 16859.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.298, pruned_loss=0.06663, over 3069472.54 frames. ], batch size: 90, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:47:47,219 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6108, 1.6363, 2.1119, 2.4959, 2.5241, 2.9054, 1.7687, 2.8380], device='cuda:0'), covar=tensor([0.0161, 0.0465, 0.0288, 0.0273, 0.0247, 0.0154, 0.0454, 0.0105], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0176, 0.0159, 0.0165, 0.0174, 0.0131, 0.0177, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-30 00:47:52,653 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5708, 5.8756, 5.6783, 5.6503, 5.2669, 5.1619, 5.3025, 6.0175], device='cuda:0'), covar=tensor([0.1164, 0.0833, 0.0908, 0.0821, 0.0827, 0.0651, 0.1035, 0.0840], device='cuda:0'), in_proj_covar=tensor([0.0582, 0.0717, 0.0592, 0.0516, 0.0452, 0.0467, 0.0600, 0.0549], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 00:47:57,376 INFO [optim.py:368] (0/8) 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 00:48:06,810 INFO [train.py:904] (0/8) Epoch 14, batch 7750, loss[loss=0.2392, simple_loss=0.3257, pruned_loss=0.07635, over 16731.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2975, pruned_loss=0.06594, over 3078957.30 frames. ], batch size: 124, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:24,619 INFO [train.py:904] (0/8) Epoch 14, batch 7800, loss[loss=0.1894, simple_loss=0.2871, pruned_loss=0.04583, over 16485.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2988, pruned_loss=0.06722, over 3056832.69 frames. ], batch size: 75, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:49:51,627 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139769.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:01,849 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4710, 3.4471, 3.3989, 2.7441, 3.3293, 2.1696, 3.1094, 2.7954], device='cuda:0'), covar=tensor([0.0163, 0.0134, 0.0153, 0.0240, 0.0105, 0.2012, 0.0146, 0.0187], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0127, 0.0173, 0.0161, 0.0146, 0.0187, 0.0161, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 00:50:21,926 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:32,561 INFO [optim.py:368] (0/8) 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 00:50:41,163 INFO [train.py:904] (0/8) Epoch 14, batch 7850, loss[loss=0.2314, simple_loss=0.32, pruned_loss=0.0714, over 15474.00 frames. ], tot_loss[loss=0.2164, simple_loss=0.299, pruned_loss=0.06689, over 3052345.88 frames. ], batch size: 191, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:50:43,288 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8566, 5.1425, 5.4218, 5.2021, 5.2261, 5.7648, 5.2371, 5.0264], device='cuda:0'), covar=tensor([0.1023, 0.1728, 0.2193, 0.1835, 0.2539, 0.1010, 0.1503, 0.2457], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0523, 0.0578, 0.0446, 0.0599, 0.0600, 0.0453, 0.0607], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 00:50:45,357 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:50:48,512 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:23,909 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,616 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:53,630 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139850.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:51:56,204 INFO [train.py:904] (0/8) Epoch 14, batch 7900, loss[loss=0.2043, simple_loss=0.2928, pruned_loss=0.05793, over 16577.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2974, pruned_loss=0.06566, over 3076197.28 frames. ], batch size: 68, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:51:59,873 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.67 vs. limit=5.0 2023-04-30 00:52:15,863 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:18,370 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139867.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:52:33,938 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.64 vs. limit=5.0 2023-04-30 00:53:01,524 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139894.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:03,663 INFO [optim.py:368] (0/8) 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 00:53:12,914 INFO [train.py:904] (0/8) Epoch 14, batch 7950, loss[loss=0.2691, simple_loss=0.3315, pruned_loss=0.1034, over 11682.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2973, pruned_loss=0.06574, over 3092156.11 frames. ], batch size: 247, lr: 4.82e-03, grad_scale: 4.0 2023-04-30 00:53:27,462 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=139911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:51,349 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:53:59,029 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139932.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 00:54:00,377 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 00:54:08,363 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=139938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:13,655 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:17,760 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=139945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:54:24,149 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 00:54:28,361 INFO [train.py:904] (0/8) Epoch 14, batch 8000, loss[loss=0.1905, simple_loss=0.2818, pruned_loss=0.04956, over 16239.00 frames. ], tot_loss[loss=0.2156, simple_loss=0.2981, pruned_loss=0.06654, over 3078638.83 frames. ], batch size: 165, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:55:07,147 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8158, 4.8047, 4.6174, 4.3654, 4.2577, 4.6996, 4.6103, 4.3896], device='cuda:0'), covar=tensor([0.0548, 0.0431, 0.0297, 0.0269, 0.1030, 0.0408, 0.0365, 0.0633], device='cuda:0'), in_proj_covar=tensor([0.0251, 0.0343, 0.0300, 0.0277, 0.0312, 0.0327, 0.0205, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 00:55:10,469 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139980.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:30,355 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=139993.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:55:34,112 INFO [optim.py:368] (0/8) 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 00:55:40,544 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-140000.pt 2023-04-30 00:55:46,032 INFO [train.py:904] (0/8) Epoch 14, batch 8050, loss[loss=0.2477, simple_loss=0.3119, pruned_loss=0.09176, over 11710.00 frames. ], tot_loss[loss=0.2166, simple_loss=0.2989, pruned_loss=0.06714, over 3063935.40 frames. ], batch size: 248, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:56:33,027 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7777, 3.8868, 2.3384, 4.5510, 2.8509, 4.4696, 2.3600, 2.9057], device='cuda:0'), covar=tensor([0.0239, 0.0366, 0.1600, 0.0179, 0.0763, 0.0474, 0.1594, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0169, 0.0190, 0.0139, 0.0167, 0.0209, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 00:56:33,314 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 00:56:59,051 INFO [train.py:904] (0/8) Epoch 14, batch 8100, loss[loss=0.2056, simple_loss=0.292, pruned_loss=0.0596, over 16531.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2987, pruned_loss=0.06644, over 3072634.47 frames. ], batch size: 57, lr: 4.82e-03, grad_scale: 8.0 2023-04-30 00:58:06,573 INFO [optim.py:368] (0/8) 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 00:58:17,064 INFO [train.py:904] (0/8) Epoch 14, batch 8150, loss[loss=0.2361, simple_loss=0.2972, pruned_loss=0.08754, over 11306.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2968, pruned_loss=0.06591, over 3061269.52 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:58:18,874 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:20,726 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-30 00:58:29,454 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140110.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:58:52,079 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:23,736 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:34,237 INFO [train.py:904] (0/8) Epoch 14, batch 8200, loss[loss=0.1927, simple_loss=0.2858, pruned_loss=0.04985, over 16891.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2941, pruned_loss=0.06521, over 3078681.42 frames. ], batch size: 102, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 00:59:47,447 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140160.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:50,140 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140162.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 00:59:53,885 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:05,623 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:00:32,257 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-30 01:00:36,414 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8797, 3.8461, 3.9432, 3.7594, 3.9365, 4.3153, 3.9733, 3.6511], device='cuda:0'), covar=tensor([0.1878, 0.1977, 0.1991, 0.2337, 0.2484, 0.1476, 0.1416, 0.2451], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0515, 0.0572, 0.0442, 0.0590, 0.0596, 0.0448, 0.0598], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 01:00:44,847 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.704e+02 2.531e+02 3.267e+02 4.112e+02 9.606e+02, threshold=6.535e+02, percent-clipped=4.0 2023-04-30 01:00:55,640 INFO [train.py:904] (0/8) Epoch 14, batch 8250, loss[loss=0.1661, simple_loss=0.2642, pruned_loss=0.03402, over 16600.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2924, pruned_loss=0.06276, over 3046810.93 frames. ], batch size: 76, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:01:02,326 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140206.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:37,276 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:01:55,110 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:16,388 INFO [train.py:904] (0/8) Epoch 14, batch 8300, loss[loss=0.1847, simple_loss=0.2798, pruned_loss=0.04479, over 16659.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2893, pruned_loss=0.05908, over 3046919.74 frames. ], batch size: 57, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:02:29,041 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140259.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:02:54,967 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:12,773 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:03:27,886 INFO [optim.py:368] (0/8) 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 01:03:28,701 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4397, 3.0277, 2.7285, 2.1496, 2.1611, 2.2518, 2.8257, 2.9188], device='cuda:0'), covar=tensor([0.2389, 0.0822, 0.1461, 0.2589, 0.2363, 0.2070, 0.0446, 0.1186], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0256, 0.0284, 0.0285, 0.0279, 0.0228, 0.0269, 0.0303], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:03:36,683 INFO [train.py:904] (0/8) Epoch 14, batch 8350, loss[loss=0.212, simple_loss=0.2871, pruned_loss=0.06843, over 12225.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.288, pruned_loss=0.05673, over 3038062.02 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:03:41,729 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 01:04:07,924 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:04:57,945 INFO [train.py:904] (0/8) Epoch 14, batch 8400, loss[loss=0.1793, simple_loss=0.2612, pruned_loss=0.04877, over 12158.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2851, pruned_loss=0.05472, over 3022402.89 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:05:04,224 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 01:05:54,566 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140387.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:06:08,149 INFO [optim.py:368] (0/8) 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 01:06:18,709 INFO [train.py:904] (0/8) Epoch 14, batch 8450, loss[loss=0.1807, simple_loss=0.2779, pruned_loss=0.04172, over 16668.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.284, pruned_loss=0.05349, over 3007160.68 frames. ], batch size: 134, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:06:55,823 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:27,242 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:27,361 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5952, 3.9302, 3.8665, 2.9250, 3.4474, 3.9298, 3.6640, 2.1839], device='cuda:0'), covar=tensor([0.0376, 0.0031, 0.0035, 0.0275, 0.0081, 0.0064, 0.0061, 0.0424], device='cuda:0'), in_proj_covar=tensor([0.0127, 0.0070, 0.0071, 0.0126, 0.0083, 0.0093, 0.0081, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 01:07:32,365 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:37,967 INFO [train.py:904] (0/8) Epoch 14, batch 8500, loss[loss=0.1864, simple_loss=0.2868, pruned_loss=0.04301, over 16374.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2803, pruned_loss=0.05099, over 3020398.17 frames. ], batch size: 146, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:07:50,018 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:51,462 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:07:54,725 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:01,384 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:12,868 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:38,531 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140489.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:46,068 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:08:51,332 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.625e+02 2.295e+02 2.729e+02 3.409e+02 8.118e+02, threshold=5.459e+02, percent-clipped=4.0 2023-04-30 01:09:01,487 INFO [train.py:904] (0/8) Epoch 14, batch 8550, loss[loss=0.178, simple_loss=0.2772, pruned_loss=0.03947, over 16871.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2783, pruned_loss=0.04973, over 3021594.81 frames. ], batch size: 96, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:09:11,504 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:15,280 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140508.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:17,255 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:09:18,514 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140510.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:39,004 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140550.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:10:41,526 INFO [train.py:904] (0/8) Epoch 14, batch 8600, loss[loss=0.1783, simple_loss=0.2728, pruned_loss=0.04186, over 16625.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2785, pruned_loss=0.04872, over 3007861.34 frames. ], batch size: 62, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:10:46,496 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140554.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:19,636 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:11:33,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7978, 3.8084, 3.9289, 3.7272, 3.8729, 4.2673, 3.9836, 3.6583], device='cuda:0'), covar=tensor([0.2226, 0.2375, 0.2332, 0.2535, 0.3084, 0.1826, 0.1523, 0.2800], device='cuda:0'), in_proj_covar=tensor([0.0362, 0.0507, 0.0559, 0.0430, 0.0577, 0.0586, 0.0443, 0.0583], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 01:12:04,260 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0282, 3.1335, 1.9038, 3.3206, 2.2404, 3.2945, 2.0082, 2.5182], device='cuda:0'), covar=tensor([0.0289, 0.0367, 0.1532, 0.0219, 0.0893, 0.0514, 0.1499, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0163, 0.0184, 0.0134, 0.0163, 0.0201, 0.0192, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 01:12:06,769 INFO [optim.py:368] (0/8) 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 01:12:18,458 INFO [train.py:904] (0/8) Epoch 14, batch 8650, loss[loss=0.1703, simple_loss=0.2613, pruned_loss=0.03966, over 12515.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2766, pruned_loss=0.04712, over 3008921.56 frames. ], batch size: 246, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:12:52,305 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:14:05,329 INFO [train.py:904] (0/8) Epoch 14, batch 8700, loss[loss=0.2004, simple_loss=0.2819, pruned_loss=0.05947, over 12918.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2741, pruned_loss=0.04573, over 3031074.33 frames. ], batch size: 248, lr: 4.81e-03, grad_scale: 8.0 2023-04-30 01:14:30,940 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6536, 4.0700, 3.6163, 3.9811, 3.5716, 3.6945, 3.6343, 4.0602], device='cuda:0'), covar=tensor([0.2659, 0.1874, 0.3270, 0.1622, 0.2011, 0.3059, 0.2496, 0.1904], device='cuda:0'), in_proj_covar=tensor([0.0571, 0.0702, 0.0578, 0.0506, 0.0444, 0.0458, 0.0589, 0.0533], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:14:42,302 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 01:15:06,603 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6000, 2.4824, 2.1961, 4.0593, 2.2655, 3.9119, 1.4292, 2.6814], device='cuda:0'), covar=tensor([0.1577, 0.0889, 0.1524, 0.0143, 0.0158, 0.0455, 0.1816, 0.0979], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0161, 0.0182, 0.0157, 0.0197, 0.0207, 0.0186, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 01:15:24,090 INFO [optim.py:368] (0/8) 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 01:15:38,751 INFO [train.py:904] (0/8) Epoch 14, batch 8750, loss[loss=0.1987, simple_loss=0.2952, pruned_loss=0.05108, over 16132.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2737, pruned_loss=0.04512, over 3041023.93 frames. ], batch size: 165, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:16:30,745 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140723.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:11,152 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140743.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:28,581 INFO [train.py:904] (0/8) Epoch 14, batch 8800, loss[loss=0.1886, simple_loss=0.2795, pruned_loss=0.04884, over 16519.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2722, pruned_loss=0.04424, over 3046188.80 frames. ], batch size: 147, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:17:43,127 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:17:57,044 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:18:18,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7520, 5.0836, 5.1835, 5.0171, 4.9894, 5.5926, 5.1469, 4.9316], device='cuda:0'), covar=tensor([0.0962, 0.1491, 0.1350, 0.1762, 0.2525, 0.0849, 0.1450, 0.2296], device='cuda:0'), in_proj_covar=tensor([0.0357, 0.0498, 0.0550, 0.0424, 0.0570, 0.0580, 0.0438, 0.0575], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 01:18:35,300 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:00,061 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.211e+02 2.894e+02 3.689e+02 7.406e+02, threshold=5.789e+02, percent-clipped=2.0 2023-04-30 01:19:12,357 INFO [train.py:904] (0/8) Epoch 14, batch 8850, loss[loss=0.1763, simple_loss=0.2786, pruned_loss=0.03702, over 15405.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2734, pruned_loss=0.04354, over 3013579.29 frames. ], batch size: 190, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:19:23,456 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:27,191 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140809.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:19:37,092 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:20:43,598 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140845.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:20:47,668 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0477, 4.0196, 3.9011, 3.3840, 3.9595, 1.7334, 3.7685, 3.6587], device='cuda:0'), covar=tensor([0.0083, 0.0073, 0.0158, 0.0222, 0.0083, 0.2398, 0.0113, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0124, 0.0169, 0.0154, 0.0142, 0.0186, 0.0157, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:20:58,219 INFO [train.py:904] (0/8) Epoch 14, batch 8900, loss[loss=0.1706, simple_loss=0.2729, pruned_loss=0.03417, over 16866.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.274, pruned_loss=0.04269, over 3043515.63 frames. ], batch size: 96, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:21:19,843 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1086, 5.0764, 4.9181, 4.5372, 4.5979, 4.9774, 4.8941, 4.6371], device='cuda:0'), covar=tensor([0.0478, 0.0385, 0.0261, 0.0275, 0.0952, 0.0408, 0.0250, 0.0610], device='cuda:0'), in_proj_covar=tensor([0.0245, 0.0334, 0.0295, 0.0271, 0.0303, 0.0319, 0.0202, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:21:25,126 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=140865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:21:34,672 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:22:46,891 INFO [optim.py:368] (0/8) 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 01:23:01,556 INFO [train.py:904] (0/8) Epoch 14, batch 8950, loss[loss=0.1831, simple_loss=0.272, pruned_loss=0.04709, over 15290.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2739, pruned_loss=0.04337, over 3052745.47 frames. ], batch size: 190, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:23:29,068 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=140915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:23:32,807 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:27,188 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2023-04-30 01:24:38,438 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=140947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:24:48,138 INFO [train.py:904] (0/8) Epoch 14, batch 9000, loss[loss=0.1603, simple_loss=0.2591, pruned_loss=0.03076, over 15258.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2708, pruned_loss=0.04169, over 3064915.87 frames. ], batch size: 190, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:24:48,139 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 01:24:58,095 INFO [train.py:938] (0/8) Epoch 14, validation: loss=0.1514, simple_loss=0.2553, pruned_loss=0.0237, over 944034.00 frames. 2023-04-30 01:24:58,096 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 01:25:21,243 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=140963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:25:52,474 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=140978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:26:29,793 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.243e+02 2.199e+02 2.701e+02 3.272e+02 7.252e+02, threshold=5.402e+02, percent-clipped=4.0 2023-04-30 01:26:39,930 INFO [train.py:904] (0/8) Epoch 14, batch 9050, loss[loss=0.1897, simple_loss=0.2837, pruned_loss=0.04781, over 17004.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2717, pruned_loss=0.0423, over 3069648.37 frames. ], batch size: 55, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:26:54,544 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141008.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:07,560 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141043.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:28:25,851 INFO [train.py:904] (0/8) Epoch 14, batch 9100, loss[loss=0.1758, simple_loss=0.2743, pruned_loss=0.03865, over 16159.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2719, pruned_loss=0.04297, over 3061174.82 frames. ], batch size: 165, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:29:26,140 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7701, 3.6588, 4.0267, 2.0004, 4.1984, 4.2684, 3.1812, 3.1580], device='cuda:0'), covar=tensor([0.0654, 0.0204, 0.0164, 0.1151, 0.0043, 0.0086, 0.0330, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0099, 0.0085, 0.0133, 0.0068, 0.0107, 0.0118, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 01:29:30,841 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:29:58,703 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:30:13,184 INFO [optim.py:368] (0/8) 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 01:30:22,853 INFO [train.py:904] (0/8) Epoch 14, batch 9150, loss[loss=0.1568, simple_loss=0.2566, pruned_loss=0.02849, over 16944.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2724, pruned_loss=0.04266, over 3053689.96 frames. ], batch size: 102, lr: 4.80e-03, grad_scale: 4.0 2023-04-30 01:30:49,990 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.41 vs. limit=5.0 2023-04-30 01:31:24,733 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0915, 2.6965, 2.8828, 2.0426, 2.6576, 2.1510, 2.7356, 2.8698], device='cuda:0'), covar=tensor([0.0302, 0.0830, 0.0526, 0.1729, 0.0805, 0.0906, 0.0594, 0.0731], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0145, 0.0156, 0.0144, 0.0136, 0.0124, 0.0135, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 01:31:51,273 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:31:54,383 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:04,858 INFO [train.py:904] (0/8) Epoch 14, batch 9200, loss[loss=0.1853, simple_loss=0.2777, pruned_loss=0.04641, over 15420.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2681, pruned_loss=0.04162, over 3082558.44 frames. ], batch size: 191, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:32:28,732 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:32:28,842 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:23,829 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:33:31,531 INFO [optim.py:368] (0/8) 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 01:33:40,997 INFO [train.py:904] (0/8) Epoch 14, batch 9250, loss[loss=0.164, simple_loss=0.2601, pruned_loss=0.03397, over 15469.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2681, pruned_loss=0.04152, over 3100657.87 frames. ], batch size: 191, lr: 4.80e-03, grad_scale: 8.0 2023-04-30 01:33:46,060 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:35:30,130 INFO [train.py:904] (0/8) Epoch 14, batch 9300, loss[loss=0.1581, simple_loss=0.2483, pruned_loss=0.03389, over 16198.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2664, pruned_loss=0.041, over 3092332.72 frames. ], batch size: 165, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:35:33,040 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9480, 4.9411, 4.5993, 4.0709, 4.7114, 1.8396, 4.4903, 4.6034], device='cuda:0'), covar=tensor([0.0072, 0.0065, 0.0189, 0.0325, 0.0085, 0.2366, 0.0123, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0123, 0.0166, 0.0151, 0.0141, 0.0183, 0.0155, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:35:59,517 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.65 vs. limit=2.0 2023-04-30 01:36:20,193 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:37:05,692 INFO [optim.py:368] (0/8) 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 01:37:14,532 INFO [train.py:904] (0/8) Epoch 14, batch 9350, loss[loss=0.1887, simple_loss=0.286, pruned_loss=0.04565, over 16207.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.266, pruned_loss=0.04049, over 3116819.52 frames. ], batch size: 165, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:37:16,941 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:08,691 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:38:29,244 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-04-30 01:38:54,843 INFO [train.py:904] (0/8) Epoch 14, batch 9400, loss[loss=0.1514, simple_loss=0.2382, pruned_loss=0.03235, over 12667.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2649, pruned_loss=0.0401, over 3085283.23 frames. ], batch size: 248, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:39:49,395 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:12,146 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:40:25,325 INFO [optim.py:368] (0/8) 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 01:40:31,042 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6554, 4.7046, 4.5154, 4.1676, 4.1678, 4.6137, 4.4206, 4.3309], device='cuda:0'), covar=tensor([0.0482, 0.0426, 0.0279, 0.0272, 0.0871, 0.0436, 0.0411, 0.0600], device='cuda:0'), in_proj_covar=tensor([0.0242, 0.0329, 0.0290, 0.0269, 0.0300, 0.0315, 0.0200, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-30 01:40:33,422 INFO [train.py:904] (0/8) Epoch 14, batch 9450, loss[loss=0.1789, simple_loss=0.2741, pruned_loss=0.04185, over 15305.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2666, pruned_loss=0.04047, over 3077544.77 frames. ], batch size: 190, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:41:03,051 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141417.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 01:41:24,366 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141427.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:41:33,129 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6438, 4.8138, 5.0037, 4.8212, 4.8716, 5.4070, 4.9243, 4.6073], device='cuda:0'), covar=tensor([0.0985, 0.1772, 0.1705, 0.1756, 0.2461, 0.0919, 0.1373, 0.2273], device='cuda:0'), in_proj_covar=tensor([0.0349, 0.0496, 0.0545, 0.0419, 0.0564, 0.0574, 0.0432, 0.0569], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 01:41:54,102 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141441.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:42:14,342 INFO [train.py:904] (0/8) Epoch 14, batch 9500, loss[loss=0.1416, simple_loss=0.2285, pruned_loss=0.02737, over 12656.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2661, pruned_loss=0.03999, over 3089471.91 frames. ], batch size: 248, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:42:17,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4390, 4.4227, 4.2757, 3.8518, 4.3140, 1.7327, 4.0860, 4.1323], device='cuda:0'), covar=tensor([0.0066, 0.0079, 0.0140, 0.0232, 0.0082, 0.2348, 0.0111, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0123, 0.0166, 0.0151, 0.0141, 0.0183, 0.0155, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:42:43,669 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:43:08,421 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141478.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 01:43:47,718 INFO [optim.py:368] (0/8) 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 01:43:55,179 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141499.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:43:59,701 INFO [train.py:904] (0/8) Epoch 14, batch 9550, loss[loss=0.1971, simple_loss=0.2924, pruned_loss=0.05086, over 16282.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2666, pruned_loss=0.04045, over 3094032.79 frames. ], batch size: 165, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:44:00,638 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141502.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:44:23,850 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:45:41,075 INFO [train.py:904] (0/8) Epoch 14, batch 9600, loss[loss=0.1794, simple_loss=0.2707, pruned_loss=0.0441, over 16662.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2675, pruned_loss=0.04105, over 3075359.36 frames. ], batch size: 62, lr: 4.79e-03, grad_scale: 8.0 2023-04-30 01:46:22,397 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:47:19,265 INFO [optim.py:368] (0/8) 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 01:47:30,279 INFO [train.py:904] (0/8) Epoch 14, batch 9650, loss[loss=0.1694, simple_loss=0.265, pruned_loss=0.03685, over 16908.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2695, pruned_loss=0.04154, over 3057389.66 frames. ], batch size: 102, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:47:34,534 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141603.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:48:16,274 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:49:16,992 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141651.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:49:17,848 INFO [train.py:904] (0/8) Epoch 14, batch 9700, loss[loss=0.1561, simple_loss=0.2536, pruned_loss=0.0293, over 16511.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2685, pruned_loss=0.04108, over 3074585.26 frames. ], batch size: 68, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:49:28,845 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5821, 3.5481, 3.4990, 2.9536, 3.4436, 2.0516, 3.2534, 2.9737], device='cuda:0'), covar=tensor([0.0112, 0.0090, 0.0141, 0.0181, 0.0083, 0.2080, 0.0108, 0.0163], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0124, 0.0167, 0.0151, 0.0141, 0.0185, 0.0156, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:50:28,042 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 01:51:00,295 INFO [train.py:904] (0/8) Epoch 14, batch 9750, loss[loss=0.1693, simple_loss=0.2639, pruned_loss=0.03735, over 15395.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2669, pruned_loss=0.04133, over 3053119.92 frames. ], batch size: 190, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:51:38,335 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9515, 3.1676, 3.0112, 2.0764, 2.9377, 3.2352, 3.0467, 1.6396], device='cuda:0'), covar=tensor([0.0491, 0.0052, 0.0064, 0.0395, 0.0098, 0.0077, 0.0078, 0.0584], device='cuda:0'), in_proj_covar=tensor([0.0126, 0.0068, 0.0070, 0.0125, 0.0082, 0.0091, 0.0080, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 01:51:43,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5019, 3.5855, 2.7968, 2.0974, 2.2211, 2.3339, 3.7228, 3.1661], device='cuda:0'), covar=tensor([0.2874, 0.0635, 0.1570, 0.2806, 0.2885, 0.1973, 0.0386, 0.1301], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0254, 0.0284, 0.0283, 0.0267, 0.0228, 0.0265, 0.0296], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 01:52:38,200 INFO [train.py:904] (0/8) Epoch 14, batch 9800, loss[loss=0.1653, simple_loss=0.2508, pruned_loss=0.03985, over 12252.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2675, pruned_loss=0.04055, over 3062552.23 frames. ], batch size: 247, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:53:18,074 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141773.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 01:54:11,652 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 01:54:13,234 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=141797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:54:13,945 INFO [optim.py:368] (0/8) 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 01:54:16,672 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141799.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:54:21,457 INFO [train.py:904] (0/8) Epoch 14, batch 9850, loss[loss=0.1736, simple_loss=0.275, pruned_loss=0.03613, over 16945.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2687, pruned_loss=0.04053, over 3069312.04 frames. ], batch size: 102, lr: 4.79e-03, grad_scale: 4.0 2023-04-30 01:54:39,243 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0008, 2.5964, 2.8035, 2.0604, 2.5826, 2.0475, 2.6112, 2.7662], device='cuda:0'), covar=tensor([0.0312, 0.0852, 0.0579, 0.1775, 0.0827, 0.1047, 0.0620, 0.0821], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0143, 0.0156, 0.0143, 0.0135, 0.0123, 0.0134, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 01:56:04,789 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=141847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:56:14,794 INFO [train.py:904] (0/8) Epoch 14, batch 9900, loss[loss=0.1663, simple_loss=0.2527, pruned_loss=0.03993, over 12680.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2691, pruned_loss=0.04058, over 3052086.89 frames. ], batch size: 250, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:57:39,928 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-30 01:58:03,748 INFO [optim.py:368] (0/8) 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 01:58:13,948 INFO [train.py:904] (0/8) Epoch 14, batch 9950, loss[loss=0.19, simple_loss=0.287, pruned_loss=0.04647, over 16415.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2704, pruned_loss=0.04089, over 3033212.82 frames. ], batch size: 146, lr: 4.78e-03, grad_scale: 4.0 2023-04-30 01:58:48,305 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:21,071 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=141928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 01:59:31,838 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 01:59:39,974 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8326, 1.6794, 2.2245, 2.7176, 2.4873, 3.0270, 2.0489, 3.1057], device='cuda:0'), covar=tensor([0.0153, 0.0461, 0.0297, 0.0231, 0.0277, 0.0146, 0.0391, 0.0119], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0175, 0.0158, 0.0161, 0.0172, 0.0128, 0.0175, 0.0118], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-30 02:00:00,563 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4239, 4.7097, 4.5225, 4.5380, 4.2667, 4.2292, 4.1970, 4.7246], device='cuda:0'), covar=tensor([0.1017, 0.0788, 0.0880, 0.0694, 0.0676, 0.1225, 0.1018, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0560, 0.0691, 0.0566, 0.0500, 0.0440, 0.0451, 0.0579, 0.0527], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:00:16,387 INFO [train.py:904] (0/8) Epoch 14, batch 10000, loss[loss=0.1427, simple_loss=0.2321, pruned_loss=0.02664, over 17198.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.269, pruned_loss=0.0403, over 3054656.49 frames. ], batch size: 46, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:01:06,750 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:25,925 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=141985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:34,082 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=141989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:01:50,851 INFO [optim.py:368] (0/8) 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 02:01:54,832 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-142000.pt 2023-04-30 02:02:01,167 INFO [train.py:904] (0/8) Epoch 14, batch 10050, loss[loss=0.1854, simple_loss=0.2866, pruned_loss=0.04204, over 15169.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2697, pruned_loss=0.04022, over 3054209.57 frames. ], batch size: 190, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:02:37,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5938, 3.8159, 2.8151, 2.1071, 2.4537, 2.3227, 4.0833, 3.3424], device='cuda:0'), covar=tensor([0.2885, 0.0731, 0.1834, 0.2942, 0.2883, 0.2006, 0.0449, 0.1313], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0252, 0.0283, 0.0280, 0.0263, 0.0226, 0.0263, 0.0293], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:03:00,955 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142033.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:03:34,650 INFO [train.py:904] (0/8) Epoch 14, batch 10100, loss[loss=0.1662, simple_loss=0.2615, pruned_loss=0.03547, over 16203.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2704, pruned_loss=0.04069, over 3048703.37 frames. ], batch size: 165, lr: 4.78e-03, grad_scale: 8.0 2023-04-30 02:04:19,935 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142073.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:04:49,717 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142097.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:04:50,460 INFO [optim.py:368] (0/8) 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 02:04:55,569 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-14.pt 2023-04-30 02:05:19,918 INFO [train.py:904] (0/8) Epoch 15, batch 0, loss[loss=0.2786, simple_loss=0.332, pruned_loss=0.1126, over 16471.00 frames. ], tot_loss[loss=0.2786, simple_loss=0.332, pruned_loss=0.1126, over 16471.00 frames. ], batch size: 146, lr: 4.62e-03, grad_scale: 8.0 2023-04-30 02:05:19,919 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 02:05:27,353 INFO [train.py:938] (0/8) Epoch 15, validation: loss=0.1501, simple_loss=0.2536, pruned_loss=0.02333, over 944034.00 frames. 2023-04-30 02:05:27,354 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 02:05:53,973 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142121.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:06:27,855 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 15, batch 50, loss[loss=0.1713, simple_loss=0.2669, pruned_loss=0.03785, over 17192.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2825, pruned_loss=0.05668, over 745324.03 frames. ], batch size: 45, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:07:44,552 INFO [optim.py:368] (0/8) 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 02:07:47,922 INFO [train.py:904] (0/8) Epoch 15, batch 100, loss[loss=0.1623, simple_loss=0.2453, pruned_loss=0.03968, over 16765.00 frames. ], tot_loss[loss=0.1926, simple_loss=0.2769, pruned_loss=0.05418, over 1319827.08 frames. ], batch size: 83, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:08:11,325 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 02:08:56,701 INFO [train.py:904] (0/8) Epoch 15, batch 150, loss[loss=0.1743, simple_loss=0.274, pruned_loss=0.03735, over 17259.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2739, pruned_loss=0.05268, over 1764769.06 frames. ], batch size: 52, lr: 4.62e-03, grad_scale: 2.0 2023-04-30 02:09:25,358 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:09:42,917 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:10:04,595 INFO [optim.py:368] (0/8) 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 02:10:07,779 INFO [train.py:904] (0/8) Epoch 15, batch 200, loss[loss=0.1945, simple_loss=0.2654, pruned_loss=0.06184, over 16990.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2727, pruned_loss=0.05326, over 2109746.06 frames. ], batch size: 41, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:10:24,359 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-04-30 02:10:45,774 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2007, 3.1312, 3.4097, 1.7871, 3.5111, 3.4600, 2.8363, 2.6650], device='cuda:0'), covar=tensor([0.0785, 0.0202, 0.0168, 0.1179, 0.0078, 0.0193, 0.0444, 0.0408], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0100, 0.0086, 0.0137, 0.0070, 0.0110, 0.0122, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0002, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 02:11:14,834 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 02:11:17,117 INFO [train.py:904] (0/8) Epoch 15, batch 250, loss[loss=0.1621, simple_loss=0.2541, pruned_loss=0.035, over 17160.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2701, pruned_loss=0.05222, over 2375861.37 frames. ], batch size: 46, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:11:18,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0067, 5.4239, 5.5116, 5.3056, 5.3855, 5.9619, 5.4131, 5.1036], device='cuda:0'), covar=tensor([0.0953, 0.1696, 0.2123, 0.2112, 0.2588, 0.1007, 0.1437, 0.2417], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0517, 0.0570, 0.0439, 0.0591, 0.0594, 0.0450, 0.0590], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 02:12:22,522 INFO [optim.py:368] (0/8) 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 02:12:25,443 INFO [train.py:904] (0/8) Epoch 15, batch 300, loss[loss=0.184, simple_loss=0.2587, pruned_loss=0.05462, over 16896.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2672, pruned_loss=0.05077, over 2582511.81 frames. ], batch size: 116, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:12:51,387 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4154, 5.3385, 5.1907, 4.7017, 4.7690, 5.2074, 5.2123, 4.8609], device='cuda:0'), covar=tensor([0.0545, 0.0419, 0.0305, 0.0307, 0.1091, 0.0424, 0.0282, 0.0739], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0348, 0.0304, 0.0283, 0.0317, 0.0331, 0.0209, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:13:36,236 INFO [train.py:904] (0/8) Epoch 15, batch 350, loss[loss=0.1781, simple_loss=0.264, pruned_loss=0.04607, over 17232.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2655, pruned_loss=0.04918, over 2756197.28 frames. ], batch size: 45, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:13:53,448 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9073, 3.0888, 3.2577, 2.0803, 2.7797, 2.3319, 3.4247, 3.3552], device='cuda:0'), covar=tensor([0.0245, 0.0875, 0.0574, 0.1679, 0.0805, 0.0899, 0.0495, 0.0840], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0147, 0.0158, 0.0145, 0.0137, 0.0125, 0.0136, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:14:12,767 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4021, 4.4223, 4.5441, 4.4200, 4.4224, 4.9920, 4.5124, 4.1828], device='cuda:0'), covar=tensor([0.1609, 0.1985, 0.2352, 0.2393, 0.3049, 0.1334, 0.1674, 0.2898], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0523, 0.0576, 0.0445, 0.0601, 0.0601, 0.0456, 0.0595], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 02:14:19,143 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9241, 4.2018, 3.0809, 2.2742, 2.7995, 2.5633, 4.5507, 3.6649], device='cuda:0'), covar=tensor([0.2394, 0.0588, 0.1560, 0.2469, 0.2502, 0.1853, 0.0354, 0.1142], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0257, 0.0288, 0.0286, 0.0273, 0.0231, 0.0269, 0.0304], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:14:36,208 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7774, 2.9216, 2.5236, 4.7059, 3.8984, 4.3204, 1.7510, 3.1934], device='cuda:0'), covar=tensor([0.1331, 0.0705, 0.1251, 0.0193, 0.0243, 0.0406, 0.1455, 0.0728], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0164, 0.0185, 0.0163, 0.0194, 0.0210, 0.0190, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:14:42,906 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.497e+02 2.253e+02 2.722e+02 3.272e+02 5.602e+02, threshold=5.443e+02, percent-clipped=1.0 2023-04-30 02:14:45,198 INFO [train.py:904] (0/8) Epoch 15, batch 400, loss[loss=0.1826, simple_loss=0.2743, pruned_loss=0.04539, over 17083.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2646, pruned_loss=0.04966, over 2882804.79 frames. ], batch size: 55, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:15:54,230 INFO [train.py:904] (0/8) Epoch 15, batch 450, loss[loss=0.1787, simple_loss=0.2721, pruned_loss=0.04265, over 17060.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2632, pruned_loss=0.04855, over 2986472.11 frames. ], batch size: 53, lr: 4.61e-03, grad_scale: 4.0 2023-04-30 02:16:02,688 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 02:16:17,087 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6484, 4.6065, 5.0428, 5.0259, 5.0340, 4.7313, 4.6474, 4.4907], device='cuda:0'), covar=tensor([0.0326, 0.0611, 0.0352, 0.0382, 0.0483, 0.0372, 0.0893, 0.0527], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0381, 0.0380, 0.0359, 0.0424, 0.0404, 0.0490, 0.0322], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 02:16:22,477 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0231, 4.4174, 3.2705, 2.3975, 2.8984, 2.5474, 4.7313, 3.7891], device='cuda:0'), covar=tensor([0.2356, 0.0538, 0.1475, 0.2509, 0.2553, 0.1928, 0.0342, 0.1188], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0260, 0.0291, 0.0289, 0.0277, 0.0234, 0.0272, 0.0308], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:16:23,443 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:24,666 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:16:40,920 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=142584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:03,320 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 15, batch 500, loss[loss=0.167, simple_loss=0.2491, pruned_loss=0.04248, over 16800.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2614, pruned_loss=0.0474, over 3067488.86 frames. ], batch size: 102, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:17:19,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7606, 2.3665, 2.4362, 4.6412, 2.3486, 2.8822, 2.4638, 2.5949], device='cuda:0'), covar=tensor([0.1010, 0.3555, 0.2568, 0.0354, 0.3861, 0.2313, 0.3244, 0.3418], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0410, 0.0344, 0.0320, 0.0420, 0.0470, 0.0376, 0.0478], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:17:28,779 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142620.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:45,636 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=142632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:49,342 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:17:52,905 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6274, 2.5523, 2.1811, 2.4522, 2.9588, 2.7944, 3.4497, 3.2589], device='cuda:0'), covar=tensor([0.0111, 0.0367, 0.0435, 0.0378, 0.0239, 0.0328, 0.0191, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0221, 0.0212, 0.0213, 0.0219, 0.0220, 0.0224, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:18:13,797 INFO [train.py:904] (0/8) Epoch 15, batch 550, loss[loss=0.1732, simple_loss=0.2509, pruned_loss=0.04778, over 12542.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.26, pruned_loss=0.04689, over 3117410.71 frames. ], batch size: 246, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:22,100 INFO [optim.py:368] (0/8) 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 02:19:23,211 INFO [train.py:904] (0/8) Epoch 15, batch 600, loss[loss=0.1823, simple_loss=0.2672, pruned_loss=0.04866, over 17233.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2584, pruned_loss=0.04568, over 3162161.48 frames. ], batch size: 44, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:19:53,070 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=142724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:20:32,871 INFO [train.py:904] (0/8) Epoch 15, batch 650, loss[loss=0.204, simple_loss=0.2921, pruned_loss=0.05795, over 17074.00 frames. ], tot_loss[loss=0.174, simple_loss=0.257, pruned_loss=0.0455, over 3198114.22 frames. ], batch size: 53, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:20:43,407 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-04-30 02:20:48,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2926, 2.3773, 1.6346, 2.0445, 2.7781, 2.4517, 3.1551, 3.0374], device='cuda:0'), covar=tensor([0.0171, 0.0424, 0.0575, 0.0473, 0.0288, 0.0401, 0.0247, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0220, 0.0211, 0.0212, 0.0219, 0.0220, 0.0224, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:21:14,252 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5811, 5.9700, 5.7355, 5.7549, 5.3710, 5.3263, 5.3673, 6.1031], device='cuda:0'), covar=tensor([0.1508, 0.0956, 0.1075, 0.0810, 0.0957, 0.0680, 0.1167, 0.1035], device='cuda:0'), in_proj_covar=tensor([0.0605, 0.0750, 0.0612, 0.0539, 0.0473, 0.0482, 0.0629, 0.0569], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:21:17,849 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=142785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:21:18,165 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 02:21:19,147 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5581, 3.7854, 4.2417, 2.4903, 3.4502, 2.7393, 4.0547, 3.9828], device='cuda:0'), covar=tensor([0.0271, 0.0888, 0.0457, 0.1778, 0.0760, 0.0990, 0.0593, 0.1077], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0150, 0.0161, 0.0147, 0.0139, 0.0127, 0.0139, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:21:27,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6897, 2.4806, 1.9189, 2.2405, 2.8434, 2.5965, 2.8885, 2.9783], device='cuda:0'), covar=tensor([0.0185, 0.0329, 0.0449, 0.0385, 0.0200, 0.0299, 0.0206, 0.0226], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0220, 0.0211, 0.0213, 0.0219, 0.0220, 0.0224, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:21:40,883 INFO [optim.py:368] (0/8) 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 02:21:42,130 INFO [train.py:904] (0/8) Epoch 15, batch 700, loss[loss=0.1662, simple_loss=0.2563, pruned_loss=0.03804, over 17112.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2569, pruned_loss=0.04477, over 3231128.59 frames. ], batch size: 47, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:22:31,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7542, 2.7542, 2.3337, 2.7208, 3.1166, 2.9487, 3.5658, 3.3974], device='cuda:0'), covar=tensor([0.0098, 0.0342, 0.0422, 0.0372, 0.0246, 0.0330, 0.0194, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0221, 0.0212, 0.0213, 0.0220, 0.0221, 0.0225, 0.0211], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:22:50,284 INFO [train.py:904] (0/8) Epoch 15, batch 750, loss[loss=0.2318, simple_loss=0.298, pruned_loss=0.08281, over 12530.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.257, pruned_loss=0.04444, over 3246410.15 frames. ], batch size: 247, lr: 4.61e-03, grad_scale: 2.0 2023-04-30 02:23:57,718 INFO [optim.py:368] (0/8) 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 02:23:59,535 INFO [train.py:904] (0/8) Epoch 15, batch 800, loss[loss=0.193, simple_loss=0.2667, pruned_loss=0.05965, over 16400.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2564, pruned_loss=0.04486, over 3268348.12 frames. ], batch size: 146, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:24:36,483 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=142929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:25:08,654 INFO [train.py:904] (0/8) Epoch 15, batch 850, loss[loss=0.1595, simple_loss=0.2454, pruned_loss=0.03675, over 16860.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2555, pruned_loss=0.04438, over 3287733.48 frames. ], batch size: 96, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:26:15,124 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 15, batch 900, loss[loss=0.1635, simple_loss=0.2419, pruned_loss=0.04254, over 16277.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2551, pruned_loss=0.04443, over 3295025.73 frames. ], batch size: 165, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:27:16,611 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8377, 3.8551, 4.2058, 2.0930, 4.3433, 4.4392, 3.0864, 3.4950], device='cuda:0'), covar=tensor([0.0656, 0.0199, 0.0203, 0.1149, 0.0072, 0.0161, 0.0427, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0104, 0.0090, 0.0140, 0.0073, 0.0116, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 02:27:24,768 INFO [train.py:904] (0/8) Epoch 15, batch 950, loss[loss=0.1866, simple_loss=0.2583, pruned_loss=0.05739, over 16872.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2561, pruned_loss=0.04446, over 3298265.40 frames. ], batch size: 116, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:28:01,855 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143080.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:28:29,461 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2632, 4.3353, 4.6552, 4.6553, 4.6907, 4.3548, 4.3907, 4.2239], device='cuda:0'), covar=tensor([0.0369, 0.0700, 0.0458, 0.0452, 0.0524, 0.0486, 0.0870, 0.0612], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0396, 0.0393, 0.0372, 0.0441, 0.0420, 0.0506, 0.0333], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 02:28:30,230 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 15, batch 1000, loss[loss=0.1755, simple_loss=0.2624, pruned_loss=0.04434, over 16753.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2547, pruned_loss=0.0443, over 3309879.64 frames. ], batch size: 62, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:29:41,408 INFO [train.py:904] (0/8) Epoch 15, batch 1050, loss[loss=0.1628, simple_loss=0.2544, pruned_loss=0.03558, over 16734.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2545, pruned_loss=0.0439, over 3321169.09 frames. ], batch size: 57, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:30:46,963 INFO [optim.py:368] (0/8) 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 02:30:49,005 INFO [train.py:904] (0/8) Epoch 15, batch 1100, loss[loss=0.1743, simple_loss=0.255, pruned_loss=0.04681, over 16241.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2546, pruned_loss=0.04418, over 3323504.65 frames. ], batch size: 165, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:31:25,825 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143229.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:31:31,130 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 02:31:40,489 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4705, 5.4041, 5.2613, 4.7777, 4.8926, 5.3010, 5.2709, 4.9300], device='cuda:0'), covar=tensor([0.0576, 0.0378, 0.0273, 0.0326, 0.1111, 0.0397, 0.0268, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0272, 0.0372, 0.0323, 0.0304, 0.0339, 0.0354, 0.0222, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:31:58,395 INFO [train.py:904] (0/8) Epoch 15, batch 1150, loss[loss=0.1526, simple_loss=0.2455, pruned_loss=0.02989, over 17094.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.254, pruned_loss=0.04385, over 3327404.43 frames. ], batch size: 47, lr: 4.60e-03, grad_scale: 4.0 2023-04-30 02:32:02,127 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9149, 3.0749, 3.0017, 5.1662, 4.2260, 4.6299, 1.9445, 3.4197], device='cuda:0'), covar=tensor([0.1299, 0.0700, 0.1001, 0.0150, 0.0266, 0.0383, 0.1391, 0.0663], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0166, 0.0187, 0.0167, 0.0198, 0.0214, 0.0191, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:32:06,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6625, 2.4613, 2.3882, 4.1336, 3.3617, 4.0847, 1.4688, 2.8807], device='cuda:0'), covar=tensor([0.1473, 0.0740, 0.1123, 0.0170, 0.0186, 0.0354, 0.1544, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0166, 0.0187, 0.0167, 0.0198, 0.0214, 0.0191, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:32:34,546 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:33:01,283 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8278, 3.8518, 4.2881, 1.9611, 4.4727, 4.4874, 3.1430, 3.5079], device='cuda:0'), covar=tensor([0.0690, 0.0207, 0.0182, 0.1208, 0.0058, 0.0131, 0.0417, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0105, 0.0090, 0.0141, 0.0073, 0.0116, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 02:33:07,202 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.162e+02 2.475e+02 3.140e+02 4.983e+02, threshold=4.950e+02, percent-clipped=0.0 2023-04-30 02:33:08,294 INFO [train.py:904] (0/8) Epoch 15, batch 1200, loss[loss=0.1669, simple_loss=0.2456, pruned_loss=0.04408, over 16752.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2535, pruned_loss=0.04308, over 3338444.58 frames. ], batch size: 134, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:33:11,180 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8352, 3.8829, 2.3558, 4.5262, 2.8831, 4.4312, 2.6378, 3.1625], device='cuda:0'), covar=tensor([0.0246, 0.0364, 0.1564, 0.0282, 0.0827, 0.0537, 0.1386, 0.0664], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0171, 0.0192, 0.0147, 0.0170, 0.0213, 0.0200, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:33:41,638 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.45 vs. limit=5.0 2023-04-30 02:33:43,192 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5142, 3.6344, 3.9549, 1.8848, 4.0551, 3.9966, 3.1415, 2.9931], device='cuda:0'), covar=tensor([0.0735, 0.0186, 0.0152, 0.1159, 0.0073, 0.0176, 0.0355, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0104, 0.0089, 0.0140, 0.0073, 0.0116, 0.0126, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 02:33:44,423 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1024, 4.8381, 5.0766, 5.2945, 5.5160, 4.7874, 5.4351, 5.4918], device='cuda:0'), covar=tensor([0.1655, 0.1334, 0.1702, 0.0702, 0.0502, 0.0837, 0.0446, 0.0497], device='cuda:0'), in_proj_covar=tensor([0.0594, 0.0730, 0.0872, 0.0746, 0.0559, 0.0579, 0.0591, 0.0695], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:34:16,303 INFO [train.py:904] (0/8) Epoch 15, batch 1250, loss[loss=0.1765, simple_loss=0.2641, pruned_loss=0.04446, over 17083.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2536, pruned_loss=0.04347, over 3328502.07 frames. ], batch size: 53, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:34:56,636 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=143380.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:35:25,398 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143400.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:35:26,165 INFO [optim.py:368] (0/8) 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 02:35:27,821 INFO [train.py:904] (0/8) Epoch 15, batch 1300, loss[loss=0.1852, simple_loss=0.2637, pruned_loss=0.05334, over 16775.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2531, pruned_loss=0.04361, over 3320234.67 frames. ], batch size: 102, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:03,861 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=143428.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:36:05,528 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 02:36:36,276 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 02:36:37,210 INFO [train.py:904] (0/8) Epoch 15, batch 1350, loss[loss=0.1862, simple_loss=0.2763, pruned_loss=0.04808, over 17020.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2532, pruned_loss=0.04323, over 3318778.26 frames. ], batch size: 53, lr: 4.60e-03, grad_scale: 8.0 2023-04-30 02:36:39,921 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9927, 4.8264, 4.9756, 5.2370, 5.4256, 4.7621, 5.3776, 5.3960], device='cuda:0'), covar=tensor([0.1964, 0.1486, 0.1967, 0.0900, 0.0700, 0.0860, 0.0563, 0.0707], device='cuda:0'), in_proj_covar=tensor([0.0604, 0.0742, 0.0889, 0.0759, 0.0568, 0.0589, 0.0599, 0.0707], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:36:49,875 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143461.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:37:06,516 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=143473.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:37:45,745 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.266e+02 2.683e+02 3.217e+02 5.542e+02, threshold=5.366e+02, percent-clipped=1.0 2023-04-30 02:37:47,548 INFO [train.py:904] (0/8) Epoch 15, batch 1400, loss[loss=0.2061, simple_loss=0.2912, pruned_loss=0.06048, over 17041.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2542, pruned_loss=0.04363, over 3328100.01 frames. ], batch size: 53, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:38:31,332 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=143534.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:38:46,931 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-04-30 02:38:56,003 INFO [train.py:904] (0/8) Epoch 15, batch 1450, loss[loss=0.1629, simple_loss=0.2376, pruned_loss=0.04412, over 16777.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2531, pruned_loss=0.04356, over 3322024.98 frames. ], batch size: 102, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:39:59,421 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-30 02:40:05,552 INFO [optim.py:368] (0/8) 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 02:40:06,737 INFO [train.py:904] (0/8) Epoch 15, batch 1500, loss[loss=0.1715, simple_loss=0.2633, pruned_loss=0.03981, over 17027.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2535, pruned_loss=0.04431, over 3317450.10 frames. ], batch size: 50, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:14,538 INFO [train.py:904] (0/8) Epoch 15, batch 1550, loss[loss=0.159, simple_loss=0.2519, pruned_loss=0.03309, over 17127.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2555, pruned_loss=0.04558, over 3321763.60 frames. ], batch size: 48, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:41:57,429 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 02:42:22,873 INFO [optim.py:368] (0/8) 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 02:42:24,071 INFO [train.py:904] (0/8) Epoch 15, batch 1600, loss[loss=0.1927, simple_loss=0.2719, pruned_loss=0.05673, over 16711.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2579, pruned_loss=0.04625, over 3317782.27 frames. ], batch size: 134, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:00,112 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8838, 2.9900, 2.6170, 4.5306, 3.6799, 4.2629, 1.7005, 3.0715], device='cuda:0'), covar=tensor([0.1303, 0.0608, 0.1070, 0.0167, 0.0223, 0.0372, 0.1416, 0.0796], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0165, 0.0185, 0.0167, 0.0198, 0.0214, 0.0189, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:43:35,435 INFO [train.py:904] (0/8) Epoch 15, batch 1650, loss[loss=0.187, simple_loss=0.263, pruned_loss=0.05549, over 16846.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2588, pruned_loss=0.04665, over 3318248.36 frames. ], batch size: 102, lr: 4.59e-03, grad_scale: 8.0 2023-04-30 02:43:40,917 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143756.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:44:46,128 INFO [optim.py:368] (0/8) 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 02:44:46,150 INFO [train.py:904] (0/8) Epoch 15, batch 1700, loss[loss=0.1854, simple_loss=0.2785, pruned_loss=0.04617, over 17058.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2607, pruned_loss=0.04752, over 3311670.39 frames. ], batch size: 50, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:45:05,692 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 02:45:22,508 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=143829.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 02:45:54,344 INFO [train.py:904] (0/8) Epoch 15, batch 1750, loss[loss=0.1806, simple_loss=0.2713, pruned_loss=0.0449, over 16657.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2603, pruned_loss=0.04697, over 3320535.04 frames. ], batch size: 62, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:47:05,601 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 15, batch 1800, loss[loss=0.1649, simple_loss=0.249, pruned_loss=0.0404, over 17215.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2618, pruned_loss=0.04754, over 3321901.82 frames. ], batch size: 44, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:47:53,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4257, 2.1606, 2.3375, 4.1990, 2.2569, 2.6553, 2.2659, 2.4559], device='cuda:0'), covar=tensor([0.1209, 0.3697, 0.2598, 0.0545, 0.3663, 0.2361, 0.3604, 0.2857], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0416, 0.0349, 0.0327, 0.0423, 0.0480, 0.0382, 0.0487], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:48:15,544 INFO [train.py:904] (0/8) Epoch 15, batch 1850, loss[loss=0.1493, simple_loss=0.2387, pruned_loss=0.02991, over 17227.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2622, pruned_loss=0.04681, over 3321572.54 frames. ], batch size: 45, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:48:26,603 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 02:48:39,422 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9821, 5.0528, 5.4384, 5.4569, 5.4609, 5.1333, 5.0362, 4.8188], device='cuda:0'), covar=tensor([0.0288, 0.0445, 0.0402, 0.0378, 0.0456, 0.0330, 0.0848, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0399, 0.0396, 0.0376, 0.0440, 0.0419, 0.0508, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 02:48:48,166 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-04-30 02:49:07,026 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-30 02:49:23,945 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-144000.pt 2023-04-30 02:49:30,767 INFO [train.py:904] (0/8) Epoch 15, batch 1900, loss[loss=0.1822, simple_loss=0.2647, pruned_loss=0.04985, over 16912.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2619, pruned_loss=0.04638, over 3318729.92 frames. ], batch size: 109, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:49:31,844 INFO [optim.py:368] (0/8) 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 02:49:55,012 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 02:50:01,297 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4618, 2.8316, 2.9302, 1.9499, 2.6983, 2.1300, 3.0685, 3.0906], device='cuda:0'), covar=tensor([0.0282, 0.0787, 0.0591, 0.1753, 0.0793, 0.0956, 0.0588, 0.0804], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0152, 0.0162, 0.0146, 0.0139, 0.0125, 0.0139, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:50:39,888 INFO [train.py:904] (0/8) Epoch 15, batch 1950, loss[loss=0.1929, simple_loss=0.2851, pruned_loss=0.05031, over 17091.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2623, pruned_loss=0.04597, over 3315768.21 frames. ], batch size: 53, lr: 4.59e-03, grad_scale: 2.0 2023-04-30 02:50:46,117 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144056.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:51:49,560 INFO [train.py:904] (0/8) Epoch 15, batch 2000, loss[loss=0.1815, simple_loss=0.2558, pruned_loss=0.05365, over 16767.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2621, pruned_loss=0.04596, over 3307867.09 frames. ], batch size: 83, lr: 4.59e-03, grad_scale: 4.0 2023-04-30 02:51:51,368 INFO [optim.py:368] (0/8) 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 02:51:52,654 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144104.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 02:52:22,491 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9983, 3.3084, 3.1469, 2.1548, 2.7508, 2.4025, 3.3414, 3.5550], device='cuda:0'), covar=tensor([0.0324, 0.0867, 0.0592, 0.1640, 0.0837, 0.0915, 0.0655, 0.0811], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0153, 0.0162, 0.0147, 0.0139, 0.0126, 0.0139, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 02:52:27,258 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=144129.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:52:58,038 INFO [train.py:904] (0/8) Epoch 15, batch 2050, loss[loss=0.2085, simple_loss=0.2742, pruned_loss=0.07138, over 16701.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2622, pruned_loss=0.04637, over 3321146.92 frames. ], batch size: 134, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:53:32,907 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=144177.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 02:54:07,934 INFO [train.py:904] (0/8) Epoch 15, batch 2100, loss[loss=0.1867, simple_loss=0.2745, pruned_loss=0.04946, over 15497.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2639, pruned_loss=0.04732, over 3313816.21 frames. ], batch size: 190, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:54:08,982 INFO [optim.py:368] (0/8) 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 02:55:15,292 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3973, 2.3033, 1.7926, 2.0743, 2.6166, 2.3503, 2.5726, 2.7089], device='cuda:0'), covar=tensor([0.0197, 0.0303, 0.0405, 0.0366, 0.0175, 0.0276, 0.0181, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0223, 0.0215, 0.0215, 0.0225, 0.0224, 0.0230, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:55:17,941 INFO [train.py:904] (0/8) Epoch 15, batch 2150, loss[loss=0.1513, simple_loss=0.2414, pruned_loss=0.03062, over 17237.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2645, pruned_loss=0.04771, over 3320073.27 frames. ], batch size: 43, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:55:37,419 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0998, 5.0852, 4.8801, 4.2755, 4.9349, 1.6360, 4.6607, 4.7876], device='cuda:0'), covar=tensor([0.0087, 0.0076, 0.0179, 0.0406, 0.0095, 0.2955, 0.0137, 0.0201], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0138, 0.0188, 0.0171, 0.0158, 0.0199, 0.0174, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:56:20,020 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0891, 3.9639, 4.1658, 4.2953, 4.3739, 3.9404, 4.1096, 4.3669], device='cuda:0'), covar=tensor([0.1495, 0.1111, 0.1206, 0.0625, 0.0574, 0.1402, 0.2930, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0607, 0.0750, 0.0901, 0.0773, 0.0577, 0.0599, 0.0609, 0.0715], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 02:56:25,365 INFO [train.py:904] (0/8) Epoch 15, batch 2200, loss[loss=0.2035, simple_loss=0.2666, pruned_loss=0.07019, over 16550.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2645, pruned_loss=0.04776, over 3319788.80 frames. ], batch size: 146, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:56:27,078 INFO [optim.py:368] (0/8) 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 02:56:52,936 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 02:57:36,228 INFO [train.py:904] (0/8) Epoch 15, batch 2250, loss[loss=0.1592, simple_loss=0.2446, pruned_loss=0.03692, over 17172.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.264, pruned_loss=0.04714, over 3330265.92 frames. ], batch size: 46, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:37,050 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 02:58:46,693 INFO [train.py:904] (0/8) Epoch 15, batch 2300, loss[loss=0.175, simple_loss=0.2665, pruned_loss=0.04177, over 17222.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.265, pruned_loss=0.0476, over 3332898.51 frames. ], batch size: 45, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 02:58:47,878 INFO [optim.py:368] (0/8) 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 02:58:58,071 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144411.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 02:59:53,217 INFO [train.py:904] (0/8) Epoch 15, batch 2350, loss[loss=0.1827, simple_loss=0.2734, pruned_loss=0.04598, over 17212.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2648, pruned_loss=0.04735, over 3332044.45 frames. ], batch size: 45, lr: 4.58e-03, grad_scale: 4.0 2023-04-30 03:00:20,676 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=144472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:01:02,749 INFO [train.py:904] (0/8) Epoch 15, batch 2400, loss[loss=0.16, simple_loss=0.2535, pruned_loss=0.03325, over 17205.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2655, pruned_loss=0.04777, over 3337142.53 frames. ], batch size: 46, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:01:04,727 INFO [optim.py:368] (0/8) 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 03:01:22,909 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 03:02:09,931 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-30 03:02:10,324 INFO [train.py:904] (0/8) Epoch 15, batch 2450, loss[loss=0.1732, simple_loss=0.2548, pruned_loss=0.0458, over 16132.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2665, pruned_loss=0.04781, over 3339366.30 frames. ], batch size: 164, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:02:44,164 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4839, 2.2062, 2.3252, 4.4681, 2.1934, 2.7087, 2.3243, 2.4000], device='cuda:0'), covar=tensor([0.1117, 0.3741, 0.2727, 0.0422, 0.3973, 0.2473, 0.3417, 0.3554], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0417, 0.0349, 0.0327, 0.0423, 0.0480, 0.0382, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:03:17,681 INFO [train.py:904] (0/8) Epoch 15, batch 2500, loss[loss=0.1526, simple_loss=0.233, pruned_loss=0.03611, over 16782.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.266, pruned_loss=0.04774, over 3335495.30 frames. ], batch size: 39, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:03:18,674 INFO [optim.py:368] (0/8) 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 03:03:19,142 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2671, 2.0433, 2.7252, 3.2031, 2.9251, 3.5666, 2.1103, 3.5483], device='cuda:0'), covar=tensor([0.0150, 0.0414, 0.0237, 0.0207, 0.0221, 0.0144, 0.0469, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0183, 0.0169, 0.0172, 0.0181, 0.0138, 0.0183, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:04:26,857 INFO [train.py:904] (0/8) Epoch 15, batch 2550, loss[loss=0.1773, simple_loss=0.2736, pruned_loss=0.04049, over 17139.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2662, pruned_loss=0.04766, over 3321999.90 frames. ], batch size: 48, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:34,895 INFO [train.py:904] (0/8) Epoch 15, batch 2600, loss[loss=0.1751, simple_loss=0.2694, pruned_loss=0.04043, over 17127.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2655, pruned_loss=0.04709, over 3328536.46 frames. ], batch size: 48, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:05:36,053 INFO [optim.py:368] (0/8) 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 03:05:36,403 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1795, 3.9813, 4.2173, 4.3720, 4.4539, 4.0371, 4.2332, 4.4291], device='cuda:0'), covar=tensor([0.1491, 0.1158, 0.1331, 0.0664, 0.0615, 0.1222, 0.1613, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0611, 0.0755, 0.0903, 0.0775, 0.0581, 0.0604, 0.0610, 0.0716], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:06:43,560 INFO [train.py:904] (0/8) Epoch 15, batch 2650, loss[loss=0.179, simple_loss=0.2715, pruned_loss=0.04326, over 17043.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2653, pruned_loss=0.04639, over 3329374.78 frames. ], batch size: 50, lr: 4.58e-03, grad_scale: 8.0 2023-04-30 03:07:05,953 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=144767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:07:50,275 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2284, 3.9730, 4.4144, 2.1726, 4.6248, 4.6826, 3.2409, 3.6546], device='cuda:0'), covar=tensor([0.0635, 0.0224, 0.0193, 0.1122, 0.0067, 0.0134, 0.0436, 0.0344], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0105, 0.0091, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 03:07:53,566 INFO [train.py:904] (0/8) Epoch 15, batch 2700, loss[loss=0.1705, simple_loss=0.2644, pruned_loss=0.03827, over 16446.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2656, pruned_loss=0.04619, over 3324686.96 frames. ], batch size: 68, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:07:54,731 INFO [optim.py:368] (0/8) 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 03:08:57,046 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6889, 2.2912, 2.4023, 4.7290, 2.3242, 2.7847, 2.3860, 2.5301], device='cuda:0'), covar=tensor([0.1037, 0.3620, 0.2646, 0.0395, 0.3859, 0.2374, 0.3184, 0.3536], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0416, 0.0347, 0.0327, 0.0422, 0.0480, 0.0380, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:09:02,454 INFO [train.py:904] (0/8) Epoch 15, batch 2750, loss[loss=0.1789, simple_loss=0.2632, pruned_loss=0.04728, over 16801.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.266, pruned_loss=0.04578, over 3330835.90 frames. ], batch size: 83, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:09:04,193 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0232, 3.1472, 2.6400, 5.0294, 3.8411, 4.3457, 2.1997, 3.1627], device='cuda:0'), covar=tensor([0.1353, 0.0787, 0.1351, 0.0175, 0.0432, 0.0430, 0.1367, 0.0896], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0170, 0.0201, 0.0214, 0.0189, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:10:11,032 INFO [train.py:904] (0/8) Epoch 15, batch 2800, loss[loss=0.1655, simple_loss=0.2564, pruned_loss=0.03731, over 17230.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2653, pruned_loss=0.04554, over 3336033.83 frames. ], batch size: 44, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:10:12,144 INFO [optim.py:368] (0/8) 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 03:10:26,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3799, 2.2115, 1.6915, 2.0582, 2.5925, 2.3617, 2.5837, 2.7167], device='cuda:0'), covar=tensor([0.0188, 0.0348, 0.0478, 0.0399, 0.0220, 0.0289, 0.0194, 0.0229], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0224, 0.0213, 0.0215, 0.0225, 0.0223, 0.0230, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:11:21,063 INFO [train.py:904] (0/8) Epoch 15, batch 2850, loss[loss=0.1659, simple_loss=0.2443, pruned_loss=0.04372, over 16643.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2643, pruned_loss=0.04548, over 3330233.35 frames. ], batch size: 89, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:22,125 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=144996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:12:31,824 INFO [train.py:904] (0/8) Epoch 15, batch 2900, loss[loss=0.1621, simple_loss=0.2359, pruned_loss=0.04411, over 16931.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2638, pruned_loss=0.04566, over 3328573.71 frames. ], batch size: 90, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:12:33,023 INFO [optim.py:368] (0/8) 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 03:13:06,357 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:13:38,100 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 03:13:40,928 INFO [train.py:904] (0/8) Epoch 15, batch 2950, loss[loss=0.2069, simple_loss=0.2722, pruned_loss=0.07085, over 16723.00 frames. ], tot_loss[loss=0.179, simple_loss=0.264, pruned_loss=0.04702, over 3321516.60 frames. ], batch size: 124, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:13:47,652 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145057.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:14:01,032 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:02,210 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145067.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:31,832 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:14:49,684 INFO [train.py:904] (0/8) Epoch 15, batch 3000, loss[loss=0.176, simple_loss=0.2641, pruned_loss=0.04391, over 16681.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.265, pruned_loss=0.04805, over 3319652.91 frames. ], batch size: 83, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:14:49,685 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 03:14:58,802 INFO [train.py:938] (0/8) Epoch 15, validation: loss=0.138, simple_loss=0.2438, pruned_loss=0.01616, over 944034.00 frames. 2023-04-30 03:14:58,802 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 03:15:00,802 INFO [optim.py:368] (0/8) 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 03:15:01,372 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0733, 4.6110, 3.4438, 2.6089, 3.1358, 2.9138, 4.9279, 3.9760], device='cuda:0'), covar=tensor([0.2447, 0.0480, 0.1494, 0.2216, 0.2316, 0.1595, 0.0269, 0.1014], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0265, 0.0294, 0.0293, 0.0287, 0.0238, 0.0279, 0.0318], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 03:15:04,446 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.68 vs. limit=5.0 2023-04-30 03:15:16,807 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145115.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:15:34,512 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145127.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:15:43,445 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 03:16:07,552 INFO [train.py:904] (0/8) Epoch 15, batch 3050, loss[loss=0.1781, simple_loss=0.2755, pruned_loss=0.04035, over 17012.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2644, pruned_loss=0.04751, over 3320756.60 frames. ], batch size: 50, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:16:38,954 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9440, 1.9305, 2.4506, 2.8052, 2.7504, 2.9057, 1.9780, 3.0566], device='cuda:0'), covar=tensor([0.0143, 0.0428, 0.0270, 0.0237, 0.0250, 0.0210, 0.0416, 0.0128], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0186, 0.0171, 0.0175, 0.0184, 0.0140, 0.0186, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:17:03,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9918, 3.1850, 2.9490, 5.1711, 4.3787, 4.6541, 1.7430, 3.5134], device='cuda:0'), covar=tensor([0.1256, 0.0670, 0.1044, 0.0173, 0.0233, 0.0344, 0.1458, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0171, 0.0201, 0.0213, 0.0189, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:17:17,942 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 03:17:18,180 INFO [train.py:904] (0/8) Epoch 15, batch 3100, loss[loss=0.1862, simple_loss=0.2667, pruned_loss=0.05288, over 16855.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2638, pruned_loss=0.04761, over 3325830.33 frames. ], batch size: 83, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:17:19,339 INFO [optim.py:368] (0/8) 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 03:18:28,441 INFO [train.py:904] (0/8) Epoch 15, batch 3150, loss[loss=0.1536, simple_loss=0.2382, pruned_loss=0.03454, over 16747.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2627, pruned_loss=0.04719, over 3321589.40 frames. ], batch size: 39, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:18:34,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8033, 4.9878, 5.1283, 4.9631, 4.9556, 5.5934, 5.1453, 4.8467], device='cuda:0'), covar=tensor([0.1222, 0.1718, 0.1969, 0.1982, 0.2625, 0.0962, 0.1397, 0.2349], device='cuda:0'), in_proj_covar=tensor([0.0389, 0.0553, 0.0607, 0.0470, 0.0631, 0.0633, 0.0479, 0.0626], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 03:18:57,440 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8095, 3.9973, 2.1214, 4.5602, 2.9714, 4.4258, 1.9782, 2.9476], device='cuda:0'), covar=tensor([0.0252, 0.0295, 0.1655, 0.0223, 0.0719, 0.0361, 0.1848, 0.0723], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0172, 0.0190, 0.0151, 0.0169, 0.0215, 0.0199, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:19:37,247 INFO [train.py:904] (0/8) Epoch 15, batch 3200, loss[loss=0.1906, simple_loss=0.276, pruned_loss=0.05259, over 17079.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2622, pruned_loss=0.04659, over 3325839.81 frames. ], batch size: 53, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:19:38,469 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.241e+02 2.735e+02 3.234e+02 5.514e+02, threshold=5.469e+02, percent-clipped=0.0 2023-04-30 03:20:46,520 INFO [train.py:904] (0/8) Epoch 15, batch 3250, loss[loss=0.2084, simple_loss=0.2876, pruned_loss=0.06456, over 16817.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2621, pruned_loss=0.04619, over 3329964.21 frames. ], batch size: 102, lr: 4.57e-03, grad_scale: 8.0 2023-04-30 03:20:46,750 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145352.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:21:13,348 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:30,579 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145383.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:21:57,382 INFO [train.py:904] (0/8) Epoch 15, batch 3300, loss[loss=0.2078, simple_loss=0.3006, pruned_loss=0.05752, over 17106.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2636, pruned_loss=0.0467, over 3306897.14 frames. ], batch size: 49, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:21:58,628 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.241e+02 2.723e+02 3.266e+02 5.157e+02, threshold=5.447e+02, percent-clipped=0.0 2023-04-30 03:22:25,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:22:38,847 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:22:58,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3827, 3.6604, 3.8654, 2.1287, 3.0770, 2.3994, 3.8360, 3.7499], device='cuda:0'), covar=tensor([0.0304, 0.0884, 0.0487, 0.1842, 0.0811, 0.0941, 0.0636, 0.1068], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0155, 0.0162, 0.0147, 0.0139, 0.0126, 0.0140, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:23:06,158 INFO [train.py:904] (0/8) Epoch 15, batch 3350, loss[loss=0.1639, simple_loss=0.2525, pruned_loss=0.03764, over 17241.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2644, pruned_loss=0.04706, over 3299369.28 frames. ], batch size: 45, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:17,515 INFO [train.py:904] (0/8) Epoch 15, batch 3400, loss[loss=0.2108, simple_loss=0.2885, pruned_loss=0.06649, over 16415.00 frames. ], tot_loss[loss=0.179, simple_loss=0.264, pruned_loss=0.04704, over 3300697.66 frames. ], batch size: 146, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:24:18,612 INFO [optim.py:368] (0/8) 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 03:25:13,773 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 03:25:15,484 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2478, 4.9969, 5.2519, 5.4170, 5.6397, 4.9108, 5.5571, 5.5890], device='cuda:0'), covar=tensor([0.1669, 0.1186, 0.1606, 0.0713, 0.0481, 0.0773, 0.0435, 0.0511], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0768, 0.0919, 0.0787, 0.0587, 0.0613, 0.0618, 0.0726], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:25:28,521 INFO [train.py:904] (0/8) Epoch 15, batch 3450, loss[loss=0.1721, simple_loss=0.2488, pruned_loss=0.04769, over 16317.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2621, pruned_loss=0.04586, over 3311444.27 frames. ], batch size: 165, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:26:27,257 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145594.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:26:38,108 INFO [train.py:904] (0/8) Epoch 15, batch 3500, loss[loss=0.1753, simple_loss=0.2589, pruned_loss=0.04586, over 16750.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2599, pruned_loss=0.0451, over 3317039.13 frames. ], batch size: 124, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:26:39,235 INFO [optim.py:368] (0/8) 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 03:27:36,027 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:27:47,271 INFO [train.py:904] (0/8) Epoch 15, batch 3550, loss[loss=0.1846, simple_loss=0.2854, pruned_loss=0.04192, over 17060.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2581, pruned_loss=0.04414, over 3322367.70 frames. ], batch size: 53, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:27:47,689 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145652.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:27:52,801 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145655.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:28:13,823 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9771, 4.1787, 2.4424, 4.7885, 3.1599, 4.7259, 2.6674, 3.4050], device='cuda:0'), covar=tensor([0.0275, 0.0346, 0.1527, 0.0195, 0.0774, 0.0460, 0.1383, 0.0617], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0152, 0.0170, 0.0216, 0.0200, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:28:15,000 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2355, 3.4525, 2.8657, 5.2724, 4.6137, 4.5698, 1.9645, 3.1481], device='cuda:0'), covar=tensor([0.1027, 0.0554, 0.1019, 0.0172, 0.0166, 0.0392, 0.1240, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0165, 0.0185, 0.0171, 0.0201, 0.0213, 0.0188, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:28:31,762 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145683.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:35,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8264, 3.7724, 4.1910, 2.0263, 4.3132, 4.4267, 3.2579, 3.4025], device='cuda:0'), covar=tensor([0.0679, 0.0232, 0.0172, 0.1164, 0.0081, 0.0151, 0.0388, 0.0368], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0106, 0.0091, 0.0139, 0.0073, 0.0118, 0.0123, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 03:28:54,824 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145700.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:28:57,596 INFO [train.py:904] (0/8) Epoch 15, batch 3600, loss[loss=0.188, simple_loss=0.2656, pruned_loss=0.05523, over 16655.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2571, pruned_loss=0.04426, over 3298006.20 frames. ], batch size: 89, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:28:58,729 INFO [optim.py:368] (0/8) 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 03:29:00,369 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:27,022 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=145722.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:34,323 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145727.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:29:39,120 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145731.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:30:10,706 INFO [train.py:904] (0/8) Epoch 15, batch 3650, loss[loss=0.1587, simple_loss=0.2258, pruned_loss=0.04578, over 16720.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2561, pruned_loss=0.04475, over 3303767.68 frames. ], batch size: 83, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:30:38,218 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=145770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:31:24,506 INFO [train.py:904] (0/8) Epoch 15, batch 3700, loss[loss=0.2004, simple_loss=0.2676, pruned_loss=0.0666, over 16883.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2558, pruned_loss=0.04664, over 3292541.78 frames. ], batch size: 109, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:31:26,281 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.174e+02 2.701e+02 3.170e+02 5.249e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 03:31:53,351 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=145821.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:32:11,600 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3679, 3.2647, 3.5105, 1.8251, 3.6191, 3.6053, 2.9728, 2.6481], device='cuda:0'), covar=tensor([0.0729, 0.0205, 0.0160, 0.1065, 0.0092, 0.0167, 0.0374, 0.0422], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0106, 0.0091, 0.0138, 0.0073, 0.0118, 0.0123, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 03:32:38,915 INFO [train.py:904] (0/8) Epoch 15, batch 3750, loss[loss=0.1767, simple_loss=0.277, pruned_loss=0.03822, over 17009.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2573, pruned_loss=0.04845, over 3281949.96 frames. ], batch size: 41, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:23,427 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=145882.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 03:33:48,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0720, 2.0650, 2.5422, 3.0419, 2.9676, 3.3637, 2.1065, 3.3181], device='cuda:0'), covar=tensor([0.0154, 0.0413, 0.0262, 0.0202, 0.0210, 0.0136, 0.0432, 0.0083], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0184, 0.0169, 0.0174, 0.0182, 0.0140, 0.0185, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:33:51,809 INFO [train.py:904] (0/8) Epoch 15, batch 3800, loss[loss=0.1988, simple_loss=0.2641, pruned_loss=0.06674, over 16917.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2587, pruned_loss=0.05007, over 3283699.73 frames. ], batch size: 109, lr: 4.56e-03, grad_scale: 8.0 2023-04-30 03:33:53,682 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.566e+02 2.221e+02 2.643e+02 3.115e+02 5.182e+02, threshold=5.286e+02, percent-clipped=0.0 2023-04-30 03:34:30,661 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7360, 4.5956, 4.7398, 4.9068, 5.0622, 4.5659, 4.9597, 5.0602], device='cuda:0'), covar=tensor([0.1590, 0.1055, 0.1495, 0.0709, 0.0563, 0.0904, 0.0919, 0.0598], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0748, 0.0896, 0.0769, 0.0574, 0.0599, 0.0606, 0.0710], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:34:31,953 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8432, 1.8842, 2.3555, 2.7031, 2.7264, 2.7350, 1.9686, 2.9305], device='cuda:0'), covar=tensor([0.0135, 0.0399, 0.0277, 0.0220, 0.0223, 0.0205, 0.0413, 0.0112], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0184, 0.0169, 0.0175, 0.0183, 0.0140, 0.0185, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:35:00,909 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145950.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 03:35:04,375 INFO [train.py:904] (0/8) Epoch 15, batch 3850, loss[loss=0.2113, simple_loss=0.2883, pruned_loss=0.0672, over 12508.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2587, pruned_loss=0.05047, over 3285422.45 frames. ], batch size: 246, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:13,426 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=145999.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:36:13,619 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6760, 3.7600, 4.2558, 2.0539, 4.5095, 4.6462, 3.1457, 3.2141], device='cuda:0'), covar=tensor([0.0790, 0.0249, 0.0184, 0.1130, 0.0047, 0.0073, 0.0410, 0.0420], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0106, 0.0091, 0.0138, 0.0073, 0.0117, 0.0123, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 03:36:14,678 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-146000.pt 2023-04-30 03:36:21,006 INFO [train.py:904] (0/8) Epoch 15, batch 3900, loss[loss=0.1631, simple_loss=0.2452, pruned_loss=0.04055, over 16499.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2581, pruned_loss=0.05065, over 3280903.26 frames. ], batch size: 68, lr: 4.56e-03, grad_scale: 16.0 2023-04-30 03:36:22,209 INFO [optim.py:368] (0/8) 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 03:36:57,929 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146027.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:37:32,699 INFO [train.py:904] (0/8) Epoch 15, batch 3950, loss[loss=0.1734, simple_loss=0.2478, pruned_loss=0.04951, over 16425.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2581, pruned_loss=0.05169, over 3288913.35 frames. ], batch size: 146, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:06,907 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:38:46,135 INFO [train.py:904] (0/8) Epoch 15, batch 4000, loss[loss=0.1652, simple_loss=0.2475, pruned_loss=0.04143, over 16980.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2574, pruned_loss=0.0516, over 3297121.28 frames. ], batch size: 41, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:38:47,410 INFO [optim.py:368] (0/8) 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 03:39:27,220 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3844, 3.3516, 3.4614, 3.5152, 3.5945, 3.3000, 3.5247, 3.6394], device='cuda:0'), covar=tensor([0.1352, 0.0916, 0.1076, 0.0604, 0.0570, 0.2297, 0.1048, 0.0683], device='cuda:0'), in_proj_covar=tensor([0.0607, 0.0747, 0.0894, 0.0767, 0.0574, 0.0600, 0.0606, 0.0708], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:39:56,266 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:39:59,985 INFO [train.py:904] (0/8) Epoch 15, batch 4050, loss[loss=0.1974, simple_loss=0.2744, pruned_loss=0.06025, over 16791.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2582, pruned_loss=0.05092, over 3289793.81 frames. ], batch size: 124, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:40:36,977 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146177.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 03:40:50,137 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0722, 2.0648, 2.1390, 3.6213, 2.0476, 2.4048, 2.1915, 2.2328], device='cuda:0'), covar=tensor([0.1237, 0.3606, 0.2592, 0.0537, 0.3889, 0.2420, 0.3322, 0.3135], device='cuda:0'), in_proj_covar=tensor([0.0385, 0.0425, 0.0352, 0.0328, 0.0427, 0.0490, 0.0388, 0.0498], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:41:13,948 INFO [train.py:904] (0/8) Epoch 15, batch 4100, loss[loss=0.1734, simple_loss=0.2667, pruned_loss=0.04, over 16733.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2596, pruned_loss=0.05027, over 3269217.83 frames. ], batch size: 83, lr: 4.55e-03, grad_scale: 16.0 2023-04-30 03:41:15,749 INFO [optim.py:368] (0/8) 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 03:41:26,635 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146210.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:30,197 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146250.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:42:32,946 INFO [train.py:904] (0/8) Epoch 15, batch 4150, loss[loss=0.2512, simple_loss=0.3229, pruned_loss=0.08976, over 11410.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2676, pruned_loss=0.05355, over 3210747.27 frames. ], batch size: 247, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:42:49,034 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8416, 4.0298, 4.2174, 4.1992, 4.2253, 4.0006, 3.8167, 3.9064], device='cuda:0'), covar=tensor([0.0412, 0.0551, 0.0466, 0.0506, 0.0495, 0.0523, 0.1199, 0.0579], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0397, 0.0390, 0.0370, 0.0438, 0.0411, 0.0505, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 03:43:45,065 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:46,721 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146299.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:43:50,411 INFO [train.py:904] (0/8) Epoch 15, batch 4200, loss[loss=0.2489, simple_loss=0.3196, pruned_loss=0.08908, over 11377.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2753, pruned_loss=0.0553, over 3192317.42 frames. ], batch size: 247, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:43:53,468 INFO [optim.py:368] (0/8) 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 03:44:58,659 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 15, batch 4250, loss[loss=0.165, simple_loss=0.2614, pruned_loss=0.03426, over 16744.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2786, pruned_loss=0.05547, over 3161116.74 frames. ], batch size: 124, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:19,375 INFO [train.py:904] (0/8) Epoch 15, batch 4300, loss[loss=0.203, simple_loss=0.3059, pruned_loss=0.04998, over 16879.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2801, pruned_loss=0.05472, over 3159722.87 frames. ], batch size: 96, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:46:23,348 INFO [optim.py:368] (0/8) 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 03:46:34,238 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-04-30 03:47:31,148 INFO [train.py:904] (0/8) Epoch 15, batch 4350, loss[loss=0.2011, simple_loss=0.2853, pruned_loss=0.05844, over 17222.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2831, pruned_loss=0.05532, over 3166220.85 frames. ], batch size: 44, lr: 4.55e-03, grad_scale: 4.0 2023-04-30 03:48:08,890 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146477.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:48:45,766 INFO [train.py:904] (0/8) Epoch 15, batch 4400, loss[loss=0.2439, simple_loss=0.312, pruned_loss=0.08788, over 11756.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.285, pruned_loss=0.05641, over 3178900.40 frames. ], batch size: 248, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:48:50,402 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.716e+02 2.574e+02 2.972e+02 3.574e+02 6.742e+02, threshold=5.944e+02, percent-clipped=2.0 2023-04-30 03:48:50,748 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:49:21,076 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146525.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:49:58,647 INFO [train.py:904] (0/8) Epoch 15, batch 4450, loss[loss=0.2044, simple_loss=0.2859, pruned_loss=0.06147, over 16621.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2881, pruned_loss=0.05756, over 3172859.87 frames. ], batch size: 62, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:50:10,118 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 03:50:42,140 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2186, 5.4978, 5.2666, 5.3458, 5.0652, 4.8759, 4.9764, 5.6493], device='cuda:0'), covar=tensor([0.1039, 0.0796, 0.0934, 0.0733, 0.0657, 0.0676, 0.0941, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0597, 0.0744, 0.0614, 0.0537, 0.0468, 0.0478, 0.0618, 0.0564], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:50:48,974 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7284, 4.9346, 5.1479, 4.8708, 4.9468, 5.5678, 5.0464, 4.7077], device='cuda:0'), covar=tensor([0.0963, 0.1549, 0.1587, 0.1745, 0.2343, 0.0795, 0.1198, 0.2151], device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0534, 0.0580, 0.0453, 0.0604, 0.0607, 0.0459, 0.0605], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 03:50:57,091 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146591.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:51:12,450 INFO [train.py:904] (0/8) Epoch 15, batch 4500, loss[loss=0.1861, simple_loss=0.2782, pruned_loss=0.04705, over 16795.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2882, pruned_loss=0.05777, over 3195375.02 frames. ], batch size: 102, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:51:15,379 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1225, 4.1656, 2.5611, 5.0589, 3.4455, 4.9270, 3.0117, 3.3720], device='cuda:0'), covar=tensor([0.0206, 0.0275, 0.1581, 0.0092, 0.0617, 0.0271, 0.1178, 0.0647], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0191, 0.0145, 0.0167, 0.0211, 0.0197, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:51:16,068 INFO [optim.py:368] (0/8) 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 03:51:47,966 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146626.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:52:25,476 INFO [train.py:904] (0/8) Epoch 15, batch 4550, loss[loss=0.2071, simple_loss=0.285, pruned_loss=0.06455, over 11624.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2886, pruned_loss=0.05872, over 3192061.56 frames. ], batch size: 248, lr: 4.55e-03, grad_scale: 8.0 2023-04-30 03:52:25,975 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146652.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:53:16,043 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146687.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 03:53:37,391 INFO [train.py:904] (0/8) Epoch 15, batch 4600, loss[loss=0.2011, simple_loss=0.2851, pruned_loss=0.05858, over 16776.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2901, pruned_loss=0.05932, over 3194158.98 frames. ], batch size: 124, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:53:41,729 INFO [optim.py:368] (0/8) 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 03:54:12,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7816, 5.0333, 5.2676, 4.9327, 5.0511, 5.7452, 5.1574, 4.8174], device='cuda:0'), covar=tensor([0.0940, 0.1826, 0.1902, 0.2008, 0.2560, 0.0835, 0.1370, 0.2383], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0531, 0.0578, 0.0453, 0.0603, 0.0606, 0.0459, 0.0603], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 03:54:43,976 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.35 vs. limit=5.0 2023-04-30 03:54:49,173 INFO [train.py:904] (0/8) Epoch 15, batch 4650, loss[loss=0.1939, simple_loss=0.2767, pruned_loss=0.05554, over 16770.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.289, pruned_loss=0.05893, over 3209245.23 frames. ], batch size: 124, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:55:25,040 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2670, 3.2597, 3.5857, 1.7219, 3.7216, 3.7757, 2.8776, 2.7998], device='cuda:0'), covar=tensor([0.0816, 0.0230, 0.0161, 0.1227, 0.0071, 0.0106, 0.0425, 0.0461], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0104, 0.0091, 0.0137, 0.0072, 0.0115, 0.0122, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 03:55:48,480 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-04-30 03:56:03,160 INFO [train.py:904] (0/8) Epoch 15, batch 4700, loss[loss=0.2007, simple_loss=0.2786, pruned_loss=0.06137, over 11944.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2859, pruned_loss=0.05737, over 3198501.68 frames. ], batch size: 246, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:56:07,891 INFO [optim.py:368] (0/8) 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 03:56:08,264 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=146805.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:56:34,438 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 03:56:51,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6868, 2.6475, 1.7966, 2.7356, 2.1287, 2.8113, 2.0417, 2.3233], device='cuda:0'), covar=tensor([0.0327, 0.0394, 0.1409, 0.0195, 0.0778, 0.0455, 0.1264, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0191, 0.0145, 0.0169, 0.0211, 0.0198, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 03:57:07,645 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=146844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:18,587 INFO [train.py:904] (0/8) Epoch 15, batch 4750, loss[loss=0.1709, simple_loss=0.2619, pruned_loss=0.03996, over 16665.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2817, pruned_loss=0.05527, over 3193903.51 frames. ], batch size: 57, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:57:20,528 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=146853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:57:51,688 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6637, 5.9886, 5.6721, 5.7887, 5.4591, 5.2533, 5.3320, 6.0716], device='cuda:0'), covar=tensor([0.1111, 0.0789, 0.1024, 0.0713, 0.0766, 0.0617, 0.1077, 0.0867], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0745, 0.0618, 0.0539, 0.0471, 0.0483, 0.0622, 0.0568], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 03:58:31,083 INFO [train.py:904] (0/8) Epoch 15, batch 4800, loss[loss=0.2181, simple_loss=0.2985, pruned_loss=0.06887, over 12078.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2783, pruned_loss=0.05334, over 3180520.83 frames. ], batch size: 247, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 03:58:36,182 INFO [optim.py:368] (0/8) 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 03:58:36,822 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=146905.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:59:40,653 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 03:59:47,628 INFO [train.py:904] (0/8) Epoch 15, batch 4850, loss[loss=0.198, simple_loss=0.2839, pruned_loss=0.05603, over 16538.00 frames. ], tot_loss[loss=0.1931, simple_loss=0.2798, pruned_loss=0.05319, over 3178881.25 frames. ], batch size: 68, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:00:27,686 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8994, 2.7608, 2.7230, 1.9545, 2.4649, 2.7146, 2.6201, 1.8186], device='cuda:0'), covar=tensor([0.0358, 0.0060, 0.0054, 0.0295, 0.0109, 0.0101, 0.0099, 0.0377], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0073, 0.0073, 0.0130, 0.0088, 0.0098, 0.0086, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 04:00:34,146 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=146982.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 04:01:03,665 INFO [train.py:904] (0/8) Epoch 15, batch 4900, loss[loss=0.1915, simple_loss=0.2768, pruned_loss=0.05307, over 17039.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2793, pruned_loss=0.05163, over 3180780.07 frames. ], batch size: 53, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:01:08,010 INFO [optim.py:368] (0/8) 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 04:01:53,025 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2719, 3.4160, 3.7774, 1.7896, 3.8985, 3.9231, 2.9977, 2.8522], device='cuda:0'), covar=tensor([0.0878, 0.0223, 0.0123, 0.1219, 0.0053, 0.0107, 0.0339, 0.0505], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0103, 0.0090, 0.0136, 0.0072, 0.0114, 0.0121, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 04:02:16,317 INFO [train.py:904] (0/8) Epoch 15, batch 4950, loss[loss=0.2035, simple_loss=0.29, pruned_loss=0.05856, over 16311.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.279, pruned_loss=0.05126, over 3185609.94 frames. ], batch size: 35, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:02:16,705 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8843, 5.2504, 5.3993, 5.1417, 5.1705, 5.7740, 5.2850, 4.9932], device='cuda:0'), covar=tensor([0.0856, 0.1673, 0.1832, 0.1847, 0.2457, 0.0996, 0.1434, 0.2305], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0525, 0.0571, 0.0449, 0.0594, 0.0602, 0.0455, 0.0598], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 04:02:55,710 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1899, 5.9235, 6.1365, 5.6812, 5.8340, 6.3839, 5.9486, 5.6314], device='cuda:0'), covar=tensor([0.0802, 0.1621, 0.1491, 0.1868, 0.2187, 0.0874, 0.1274, 0.2021], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0522, 0.0568, 0.0447, 0.0592, 0.0599, 0.0453, 0.0594], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 04:03:28,738 INFO [train.py:904] (0/8) Epoch 15, batch 5000, loss[loss=0.1732, simple_loss=0.2618, pruned_loss=0.04227, over 16221.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2799, pruned_loss=0.05096, over 3212172.58 frames. ], batch size: 35, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:03:32,273 INFO [optim.py:368] (0/8) 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 04:04:12,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3941, 2.2879, 2.3301, 4.0580, 2.1319, 2.6521, 2.3302, 2.4323], device='cuda:0'), covar=tensor([0.1078, 0.3285, 0.2453, 0.0443, 0.3628, 0.2252, 0.3131, 0.2940], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0418, 0.0345, 0.0320, 0.0421, 0.0479, 0.0382, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:04:23,350 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-04-30 04:04:39,308 INFO [train.py:904] (0/8) Epoch 15, batch 5050, loss[loss=0.1663, simple_loss=0.2548, pruned_loss=0.03886, over 17182.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2805, pruned_loss=0.05056, over 3214907.11 frames. ], batch size: 46, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:46,859 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:05:48,961 INFO [train.py:904] (0/8) Epoch 15, batch 5100, loss[loss=0.2233, simple_loss=0.2997, pruned_loss=0.07351, over 12253.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2787, pruned_loss=0.05006, over 3226290.16 frames. ], batch size: 246, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:05:52,968 INFO [optim.py:368] (0/8) 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 04:06:45,611 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:52,668 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147247.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:06:59,927 INFO [train.py:904] (0/8) Epoch 15, batch 5150, loss[loss=0.1949, simple_loss=0.2974, pruned_loss=0.04617, over 15557.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2787, pruned_loss=0.04951, over 3210451.91 frames. ], batch size: 190, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:07:45,012 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:03,059 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:14,142 INFO [train.py:904] (0/8) Epoch 15, batch 5200, loss[loss=0.1787, simple_loss=0.2744, pruned_loss=0.04152, over 16242.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2779, pruned_loss=0.04918, over 3210862.11 frames. ], batch size: 165, lr: 4.54e-03, grad_scale: 8.0 2023-04-30 04:08:14,610 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 04:08:25,246 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147310.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:08:55,067 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147330.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:09:27,033 INFO [train.py:904] (0/8) Epoch 15, batch 5250, loss[loss=0.1742, simple_loss=0.2505, pruned_loss=0.04893, over 16974.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2753, pruned_loss=0.04912, over 3212389.71 frames. ], batch size: 55, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:09:54,292 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147371.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:10:02,405 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1310, 1.4559, 1.8696, 2.0650, 2.2184, 2.3941, 1.6402, 2.2813], device='cuda:0'), covar=tensor([0.0182, 0.0464, 0.0244, 0.0323, 0.0262, 0.0185, 0.0475, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0182, 0.0168, 0.0173, 0.0181, 0.0138, 0.0185, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:10:37,511 INFO [train.py:904] (0/8) Epoch 15, batch 5300, loss[loss=0.1477, simple_loss=0.2366, pruned_loss=0.02942, over 16494.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2716, pruned_loss=0.04749, over 3216184.98 frames. ], batch size: 68, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:10:40,967 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.533e+02 1.977e+02 2.284e+02 2.754e+02 4.909e+02, threshold=4.569e+02, percent-clipped=0.0 2023-04-30 04:10:44,664 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147407.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:14,163 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8783, 5.2091, 5.4064, 5.1144, 5.1957, 5.7538, 5.2310, 4.8724], device='cuda:0'), covar=tensor([0.0899, 0.1675, 0.1663, 0.1691, 0.2174, 0.0872, 0.1333, 0.2154], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0526, 0.0569, 0.0448, 0.0598, 0.0606, 0.0454, 0.0597], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 04:11:23,784 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147434.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:11:49,980 INFO [train.py:904] (0/8) Epoch 15, batch 5350, loss[loss=0.184, simple_loss=0.2765, pruned_loss=0.04582, over 16398.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2701, pruned_loss=0.04672, over 3223816.75 frames. ], batch size: 146, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:12:14,768 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:12:53,051 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:00,704 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147500.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:13:03,291 INFO [train.py:904] (0/8) Epoch 15, batch 5400, loss[loss=0.1892, simple_loss=0.2765, pruned_loss=0.05097, over 17050.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2728, pruned_loss=0.0475, over 3229627.00 frames. ], batch size: 53, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:13:04,156 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-30 04:13:07,675 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.104e+02 2.571e+02 3.291e+02 5.861e+02, threshold=5.143e+02, percent-clipped=4.0 2023-04-30 04:13:34,837 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4021, 4.3985, 4.2771, 3.5427, 4.3211, 1.5062, 4.0735, 3.9721], device='cuda:0'), covar=tensor([0.0090, 0.0074, 0.0129, 0.0332, 0.0086, 0.2801, 0.0118, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0132, 0.0180, 0.0168, 0.0152, 0.0191, 0.0169, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:14:13,937 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147548.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:14:20,114 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9850, 2.3386, 2.3116, 2.9058, 2.1409, 3.2864, 1.7295, 2.7132], device='cuda:0'), covar=tensor([0.1038, 0.0573, 0.1036, 0.0154, 0.0126, 0.0350, 0.1375, 0.0661], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0166, 0.0188, 0.0168, 0.0201, 0.0212, 0.0191, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:14:20,734 INFO [train.py:904] (0/8) Epoch 15, batch 5450, loss[loss=0.2044, simple_loss=0.2852, pruned_loss=0.06183, over 12226.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2752, pruned_loss=0.04869, over 3214413.17 frames. ], batch size: 247, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:14:32,832 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1123, 5.1783, 4.9454, 4.3174, 5.0715, 1.7741, 4.7795, 4.8534], device='cuda:0'), covar=tensor([0.0090, 0.0074, 0.0150, 0.0407, 0.0078, 0.2549, 0.0119, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0132, 0.0180, 0.0169, 0.0152, 0.0191, 0.0169, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:15:32,468 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147597.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:15:40,038 INFO [train.py:904] (0/8) Epoch 15, batch 5500, loss[loss=0.2333, simple_loss=0.322, pruned_loss=0.07236, over 16478.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2825, pruned_loss=0.05332, over 3180985.88 frames. ], batch size: 75, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:15:45,686 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.456e+02 3.016e+02 4.198e+02 6.055e+02, threshold=6.032e+02, percent-clipped=7.0 2023-04-30 04:16:01,276 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4497, 3.5236, 3.3320, 2.9917, 3.1354, 3.4300, 3.2603, 3.2630], device='cuda:0'), covar=tensor([0.0595, 0.0515, 0.0307, 0.0292, 0.0641, 0.0454, 0.1097, 0.0542], device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0365, 0.0313, 0.0296, 0.0327, 0.0344, 0.0210, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:16:58,369 INFO [train.py:904] (0/8) Epoch 15, batch 5550, loss[loss=0.2299, simple_loss=0.3117, pruned_loss=0.07405, over 15381.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2896, pruned_loss=0.05832, over 3154646.86 frames. ], batch size: 190, lr: 4.53e-03, grad_scale: 4.0 2023-04-30 04:17:10,681 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:22,304 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:17:45,572 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2658, 2.3067, 1.6548, 1.8849, 2.7676, 2.3421, 2.9369, 3.0824], device='cuda:0'), covar=tensor([0.0122, 0.0428, 0.0616, 0.0541, 0.0244, 0.0425, 0.0221, 0.0233], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0217, 0.0209, 0.0210, 0.0215, 0.0216, 0.0220, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:18:21,670 INFO [train.py:904] (0/8) Epoch 15, batch 5600, loss[loss=0.2063, simple_loss=0.3002, pruned_loss=0.05624, over 16691.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2948, pruned_loss=0.06287, over 3117757.42 frames. ], batch size: 89, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:18:28,274 INFO [optim.py:368] (0/8) 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 04:18:47,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6340, 2.6090, 1.9194, 2.6982, 2.1937, 2.7449, 2.0733, 2.4062], device='cuda:0'), covar=tensor([0.0309, 0.0388, 0.1109, 0.0273, 0.0659, 0.0521, 0.1141, 0.0537], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0189, 0.0143, 0.0168, 0.0210, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:18:52,992 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:19:46,193 INFO [train.py:904] (0/8) Epoch 15, batch 5650, loss[loss=0.3245, simple_loss=0.3715, pruned_loss=0.1387, over 11166.00 frames. ], tot_loss[loss=0.2182, simple_loss=0.3007, pruned_loss=0.0678, over 3084146.90 frames. ], batch size: 250, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:20:04,691 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147763.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:20:47,705 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=147790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:21:05,477 INFO [train.py:904] (0/8) Epoch 15, batch 5700, loss[loss=0.2158, simple_loss=0.305, pruned_loss=0.06334, over 16725.00 frames. ], tot_loss[loss=0.2211, simple_loss=0.3026, pruned_loss=0.06983, over 3070729.47 frames. ], batch size: 134, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:21:11,573 INFO [optim.py:368] (0/8) 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 04:21:18,540 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9111, 4.1559, 3.9498, 4.0007, 3.7062, 3.8073, 3.8559, 4.1219], device='cuda:0'), covar=tensor([0.0981, 0.0850, 0.0997, 0.0743, 0.0729, 0.1330, 0.0811, 0.1015], device='cuda:0'), in_proj_covar=tensor([0.0587, 0.0730, 0.0606, 0.0524, 0.0458, 0.0470, 0.0602, 0.0553], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:21:26,384 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:22:23,103 INFO [train.py:904] (0/8) Epoch 15, batch 5750, loss[loss=0.204, simple_loss=0.293, pruned_loss=0.05752, over 16887.00 frames. ], tot_loss[loss=0.2246, simple_loss=0.3058, pruned_loss=0.07172, over 3042528.17 frames. ], batch size: 109, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:22:32,719 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:03,165 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:18,247 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1730, 3.2165, 1.9107, 3.4327, 2.3969, 3.4809, 2.0753, 2.5771], device='cuda:0'), covar=tensor([0.0305, 0.0425, 0.1721, 0.0286, 0.0898, 0.0621, 0.1715, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0189, 0.0143, 0.0168, 0.0210, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:23:36,766 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147897.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:23:44,677 INFO [train.py:904] (0/8) Epoch 15, batch 5800, loss[loss=0.2416, simple_loss=0.3117, pruned_loss=0.08571, over 11610.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.3054, pruned_loss=0.07081, over 3031271.25 frames. ], batch size: 247, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:23:51,421 INFO [optim.py:368] (0/8) 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 04:24:00,154 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=147911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:12,303 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147918.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:24:55,228 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=147945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:05,475 INFO [train.py:904] (0/8) Epoch 15, batch 5850, loss[loss=0.2023, simple_loss=0.2923, pruned_loss=0.05613, over 16735.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.3028, pruned_loss=0.06907, over 3034710.24 frames. ], batch size: 83, lr: 4.53e-03, grad_scale: 8.0 2023-04-30 04:25:16,288 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8224, 2.8882, 2.6208, 4.6467, 3.5329, 4.1995, 1.6174, 3.0468], device='cuda:0'), covar=tensor([0.1299, 0.0692, 0.1193, 0.0156, 0.0374, 0.0403, 0.1523, 0.0806], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0166, 0.0188, 0.0167, 0.0202, 0.0211, 0.0190, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:25:28,548 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=147966.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:25:38,389 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=147972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:26:23,345 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-148000.pt 2023-04-30 04:26:30,247 INFO [train.py:904] (0/8) Epoch 15, batch 5900, loss[loss=0.2146, simple_loss=0.2871, pruned_loss=0.07103, over 11670.00 frames. ], tot_loss[loss=0.2197, simple_loss=0.3026, pruned_loss=0.06834, over 3053557.20 frames. ], batch size: 247, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:26:39,380 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.034e+02 2.736e+02 3.208e+02 4.005e+02 8.372e+02, threshold=6.416e+02, percent-clipped=2.0 2023-04-30 04:26:52,641 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:26:54,039 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:27:35,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6858, 3.9297, 4.1539, 2.2835, 3.3092, 2.8642, 4.1356, 4.2212], device='cuda:0'), covar=tensor([0.0205, 0.0625, 0.0462, 0.1764, 0.0723, 0.0804, 0.0464, 0.0730], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0152, 0.0160, 0.0145, 0.0138, 0.0125, 0.0138, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:27:38,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7194, 4.9432, 5.1079, 4.9496, 4.9946, 5.5182, 4.9580, 4.7472], device='cuda:0'), covar=tensor([0.1015, 0.1759, 0.2270, 0.1857, 0.2393, 0.0917, 0.1570, 0.2380], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0531, 0.0579, 0.0452, 0.0602, 0.0608, 0.0460, 0.0605], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 04:27:50,063 INFO [train.py:904] (0/8) Epoch 15, batch 5950, loss[loss=0.2191, simple_loss=0.3047, pruned_loss=0.06675, over 17199.00 frames. ], tot_loss[loss=0.2178, simple_loss=0.3025, pruned_loss=0.0666, over 3066577.56 frames. ], batch size: 44, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:27:58,472 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 04:28:04,795 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-04-30 04:28:08,685 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:28:48,997 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:29:08,865 INFO [train.py:904] (0/8) Epoch 15, batch 6000, loss[loss=0.1824, simple_loss=0.2745, pruned_loss=0.04512, over 16880.00 frames. ], tot_loss[loss=0.2173, simple_loss=0.3017, pruned_loss=0.06641, over 3076841.17 frames. ], batch size: 96, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:29:08,866 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 04:29:19,438 INFO [train.py:938] (0/8) Epoch 15, validation: loss=0.1559, simple_loss=0.2691, pruned_loss=0.0214, over 944034.00 frames. 2023-04-30 04:29:19,439 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 04:29:26,131 INFO [optim.py:368] (0/8) 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 04:29:33,081 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:29:50,465 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9320, 2.7656, 2.8019, 2.0808, 2.6006, 2.1816, 2.7547, 2.9350], device='cuda:0'), covar=tensor([0.0259, 0.0663, 0.0529, 0.1676, 0.0758, 0.0856, 0.0556, 0.0650], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0154, 0.0162, 0.0147, 0.0139, 0.0127, 0.0140, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:30:15,358 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:30:36,864 INFO [train.py:904] (0/8) Epoch 15, batch 6050, loss[loss=0.1947, simple_loss=0.2983, pruned_loss=0.04557, over 16540.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.3002, pruned_loss=0.06576, over 3086939.71 frames. ], batch size: 68, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:31:07,975 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:31:20,263 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7312, 4.0721, 3.0718, 2.2731, 2.7950, 2.4880, 4.3347, 3.6375], device='cuda:0'), covar=tensor([0.2628, 0.0611, 0.1558, 0.2305, 0.2346, 0.1810, 0.0383, 0.1052], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0261, 0.0291, 0.0290, 0.0284, 0.0233, 0.0276, 0.0311], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:31:44,569 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2492, 3.3170, 1.8303, 3.5907, 2.4571, 3.5947, 1.9667, 2.5503], device='cuda:0'), covar=tensor([0.0263, 0.0365, 0.1702, 0.0243, 0.0817, 0.0572, 0.1597, 0.0801], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0190, 0.0143, 0.0169, 0.0210, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:31:59,327 INFO [train.py:904] (0/8) Epoch 15, batch 6100, loss[loss=0.1931, simple_loss=0.2795, pruned_loss=0.05334, over 16488.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2996, pruned_loss=0.06469, over 3103573.56 frames. ], batch size: 68, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:32:08,602 INFO [optim.py:368] (0/8) 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 04:32:17,954 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:33:18,246 INFO [train.py:904] (0/8) Epoch 15, batch 6150, loss[loss=0.1983, simple_loss=0.2845, pruned_loss=0.05604, over 16400.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2973, pruned_loss=0.0637, over 3114419.56 frames. ], batch size: 146, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:33:42,603 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:33:51,035 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0498, 3.0510, 3.3017, 1.6561, 3.4649, 3.4772, 2.7099, 2.6336], device='cuda:0'), covar=tensor([0.0880, 0.0239, 0.0206, 0.1259, 0.0078, 0.0152, 0.0441, 0.0485], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0104, 0.0090, 0.0137, 0.0073, 0.0115, 0.0122, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 04:34:09,397 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 04:34:14,836 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:34:37,598 INFO [train.py:904] (0/8) Epoch 15, batch 6200, loss[loss=0.2196, simple_loss=0.2988, pruned_loss=0.07024, over 16399.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2956, pruned_loss=0.06327, over 3123754.26 frames. ], batch size: 68, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:34:46,170 INFO [optim.py:368] (0/8) 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 04:34:58,638 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148315.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:13,779 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2873, 4.2428, 4.1217, 3.2835, 4.2001, 1.6907, 3.9368, 3.7703], device='cuda:0'), covar=tensor([0.0121, 0.0105, 0.0187, 0.0424, 0.0108, 0.2722, 0.0155, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0131, 0.0180, 0.0167, 0.0152, 0.0189, 0.0168, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:35:48,365 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148348.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:35:54,867 INFO [train.py:904] (0/8) Epoch 15, batch 6250, loss[loss=0.2314, simple_loss=0.307, pruned_loss=0.07792, over 11716.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2949, pruned_loss=0.06299, over 3115621.84 frames. ], batch size: 247, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:36:11,760 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148363.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:36:11,942 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6395, 3.9566, 4.4736, 1.9942, 4.6839, 4.6889, 3.3589, 3.2924], device='cuda:0'), covar=tensor([0.0831, 0.0197, 0.0121, 0.1200, 0.0039, 0.0098, 0.0312, 0.0453], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0104, 0.0090, 0.0138, 0.0072, 0.0115, 0.0122, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 04:37:11,940 INFO [train.py:904] (0/8) Epoch 15, batch 6300, loss[loss=0.2013, simple_loss=0.2908, pruned_loss=0.05585, over 16576.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2944, pruned_loss=0.06221, over 3120615.94 frames. ], batch size: 62, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:37:21,868 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.809e+02 3.245e+02 4.323e+02 1.479e+03, threshold=6.491e+02, percent-clipped=2.0 2023-04-30 04:37:26,390 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7565, 3.7072, 3.8439, 3.6060, 3.7887, 4.1933, 3.8621, 3.6298], device='cuda:0'), covar=tensor([0.2087, 0.2477, 0.2703, 0.3021, 0.3053, 0.2044, 0.1919, 0.2976], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0545, 0.0591, 0.0461, 0.0615, 0.0622, 0.0472, 0.0619], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 04:37:57,328 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148429.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:38:33,526 INFO [train.py:904] (0/8) Epoch 15, batch 6350, loss[loss=0.2007, simple_loss=0.2794, pruned_loss=0.06097, over 16686.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2964, pruned_loss=0.06457, over 3085021.46 frames. ], batch size: 57, lr: 4.52e-03, grad_scale: 4.0 2023-04-30 04:38:34,154 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2023-04-30 04:39:03,914 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:31,732 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1623, 3.7062, 3.6111, 2.4324, 3.3816, 3.6896, 3.5074, 1.9952], device='cuda:0'), covar=tensor([0.0515, 0.0040, 0.0052, 0.0386, 0.0080, 0.0100, 0.0073, 0.0443], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0075, 0.0075, 0.0133, 0.0089, 0.0099, 0.0087, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 04:39:33,582 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:45,805 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:39:51,478 INFO [train.py:904] (0/8) Epoch 15, batch 6400, loss[loss=0.213, simple_loss=0.299, pruned_loss=0.06353, over 16352.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2968, pruned_loss=0.0659, over 3063978.34 frames. ], batch size: 165, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:39:59,989 INFO [optim.py:368] (0/8) 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 04:40:08,913 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:40:18,149 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 15, batch 6450, loss[loss=0.204, simple_loss=0.2759, pruned_loss=0.06603, over 11815.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2961, pruned_loss=0.06411, over 3094353.75 frames. ], batch size: 247, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:41:19,857 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:22,168 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:41:31,958 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8610, 4.9180, 5.3013, 5.2893, 5.3003, 4.9184, 4.9198, 4.5925], device='cuda:0'), covar=tensor([0.0262, 0.0395, 0.0341, 0.0349, 0.0410, 0.0328, 0.0835, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0367, 0.0392, 0.0386, 0.0370, 0.0438, 0.0409, 0.0505, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 04:41:31,986 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:42:27,008 INFO [train.py:904] (0/8) Epoch 15, batch 6500, loss[loss=0.2083, simple_loss=0.2913, pruned_loss=0.0627, over 16849.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2941, pruned_loss=0.06342, over 3101423.34 frames. ], batch size: 116, lr: 4.52e-03, grad_scale: 8.0 2023-04-30 04:42:37,001 INFO [optim.py:368] (0/8) 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 04:42:44,403 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9834, 3.3464, 3.4408, 2.0264, 2.8201, 2.1712, 3.3077, 3.5289], device='cuda:0'), covar=tensor([0.0322, 0.0813, 0.0542, 0.1901, 0.0870, 0.1010, 0.0780, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0155, 0.0163, 0.0148, 0.0139, 0.0128, 0.0141, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:42:47,766 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:25,061 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2995, 3.2268, 3.1658, 3.4379, 3.3963, 3.1858, 3.4383, 3.4718], device='cuda:0'), covar=tensor([0.1232, 0.1123, 0.1399, 0.0771, 0.0957, 0.3119, 0.1202, 0.1059], device='cuda:0'), in_proj_covar=tensor([0.0568, 0.0701, 0.0842, 0.0712, 0.0539, 0.0565, 0.0570, 0.0666], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:43:33,201 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:46,598 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148651.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:43:47,986 INFO [train.py:904] (0/8) Epoch 15, batch 6550, loss[loss=0.202, simple_loss=0.3054, pruned_loss=0.04928, over 16371.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2968, pruned_loss=0.0639, over 3112727.58 frames. ], batch size: 146, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:44:03,206 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.70 vs. limit=2.0 2023-04-30 04:45:05,109 INFO [train.py:904] (0/8) Epoch 15, batch 6600, loss[loss=0.2304, simple_loss=0.3158, pruned_loss=0.07252, over 16715.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2986, pruned_loss=0.06403, over 3124142.01 frames. ], batch size: 134, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:45:13,946 INFO [optim.py:368] (0/8) 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 04:45:19,970 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:26,260 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=148716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:45:58,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7687, 3.8567, 2.2632, 4.5259, 2.8703, 4.4716, 2.6632, 3.1128], device='cuda:0'), covar=tensor([0.0225, 0.0321, 0.1504, 0.0163, 0.0771, 0.0371, 0.1213, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0188, 0.0142, 0.0168, 0.0209, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:46:22,128 INFO [train.py:904] (0/8) Epoch 15, batch 6650, loss[loss=0.1883, simple_loss=0.2757, pruned_loss=0.05045, over 16730.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.299, pruned_loss=0.06509, over 3098488.76 frames. ], batch size: 89, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:46:32,238 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9983, 3.0891, 1.7545, 3.2808, 2.2656, 3.3341, 2.1063, 2.5866], device='cuda:0'), covar=tensor([0.0281, 0.0365, 0.1587, 0.0168, 0.0838, 0.0467, 0.1376, 0.0657], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0188, 0.0142, 0.0168, 0.0209, 0.0195, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:46:32,297 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4403, 2.9186, 2.9312, 1.8445, 2.6008, 2.0468, 3.0306, 3.0997], device='cuda:0'), covar=tensor([0.0251, 0.0747, 0.0603, 0.1951, 0.0891, 0.0984, 0.0641, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0154, 0.0162, 0.0146, 0.0139, 0.0127, 0.0140, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:46:50,546 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9682, 4.7685, 4.9351, 5.1771, 5.3709, 4.7443, 5.3493, 5.3418], device='cuda:0'), covar=tensor([0.1684, 0.1153, 0.1571, 0.0706, 0.0513, 0.0849, 0.0466, 0.0539], device='cuda:0'), in_proj_covar=tensor([0.0569, 0.0703, 0.0844, 0.0714, 0.0537, 0.0565, 0.0572, 0.0666], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:47:00,946 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=148777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:13,456 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:47:39,124 INFO [train.py:904] (0/8) Epoch 15, batch 6700, loss[loss=0.187, simple_loss=0.2704, pruned_loss=0.05175, over 16488.00 frames. ], tot_loss[loss=0.2145, simple_loss=0.2978, pruned_loss=0.06559, over 3086754.18 frames. ], batch size: 68, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:47:47,146 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.925e+02 2.803e+02 3.506e+02 4.354e+02 9.571e+02, threshold=7.012e+02, percent-clipped=2.0 2023-04-30 04:47:57,644 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6636, 2.5906, 2.4354, 4.0322, 2.8964, 3.9455, 1.3798, 2.8103], device='cuda:0'), covar=tensor([0.1413, 0.0756, 0.1263, 0.0177, 0.0282, 0.0397, 0.1669, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0165, 0.0186, 0.0167, 0.0201, 0.0209, 0.0190, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:48:54,924 INFO [train.py:904] (0/8) Epoch 15, batch 6750, loss[loss=0.1934, simple_loss=0.2816, pruned_loss=0.05262, over 16356.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2974, pruned_loss=0.06593, over 3083423.16 frames. ], batch size: 146, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:48:58,426 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=148854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:48:58,767 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 04:49:35,153 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 04:50:07,999 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6607, 2.8547, 2.6582, 4.9569, 3.7741, 4.2650, 1.7123, 2.8254], device='cuda:0'), covar=tensor([0.1363, 0.0751, 0.1197, 0.0132, 0.0350, 0.0382, 0.1497, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0165, 0.0186, 0.0166, 0.0201, 0.0209, 0.0190, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 04:50:09,783 INFO [train.py:904] (0/8) Epoch 15, batch 6800, loss[loss=0.2076, simple_loss=0.2983, pruned_loss=0.05848, over 16721.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2966, pruned_loss=0.06542, over 3099631.92 frames. ], batch size: 134, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:50:21,236 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.768e+02 2.837e+02 3.542e+02 4.429e+02 7.153e+02, threshold=7.083e+02, percent-clipped=1.0 2023-04-30 04:50:43,582 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-30 04:51:01,878 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 04:51:15,831 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=148943.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:51:28,465 INFO [train.py:904] (0/8) Epoch 15, batch 6850, loss[loss=0.2045, simple_loss=0.2958, pruned_loss=0.05663, over 17079.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2973, pruned_loss=0.06547, over 3104155.06 frames. ], batch size: 47, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:51:32,571 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9721, 3.8477, 4.0403, 4.1983, 4.2744, 3.8046, 4.1888, 4.2740], device='cuda:0'), covar=tensor([0.1536, 0.1057, 0.1196, 0.0588, 0.0547, 0.1580, 0.0805, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0565, 0.0698, 0.0835, 0.0707, 0.0533, 0.0560, 0.0569, 0.0661], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:51:41,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 04:52:26,496 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=148991.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:43,271 INFO [train.py:904] (0/8) Epoch 15, batch 6900, loss[loss=0.1845, simple_loss=0.2816, pruned_loss=0.04372, over 16660.00 frames. ], tot_loss[loss=0.2138, simple_loss=0.299, pruned_loss=0.06428, over 3107682.49 frames. ], batch size: 62, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:52:51,562 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149007.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:52:54,701 INFO [optim.py:368] (0/8) 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 04:53:55,993 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 04:53:58,322 INFO [train.py:904] (0/8) Epoch 15, batch 6950, loss[loss=0.1943, simple_loss=0.2808, pruned_loss=0.05393, over 16277.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.3005, pruned_loss=0.06558, over 3105275.68 frames. ], batch size: 35, lr: 4.51e-03, grad_scale: 8.0 2023-04-30 04:54:27,431 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149072.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:54:42,975 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1971, 5.2042, 4.9947, 4.6353, 4.5702, 5.0676, 5.0784, 4.7530], device='cuda:0'), covar=tensor([0.0691, 0.0529, 0.0324, 0.0335, 0.1172, 0.0534, 0.0290, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0259, 0.0359, 0.0306, 0.0290, 0.0322, 0.0337, 0.0209, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 04:54:43,365 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-04-30 04:54:45,958 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:55:10,870 INFO [train.py:904] (0/8) Epoch 15, batch 7000, loss[loss=0.2326, simple_loss=0.3231, pruned_loss=0.07108, over 16635.00 frames. ], tot_loss[loss=0.2143, simple_loss=0.2999, pruned_loss=0.06434, over 3107898.21 frames. ], batch size: 134, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:55:23,365 INFO [optim.py:368] (0/8) 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 04:55:57,578 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:56:25,570 INFO [train.py:904] (0/8) Epoch 15, batch 7050, loss[loss=0.2095, simple_loss=0.2975, pruned_loss=0.06072, over 16220.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.3008, pruned_loss=0.06425, over 3118937.02 frames. ], batch size: 165, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:56:28,521 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:40,356 INFO [train.py:904] (0/8) Epoch 15, batch 7100, loss[loss=0.2127, simple_loss=0.305, pruned_loss=0.06022, over 16793.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2997, pruned_loss=0.0642, over 3116259.61 frames. ], batch size: 83, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:57:40,684 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 04:57:53,982 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 2.990e+02 3.583e+02 4.304e+02 1.223e+03, threshold=7.166e+02, percent-clipped=1.0 2023-04-30 04:58:55,138 INFO [train.py:904] (0/8) Epoch 15, batch 7150, loss[loss=0.2086, simple_loss=0.2966, pruned_loss=0.06032, over 16794.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2982, pruned_loss=0.0636, over 3132394.76 frames. ], batch size: 102, lr: 4.51e-03, grad_scale: 2.0 2023-04-30 04:59:37,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6161, 4.9793, 4.5205, 4.8826, 4.4954, 4.4286, 4.5405, 5.0680], device='cuda:0'), covar=tensor([0.2179, 0.1622, 0.2308, 0.1181, 0.1559, 0.1661, 0.1942, 0.1629], device='cuda:0'), in_proj_covar=tensor([0.0592, 0.0730, 0.0609, 0.0529, 0.0459, 0.0475, 0.0609, 0.0553], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:00:05,785 INFO [train.py:904] (0/8) Epoch 15, batch 7200, loss[loss=0.1819, simple_loss=0.2751, pruned_loss=0.04437, over 16222.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2954, pruned_loss=0.06176, over 3119488.76 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:00:13,247 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:00:17,919 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.349e+02 2.693e+02 3.090e+02 3.791e+02 7.265e+02, threshold=6.181e+02, percent-clipped=1.0 2023-04-30 05:01:26,135 INFO [train.py:904] (0/8) Epoch 15, batch 7250, loss[loss=0.1994, simple_loss=0.2791, pruned_loss=0.05988, over 16993.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2928, pruned_loss=0.06043, over 3131853.38 frames. ], batch size: 41, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:01:31,290 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149355.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:01:57,255 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=149372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:02:42,295 INFO [train.py:904] (0/8) Epoch 15, batch 7300, loss[loss=0.1943, simple_loss=0.292, pruned_loss=0.04827, over 16881.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2921, pruned_loss=0.06024, over 3131652.34 frames. ], batch size: 96, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:02:55,637 INFO [optim.py:368] (0/8) 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 05:03:09,833 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=149420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:03:58,890 INFO [train.py:904] (0/8) Epoch 15, batch 7350, loss[loss=0.2052, simple_loss=0.2955, pruned_loss=0.05741, over 16540.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2938, pruned_loss=0.0619, over 3097706.05 frames. ], batch size: 75, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:05:17,933 INFO [train.py:904] (0/8) Epoch 15, batch 7400, loss[loss=0.2478, simple_loss=0.3133, pruned_loss=0.09119, over 11417.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2953, pruned_loss=0.0631, over 3089560.39 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:05:32,178 INFO [optim.py:368] (0/8) 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 05:05:37,396 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:06:37,035 INFO [train.py:904] (0/8) Epoch 15, batch 7450, loss[loss=0.2457, simple_loss=0.3103, pruned_loss=0.0906, over 11741.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2965, pruned_loss=0.06406, over 3099114.56 frames. ], batch size: 247, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:07:00,595 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149565.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:07:16,508 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:08:00,173 INFO [train.py:904] (0/8) Epoch 15, batch 7500, loss[loss=0.1704, simple_loss=0.2567, pruned_loss=0.04206, over 16459.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2969, pruned_loss=0.06322, over 3101443.00 frames. ], batch size: 68, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:08:16,082 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.765e+02 2.987e+02 3.634e+02 4.459e+02 1.121e+03, threshold=7.267e+02, percent-clipped=5.0 2023-04-30 05:08:39,465 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149626.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:08:58,819 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3656, 3.8155, 3.7311, 2.4293, 3.5581, 3.8095, 3.5138, 1.6981], device='cuda:0'), covar=tensor([0.0474, 0.0055, 0.0056, 0.0415, 0.0095, 0.0143, 0.0092, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0074, 0.0073, 0.0132, 0.0087, 0.0097, 0.0084, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:09:18,811 INFO [train.py:904] (0/8) Epoch 15, batch 7550, loss[loss=0.2044, simple_loss=0.2835, pruned_loss=0.06271, over 16667.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2961, pruned_loss=0.06377, over 3080067.53 frames. ], batch size: 134, lr: 4.50e-03, grad_scale: 2.0 2023-04-30 05:10:35,811 INFO [train.py:904] (0/8) Epoch 15, batch 7600, loss[loss=0.2383, simple_loss=0.3049, pruned_loss=0.08582, over 11481.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2953, pruned_loss=0.0641, over 3079455.60 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:10:50,677 INFO [optim.py:368] (0/8) 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 05:11:07,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4741, 3.5362, 3.3108, 3.0140, 3.1254, 3.4066, 3.2726, 3.2179], device='cuda:0'), covar=tensor([0.0657, 0.0512, 0.0299, 0.0309, 0.0579, 0.0494, 0.1199, 0.0540], device='cuda:0'), in_proj_covar=tensor([0.0258, 0.0356, 0.0305, 0.0289, 0.0320, 0.0336, 0.0209, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:11:45,964 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=149746.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:11:55,075 INFO [train.py:904] (0/8) Epoch 15, batch 7650, loss[loss=0.2262, simple_loss=0.3027, pruned_loss=0.07483, over 15258.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2958, pruned_loss=0.06435, over 3099316.23 frames. ], batch size: 190, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:11:57,754 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-04-30 05:12:22,576 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4630, 3.4777, 2.6341, 2.1236, 2.3134, 2.2081, 3.5854, 3.1931], device='cuda:0'), covar=tensor([0.2753, 0.0688, 0.1747, 0.2559, 0.2436, 0.2033, 0.0436, 0.1109], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0264, 0.0295, 0.0296, 0.0287, 0.0237, 0.0279, 0.0313], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:12:45,839 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6628, 3.7389, 4.3087, 2.1343, 4.5664, 4.5138, 3.0740, 3.3227], device='cuda:0'), covar=tensor([0.0720, 0.0219, 0.0170, 0.1060, 0.0048, 0.0108, 0.0408, 0.0400], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0104, 0.0090, 0.0137, 0.0072, 0.0114, 0.0122, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 05:13:13,702 INFO [train.py:904] (0/8) Epoch 15, batch 7700, loss[loss=0.2088, simple_loss=0.2924, pruned_loss=0.06262, over 16226.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2959, pruned_loss=0.06523, over 3081405.56 frames. ], batch size: 165, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:13:22,111 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=149807.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:13:29,265 INFO [optim.py:368] (0/8) 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 05:14:18,392 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4493, 4.3290, 4.5840, 4.7199, 4.8994, 4.4268, 4.8617, 4.8958], device='cuda:0'), covar=tensor([0.1825, 0.1237, 0.1522, 0.0728, 0.0503, 0.0917, 0.0616, 0.0596], device='cuda:0'), in_proj_covar=tensor([0.0556, 0.0689, 0.0826, 0.0700, 0.0530, 0.0554, 0.0563, 0.0654], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:14:33,197 INFO [train.py:904] (0/8) Epoch 15, batch 7750, loss[loss=0.2081, simple_loss=0.3064, pruned_loss=0.05494, over 16902.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2965, pruned_loss=0.06544, over 3076182.51 frames. ], batch size: 96, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:15:02,346 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:15:37,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7824, 3.7672, 3.8864, 3.7315, 3.8281, 4.2174, 3.8902, 3.6308], device='cuda:0'), covar=tensor([0.1976, 0.2161, 0.2401, 0.2215, 0.2467, 0.1575, 0.1597, 0.2355], device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0541, 0.0589, 0.0456, 0.0605, 0.0617, 0.0469, 0.0608], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:15:51,923 INFO [train.py:904] (0/8) Epoch 15, batch 7800, loss[loss=0.2512, simple_loss=0.3107, pruned_loss=0.09579, over 11721.00 frames. ], tot_loss[loss=0.214, simple_loss=0.2966, pruned_loss=0.06568, over 3091224.86 frames. ], batch size: 247, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:16:07,343 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.059e+02 3.237e+02 3.890e+02 4.639e+02 7.802e+02, threshold=7.781e+02, percent-clipped=0.0 2023-04-30 05:16:20,885 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=149921.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:16:21,029 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2910, 3.8134, 4.1598, 1.7326, 4.3464, 4.4978, 3.2946, 3.0830], device='cuda:0'), covar=tensor([0.1131, 0.0185, 0.0177, 0.1370, 0.0071, 0.0126, 0.0340, 0.0574], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0103, 0.0089, 0.0135, 0.0072, 0.0113, 0.0121, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 05:17:06,641 INFO [train.py:904] (0/8) Epoch 15, batch 7850, loss[loss=0.2398, simple_loss=0.3118, pruned_loss=0.08389, over 11428.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.2976, pruned_loss=0.06557, over 3088469.55 frames. ], batch size: 248, lr: 4.50e-03, grad_scale: 4.0 2023-04-30 05:18:19,022 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-150000.pt 2023-04-30 05:18:25,140 INFO [train.py:904] (0/8) Epoch 15, batch 7900, loss[loss=0.189, simple_loss=0.2852, pruned_loss=0.04641, over 16797.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2966, pruned_loss=0.06531, over 3064509.58 frames. ], batch size: 102, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:18:40,306 INFO [optim.py:368] (0/8) 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 05:19:15,241 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3840, 3.5101, 1.5983, 3.8652, 2.4382, 3.8530, 1.7306, 2.5948], device='cuda:0'), covar=tensor([0.0284, 0.0376, 0.2221, 0.0219, 0.0920, 0.0480, 0.1991, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0192, 0.0144, 0.0171, 0.0211, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 05:19:26,625 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-04-30 05:19:27,608 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 05:19:43,698 INFO [train.py:904] (0/8) Epoch 15, batch 7950, loss[loss=0.2597, simple_loss=0.3178, pruned_loss=0.1008, over 11902.00 frames. ], tot_loss[loss=0.214, simple_loss=0.297, pruned_loss=0.06546, over 3064018.09 frames. ], batch size: 248, lr: 4.49e-03, grad_scale: 4.0 2023-04-30 05:20:25,911 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:20:58,688 INFO [train.py:904] (0/8) Epoch 15, batch 8000, loss[loss=0.2587, simple_loss=0.32, pruned_loss=0.09871, over 11278.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2975, pruned_loss=0.06581, over 3057766.45 frames. ], batch size: 248, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:20:59,808 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150102.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 05:21:14,673 INFO [optim.py:368] (0/8) 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 05:21:54,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2176, 2.1534, 2.1500, 3.9871, 2.0893, 2.5458, 2.2434, 2.3018], device='cuda:0'), covar=tensor([0.1209, 0.3557, 0.2805, 0.0461, 0.4107, 0.2459, 0.3363, 0.3370], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0416, 0.0345, 0.0320, 0.0426, 0.0478, 0.0382, 0.0486], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:21:55,934 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:22:14,259 INFO [train.py:904] (0/8) Epoch 15, batch 8050, loss[loss=0.2095, simple_loss=0.2971, pruned_loss=0.06097, over 16741.00 frames. ], tot_loss[loss=0.2127, simple_loss=0.2966, pruned_loss=0.06437, over 3088734.12 frames. ], batch size: 83, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:22:41,254 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:23:30,584 INFO [train.py:904] (0/8) Epoch 15, batch 8100, loss[loss=0.2046, simple_loss=0.2917, pruned_loss=0.05877, over 16184.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2957, pruned_loss=0.06348, over 3093224.88 frames. ], batch size: 165, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:23:45,514 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.847e+02 2.677e+02 3.415e+02 4.132e+02 1.188e+03, threshold=6.830e+02, percent-clipped=3.0 2023-04-30 05:23:55,521 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:24:00,003 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:24:01,287 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7294, 3.1598, 3.1335, 1.9206, 2.7517, 2.0948, 3.1979, 3.3573], device='cuda:0'), covar=tensor([0.0323, 0.0740, 0.0693, 0.2097, 0.0905, 0.1048, 0.0722, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0154, 0.0162, 0.0147, 0.0140, 0.0127, 0.0141, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 05:24:09,732 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8568, 4.0152, 3.1950, 2.4533, 2.8830, 2.5533, 4.4097, 3.6722], device='cuda:0'), covar=tensor([0.2579, 0.0705, 0.1516, 0.2248, 0.2406, 0.1785, 0.0383, 0.1075], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0263, 0.0295, 0.0296, 0.0289, 0.0238, 0.0279, 0.0313], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:24:45,715 INFO [train.py:904] (0/8) Epoch 15, batch 8150, loss[loss=0.2297, simple_loss=0.2942, pruned_loss=0.08262, over 11366.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2927, pruned_loss=0.06198, over 3103441.18 frames. ], batch size: 247, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:24:57,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9102, 4.0267, 3.1718, 2.4115, 2.7661, 2.5692, 4.3822, 3.5733], device='cuda:0'), covar=tensor([0.2574, 0.0771, 0.1644, 0.2428, 0.2520, 0.1837, 0.0474, 0.1213], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0263, 0.0296, 0.0296, 0.0289, 0.0238, 0.0280, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:25:11,110 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150269.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:18,980 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:25:43,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8091, 2.6221, 2.3018, 3.7135, 2.7366, 3.7963, 1.4622, 2.7491], device='cuda:0'), covar=tensor([0.1207, 0.0654, 0.1230, 0.0149, 0.0215, 0.0436, 0.1504, 0.0810], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0166, 0.0188, 0.0167, 0.0204, 0.0210, 0.0191, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 05:26:00,383 INFO [train.py:904] (0/8) Epoch 15, batch 8200, loss[loss=0.2217, simple_loss=0.3033, pruned_loss=0.07004, over 15413.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2907, pruned_loss=0.06191, over 3094092.22 frames. ], batch size: 191, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:26:16,346 INFO [optim.py:368] (0/8) 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 05:26:23,988 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-04-30 05:26:54,596 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:27:22,326 INFO [train.py:904] (0/8) Epoch 15, batch 8250, loss[loss=0.173, simple_loss=0.2626, pruned_loss=0.04174, over 12147.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2895, pruned_loss=0.05903, over 3098757.70 frames. ], batch size: 246, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:12,481 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 05:28:45,013 INFO [train.py:904] (0/8) Epoch 15, batch 8300, loss[loss=0.1755, simple_loss=0.2595, pruned_loss=0.0457, over 11893.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2868, pruned_loss=0.05632, over 3071851.14 frames. ], batch size: 246, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:28:45,691 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150402.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 05:29:00,244 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.77 vs. limit=5.0 2023-04-30 05:29:01,278 INFO [optim.py:368] (0/8) 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 05:29:13,657 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0619, 3.2199, 3.1373, 2.2487, 2.9489, 3.1747, 3.1151, 1.9772], device='cuda:0'), covar=tensor([0.0407, 0.0046, 0.0048, 0.0339, 0.0080, 0.0076, 0.0066, 0.0398], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0072, 0.0072, 0.0130, 0.0086, 0.0095, 0.0083, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:29:39,510 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:30:05,309 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150450.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:30:07,766 INFO [train.py:904] (0/8) Epoch 15, batch 8350, loss[loss=0.2201, simple_loss=0.3074, pruned_loss=0.06637, over 15241.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2871, pruned_loss=0.05499, over 3076961.73 frames. ], batch size: 190, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:17,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2953, 3.3887, 3.6252, 3.6266, 3.6165, 3.4214, 3.4521, 3.5144], device='cuda:0'), covar=tensor([0.0442, 0.0821, 0.0493, 0.0477, 0.0593, 0.0566, 0.0936, 0.0521], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0393, 0.0382, 0.0367, 0.0437, 0.0405, 0.0503, 0.0325], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 05:31:29,421 INFO [train.py:904] (0/8) Epoch 15, batch 8400, loss[loss=0.172, simple_loss=0.2695, pruned_loss=0.03723, over 16269.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2842, pruned_loss=0.05275, over 3079086.58 frames. ], batch size: 165, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:31:46,293 INFO [optim.py:368] (0/8) 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 05:32:27,430 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150537.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:32:27,755 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.92 vs. limit=5.0 2023-04-30 05:32:46,098 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9431, 2.7370, 2.5952, 2.0912, 2.4663, 2.7383, 2.6238, 1.9535], device='cuda:0'), covar=tensor([0.0325, 0.0063, 0.0056, 0.0275, 0.0093, 0.0090, 0.0086, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0073, 0.0072, 0.0130, 0.0086, 0.0095, 0.0084, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:32:49,402 INFO [train.py:904] (0/8) Epoch 15, batch 8450, loss[loss=0.1725, simple_loss=0.2737, pruned_loss=0.03563, over 16239.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.2825, pruned_loss=0.0512, over 3065851.37 frames. ], batch size: 165, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:33:04,468 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5604, 3.4056, 3.7320, 1.9267, 3.8954, 3.9423, 3.0325, 3.1679], device='cuda:0'), covar=tensor([0.0644, 0.0230, 0.0170, 0.1119, 0.0047, 0.0140, 0.0342, 0.0329], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0103, 0.0088, 0.0135, 0.0071, 0.0111, 0.0120, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 05:34:02,222 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:34:08,625 INFO [train.py:904] (0/8) Epoch 15, batch 8500, loss[loss=0.1743, simple_loss=0.2648, pruned_loss=0.04195, over 16377.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2792, pruned_loss=0.04892, over 3077448.34 frames. ], batch size: 146, lr: 4.49e-03, grad_scale: 8.0 2023-04-30 05:34:25,011 INFO [optim.py:368] (0/8) 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 05:34:47,294 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0059, 2.2847, 2.4112, 3.0225, 1.9949, 3.2823, 1.7725, 2.8288], device='cuda:0'), covar=tensor([0.1136, 0.0621, 0.0952, 0.0157, 0.0092, 0.0340, 0.1403, 0.0618], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0162, 0.0184, 0.0162, 0.0197, 0.0205, 0.0187, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-30 05:34:54,235 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:35:30,241 INFO [train.py:904] (0/8) Epoch 15, batch 8550, loss[loss=0.1989, simple_loss=0.2944, pruned_loss=0.05166, over 15251.00 frames. ], tot_loss[loss=0.1861, simple_loss=0.2766, pruned_loss=0.04785, over 3060292.45 frames. ], batch size: 191, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:36:29,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7123, 5.0666, 4.8851, 4.8792, 4.5535, 4.5766, 4.5022, 5.1349], device='cuda:0'), covar=tensor([0.1223, 0.0885, 0.0863, 0.0702, 0.0874, 0.0907, 0.1108, 0.0816], device='cuda:0'), in_proj_covar=tensor([0.0585, 0.0722, 0.0599, 0.0520, 0.0453, 0.0468, 0.0601, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:37:12,303 INFO [train.py:904] (0/8) Epoch 15, batch 8600, loss[loss=0.1966, simple_loss=0.2908, pruned_loss=0.05116, over 16655.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2766, pruned_loss=0.04708, over 3035766.91 frames. ], batch size: 134, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:37:32,420 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.969e+02 2.521e+02 3.034e+02 3.543e+02 4.971e+02, threshold=6.069e+02, percent-clipped=0.0 2023-04-30 05:38:14,922 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:19,745 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:38:51,968 INFO [train.py:904] (0/8) Epoch 15, batch 8650, loss[loss=0.1809, simple_loss=0.2827, pruned_loss=0.03956, over 16242.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2752, pruned_loss=0.04631, over 3025179.31 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:39:27,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7005, 3.6815, 4.1091, 1.9434, 4.2530, 4.3356, 3.2062, 3.4232], device='cuda:0'), covar=tensor([0.0715, 0.0233, 0.0170, 0.1236, 0.0046, 0.0091, 0.0327, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0101, 0.0086, 0.0134, 0.0070, 0.0110, 0.0119, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 05:40:03,480 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:40:23,152 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150793.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 05:40:24,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2258, 4.2890, 4.4919, 4.2495, 4.2957, 4.8327, 4.4471, 4.1623], device='cuda:0'), covar=tensor([0.1590, 0.1945, 0.1731, 0.2134, 0.2623, 0.1024, 0.1335, 0.2275], device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0515, 0.0563, 0.0432, 0.0576, 0.0592, 0.0445, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:40:39,990 INFO [train.py:904] (0/8) Epoch 15, batch 8700, loss[loss=0.1728, simple_loss=0.2654, pruned_loss=0.04008, over 16191.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2728, pruned_loss=0.04524, over 3031544.00 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:41:01,905 INFO [optim.py:368] (0/8) 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 05:41:18,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1538, 1.9993, 1.9921, 3.7799, 1.9152, 2.4472, 2.1487, 2.1633], device='cuda:0'), covar=tensor([0.1111, 0.3643, 0.2985, 0.0474, 0.4296, 0.2382, 0.3550, 0.3306], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0409, 0.0342, 0.0312, 0.0418, 0.0467, 0.0376, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:41:25,366 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8318, 5.1707, 5.3398, 5.1582, 5.1446, 5.7204, 5.2621, 5.0014], device='cuda:0'), covar=tensor([0.0893, 0.1755, 0.1768, 0.1820, 0.2219, 0.0844, 0.1205, 0.1985], device='cuda:0'), in_proj_covar=tensor([0.0363, 0.0513, 0.0560, 0.0432, 0.0573, 0.0590, 0.0444, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:41:27,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5947, 2.5940, 2.1317, 3.8370, 1.9797, 3.8299, 1.3292, 2.6974], device='cuda:0'), covar=tensor([0.1645, 0.0838, 0.1489, 0.0236, 0.0124, 0.0414, 0.1956, 0.0887], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0162, 0.0184, 0.0162, 0.0194, 0.0205, 0.0187, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004], device='cuda:0') 2023-04-30 05:42:16,612 INFO [train.py:904] (0/8) Epoch 15, batch 8750, loss[loss=0.1947, simple_loss=0.2795, pruned_loss=0.05498, over 12109.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.273, pruned_loss=0.04466, over 3042458.97 frames. ], batch size: 248, lr: 4.48e-03, grad_scale: 4.0 2023-04-30 05:42:22,808 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 05:42:57,872 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2280, 2.0277, 2.1170, 3.8536, 2.0776, 2.4414, 2.2136, 2.1850], device='cuda:0'), covar=tensor([0.1084, 0.3922, 0.2932, 0.0466, 0.4346, 0.2632, 0.3411, 0.3843], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0408, 0.0342, 0.0311, 0.0418, 0.0467, 0.0376, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:43:52,660 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=150893.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:06,398 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=150900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:44:08,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3873, 3.6387, 3.7163, 2.6793, 3.3431, 3.7005, 3.5554, 2.1540], device='cuda:0'), covar=tensor([0.0424, 0.0046, 0.0040, 0.0314, 0.0085, 0.0067, 0.0058, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0072, 0.0072, 0.0130, 0.0086, 0.0094, 0.0083, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 05:44:08,764 INFO [train.py:904] (0/8) Epoch 15, batch 8800, loss[loss=0.1828, simple_loss=0.2796, pruned_loss=0.04307, over 16893.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2716, pruned_loss=0.0435, over 3070893.30 frames. ], batch size: 116, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:44:28,835 INFO [optim.py:368] (0/8) 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 05:45:06,335 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=150930.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:45:50,526 INFO [train.py:904] (0/8) Epoch 15, batch 8850, loss[loss=0.1628, simple_loss=0.2518, pruned_loss=0.03697, over 12177.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2735, pruned_loss=0.04272, over 3056168.28 frames. ], batch size: 247, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:46:10,296 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=150961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:46:46,311 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=150978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:47:36,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8329, 4.7859, 4.5930, 4.2019, 4.7199, 1.8762, 4.4719, 4.5795], device='cuda:0'), covar=tensor([0.0073, 0.0062, 0.0189, 0.0287, 0.0085, 0.2446, 0.0114, 0.0159], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0129, 0.0176, 0.0161, 0.0149, 0.0190, 0.0164, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:47:36,875 INFO [train.py:904] (0/8) Epoch 15, batch 8900, loss[loss=0.1872, simple_loss=0.2864, pruned_loss=0.04402, over 16746.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2738, pruned_loss=0.042, over 3056170.05 frames. ], batch size: 83, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:47:59,490 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.270e+02 2.672e+02 3.305e+02 7.174e+02, threshold=5.344e+02, percent-clipped=2.0 2023-04-30 05:49:35,306 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-04-30 05:49:42,119 INFO [train.py:904] (0/8) Epoch 15, batch 8950, loss[loss=0.1716, simple_loss=0.2577, pruned_loss=0.04281, over 12733.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2733, pruned_loss=0.04225, over 3048401.45 frames. ], batch size: 246, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:50:15,275 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 05:50:28,125 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2394, 3.3462, 1.8864, 3.6487, 2.4443, 3.5992, 2.2298, 2.6773], device='cuda:0'), covar=tensor([0.0268, 0.0350, 0.1723, 0.0151, 0.0845, 0.0546, 0.1518, 0.0737], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0161, 0.0180, 0.0134, 0.0162, 0.0197, 0.0191, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 05:51:01,608 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151088.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 05:51:31,008 INFO [train.py:904] (0/8) Epoch 15, batch 9000, loss[loss=0.1695, simple_loss=0.2751, pruned_loss=0.03196, over 15643.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2702, pruned_loss=0.04055, over 3069547.08 frames. ], batch size: 194, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:51:31,010 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 05:51:40,826 INFO [train.py:938] (0/8) Epoch 15, validation: loss=0.15, simple_loss=0.2539, pruned_loss=0.02307, over 944034.00 frames. 2023-04-30 05:51:40,827 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 05:52:03,719 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.315e+02 2.164e+02 2.577e+02 3.241e+02 6.734e+02, threshold=5.154e+02, percent-clipped=2.0 2023-04-30 05:53:23,292 INFO [train.py:904] (0/8) Epoch 15, batch 9050, loss[loss=0.1489, simple_loss=0.2397, pruned_loss=0.02909, over 16818.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2712, pruned_loss=0.04131, over 3064633.02 frames. ], batch size: 90, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:54:27,754 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:54:51,449 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:55:09,981 INFO [train.py:904] (0/8) Epoch 15, batch 9100, loss[loss=0.1993, simple_loss=0.3051, pruned_loss=0.04671, over 16155.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.271, pruned_loss=0.04198, over 3054273.16 frames. ], batch size: 165, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:55:31,073 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 2.276e+02 2.684e+02 3.281e+02 5.650e+02, threshold=5.369e+02, percent-clipped=3.0 2023-04-30 05:56:44,481 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:56:54,423 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 15, batch 9150, loss[loss=0.1715, simple_loss=0.2631, pruned_loss=0.03997, over 16414.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2712, pruned_loss=0.04147, over 3060081.77 frames. ], batch size: 147, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:57:18,560 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151256.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:12,679 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:24,972 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9197, 5.1974, 4.9974, 5.0066, 4.7340, 4.6979, 4.6307, 5.2734], device='cuda:0'), covar=tensor([0.1056, 0.0874, 0.0829, 0.0717, 0.0831, 0.0866, 0.1047, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0579, 0.0717, 0.0584, 0.0514, 0.0449, 0.0466, 0.0593, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 05:58:46,511 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151298.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 05:58:51,682 INFO [train.py:904] (0/8) Epoch 15, batch 9200, loss[loss=0.1858, simple_loss=0.2764, pruned_loss=0.04762, over 16940.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2667, pruned_loss=0.04055, over 3081981.36 frames. ], batch size: 116, lr: 4.48e-03, grad_scale: 8.0 2023-04-30 05:59:12,242 INFO [optim.py:368] (0/8) 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 06:00:09,350 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151343.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:00:24,883 INFO [train.py:904] (0/8) Epoch 15, batch 9250, loss[loss=0.1522, simple_loss=0.2377, pruned_loss=0.03333, over 12572.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2663, pruned_loss=0.04062, over 3066506.19 frames. ], batch size: 247, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:00:40,392 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:00:52,544 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7517, 2.8835, 2.7733, 4.4447, 3.1005, 4.1472, 1.5118, 3.1685], device='cuda:0'), covar=tensor([0.1326, 0.0644, 0.0974, 0.0121, 0.0145, 0.0314, 0.1544, 0.0616], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0160, 0.0182, 0.0158, 0.0189, 0.0202, 0.0185, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-04-30 06:01:43,211 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:02:14,985 INFO [train.py:904] (0/8) Epoch 15, batch 9300, loss[loss=0.1537, simple_loss=0.2455, pruned_loss=0.03097, over 16729.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2649, pruned_loss=0.04014, over 3060181.81 frames. ], batch size: 83, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:02:37,915 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.525e+02 2.225e+02 2.604e+02 3.286e+02 5.862e+02, threshold=5.207e+02, percent-clipped=0.0 2023-04-30 06:03:01,222 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3198, 1.5843, 1.9392, 2.2582, 2.2919, 2.5188, 1.6541, 2.4932], device='cuda:0'), covar=tensor([0.0159, 0.0448, 0.0289, 0.0263, 0.0259, 0.0172, 0.0453, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0175, 0.0161, 0.0163, 0.0174, 0.0131, 0.0176, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-30 06:03:08,368 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9422, 4.2195, 4.0876, 4.0800, 3.7584, 3.8232, 3.8342, 4.2234], device='cuda:0'), covar=tensor([0.1116, 0.0894, 0.0880, 0.0734, 0.0858, 0.1594, 0.0959, 0.0968], device='cuda:0'), in_proj_covar=tensor([0.0574, 0.0716, 0.0580, 0.0512, 0.0448, 0.0464, 0.0591, 0.0543], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:03:10,237 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6591, 3.8376, 2.8903, 2.2101, 2.4257, 2.3156, 4.0841, 3.4290], device='cuda:0'), covar=tensor([0.2791, 0.0615, 0.1713, 0.2787, 0.2679, 0.2085, 0.0412, 0.1137], device='cuda:0'), in_proj_covar=tensor([0.0304, 0.0254, 0.0284, 0.0286, 0.0270, 0.0230, 0.0268, 0.0299], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:03:29,232 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:03:59,122 INFO [train.py:904] (0/8) Epoch 15, batch 9350, loss[loss=0.1775, simple_loss=0.2699, pruned_loss=0.04254, over 16582.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2646, pruned_loss=0.03963, over 3066712.10 frames. ], batch size: 68, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:05:36,969 INFO [train.py:904] (0/8) Epoch 15, batch 9400, loss[loss=0.1861, simple_loss=0.284, pruned_loss=0.04409, over 16772.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2644, pruned_loss=0.03944, over 3056293.57 frames. ], batch size: 124, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:05:59,183 INFO [optim.py:368] (0/8) 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 06:06:32,130 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3463, 3.4465, 3.6737, 3.6672, 3.6903, 3.4670, 3.5466, 3.5634], device='cuda:0'), covar=tensor([0.0365, 0.0728, 0.0573, 0.0628, 0.0618, 0.0637, 0.0709, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0345, 0.0369, 0.0363, 0.0346, 0.0414, 0.0384, 0.0470, 0.0308], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 06:06:55,130 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151540.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:07:17,466 INFO [train.py:904] (0/8) Epoch 15, batch 9450, loss[loss=0.1749, simple_loss=0.2669, pruned_loss=0.04144, over 16304.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.266, pruned_loss=0.03985, over 3048787.94 frames. ], batch size: 146, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:07:24,173 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151556.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:08:58,274 INFO [train.py:904] (0/8) Epoch 15, batch 9500, loss[loss=0.1676, simple_loss=0.2687, pruned_loss=0.03323, over 16863.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2649, pruned_loss=0.03924, over 3052942.45 frames. ], batch size: 102, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:09:01,598 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3932, 1.9811, 1.6581, 1.6617, 2.2675, 1.9225, 2.0470, 2.3616], device='cuda:0'), covar=tensor([0.0153, 0.0355, 0.0451, 0.0459, 0.0249, 0.0343, 0.0172, 0.0230], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0213, 0.0207, 0.0207, 0.0212, 0.0213, 0.0210, 0.0203], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:09:04,124 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:09:22,320 INFO [optim.py:368] (0/8) 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 06:10:11,842 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151638.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:10:41,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5807, 3.5218, 3.5473, 3.0036, 3.5302, 1.9401, 3.3178, 3.1131], device='cuda:0'), covar=tensor([0.0114, 0.0103, 0.0146, 0.0203, 0.0092, 0.2275, 0.0121, 0.0247], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0127, 0.0171, 0.0155, 0.0147, 0.0187, 0.0161, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:10:44,459 INFO [train.py:904] (0/8) Epoch 15, batch 9550, loss[loss=0.1854, simple_loss=0.286, pruned_loss=0.04244, over 15336.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.265, pruned_loss=0.03936, over 3062603.89 frames. ], batch size: 190, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:10:49,191 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=151654.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:11:30,387 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 06:11:58,511 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151688.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:12:24,739 INFO [train.py:904] (0/8) Epoch 15, batch 9600, loss[loss=0.1855, simple_loss=0.2898, pruned_loss=0.04057, over 16687.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2664, pruned_loss=0.04007, over 3053944.18 frames. ], batch size: 134, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:12:44,297 INFO [optim.py:368] (0/8) 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 06:14:04,336 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:14:11,698 INFO [train.py:904] (0/8) Epoch 15, batch 9650, loss[loss=0.1826, simple_loss=0.2715, pruned_loss=0.04685, over 16989.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2684, pruned_loss=0.04038, over 3064692.49 frames. ], batch size: 109, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:14:43,571 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4125, 2.1263, 2.1132, 4.0522, 2.0318, 2.5054, 2.2762, 2.2613], device='cuda:0'), covar=tensor([0.0966, 0.3533, 0.2909, 0.0429, 0.4310, 0.2550, 0.3193, 0.3699], device='cuda:0'), in_proj_covar=tensor([0.0366, 0.0402, 0.0339, 0.0308, 0.0414, 0.0457, 0.0371, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:15:57,485 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=151801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:15:58,772 INFO [train.py:904] (0/8) Epoch 15, batch 9700, loss[loss=0.1775, simple_loss=0.2614, pruned_loss=0.04678, over 12336.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2672, pruned_loss=0.04024, over 3052070.88 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:16:19,907 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.752e+02 2.225e+02 2.684e+02 3.461e+02 6.863e+02, threshold=5.368e+02, percent-clipped=3.0 2023-04-30 06:17:00,719 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 06:17:18,833 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151840.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:17:41,659 INFO [train.py:904] (0/8) Epoch 15, batch 9750, loss[loss=0.1754, simple_loss=0.2663, pruned_loss=0.04227, over 16900.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2665, pruned_loss=0.04054, over 3061818.09 frames. ], batch size: 116, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:18:01,518 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=151862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:18:51,598 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8651, 2.8086, 2.6522, 1.9712, 2.4907, 2.7533, 2.7257, 1.9389], device='cuda:0'), covar=tensor([0.0396, 0.0052, 0.0051, 0.0327, 0.0111, 0.0072, 0.0076, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0128, 0.0070, 0.0070, 0.0127, 0.0085, 0.0093, 0.0082, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 06:18:55,495 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:19:18,829 INFO [train.py:904] (0/8) Epoch 15, batch 9800, loss[loss=0.18, simple_loss=0.2642, pruned_loss=0.04786, over 12548.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2661, pruned_loss=0.03974, over 3059120.65 frames. ], batch size: 248, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:19:37,433 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-04-30 06:19:40,663 INFO [optim.py:368] (0/8) 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 06:19:58,766 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1601, 1.5145, 1.8911, 2.0337, 2.1148, 2.3024, 1.7411, 2.3099], device='cuda:0'), covar=tensor([0.0200, 0.0404, 0.0241, 0.0303, 0.0280, 0.0211, 0.0390, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0175, 0.0160, 0.0164, 0.0173, 0.0131, 0.0176, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001], device='cuda:0') 2023-04-30 06:20:30,076 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=151938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:21:03,080 INFO [train.py:904] (0/8) Epoch 15, batch 9850, loss[loss=0.1948, simple_loss=0.2849, pruned_loss=0.05235, over 16699.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2674, pruned_loss=0.03915, over 3082004.35 frames. ], batch size: 134, lr: 4.47e-03, grad_scale: 8.0 2023-04-30 06:21:08,470 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=151986.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:22:50,430 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-152000.pt 2023-04-30 06:22:57,942 INFO [train.py:904] (0/8) Epoch 15, batch 9900, loss[loss=0.1823, simple_loss=0.2796, pruned_loss=0.04256, over 16942.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2672, pruned_loss=0.03906, over 3060690.18 frames. ], batch size: 109, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:22:58,683 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152002.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:23:18,285 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 06:23:24,824 INFO [optim.py:368] (0/8) 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 06:23:25,601 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9424, 4.2416, 4.0998, 4.1469, 3.7623, 3.8215, 3.8420, 4.2485], device='cuda:0'), covar=tensor([0.1104, 0.0940, 0.0938, 0.0628, 0.0795, 0.1639, 0.0918, 0.0981], device='cuda:0'), in_proj_covar=tensor([0.0573, 0.0707, 0.0575, 0.0507, 0.0445, 0.0459, 0.0584, 0.0539], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:24:13,294 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9569, 4.0918, 3.8959, 3.6446, 3.6088, 4.0360, 3.6430, 3.7949], device='cuda:0'), covar=tensor([0.0617, 0.0475, 0.0297, 0.0263, 0.0714, 0.0439, 0.0929, 0.0497], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0338, 0.0292, 0.0276, 0.0301, 0.0319, 0.0200, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-30 06:24:37,872 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:24:55,591 INFO [train.py:904] (0/8) Epoch 15, batch 9950, loss[loss=0.1614, simple_loss=0.2575, pruned_loss=0.03264, over 16465.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.269, pruned_loss=0.03925, over 3060282.56 frames. ], batch size: 68, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:26:56,982 INFO [train.py:904] (0/8) Epoch 15, batch 10000, loss[loss=0.1713, simple_loss=0.2691, pruned_loss=0.0367, over 15370.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2676, pruned_loss=0.03901, over 3075445.59 frames. ], batch size: 190, lr: 4.46e-03, grad_scale: 8.0 2023-04-30 06:27:01,545 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8199, 3.7408, 3.9895, 3.7452, 3.9043, 4.3250, 4.0188, 3.7450], device='cuda:0'), covar=tensor([0.1994, 0.2624, 0.2149, 0.2460, 0.2717, 0.1528, 0.1519, 0.2263], device='cuda:0'), in_proj_covar=tensor([0.0362, 0.0510, 0.0559, 0.0431, 0.0571, 0.0596, 0.0441, 0.0571], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 06:27:18,767 INFO [optim.py:368] (0/8) 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 06:27:32,360 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-04-30 06:28:35,916 INFO [train.py:904] (0/8) Epoch 15, batch 10050, loss[loss=0.1771, simple_loss=0.2721, pruned_loss=0.04108, over 16500.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2682, pruned_loss=0.03929, over 3073585.60 frames. ], batch size: 68, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:28:45,753 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152157.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:28:48,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3216, 4.3415, 4.1948, 3.8756, 3.9061, 4.3092, 4.0443, 3.9793], device='cuda:0'), covar=tensor([0.0512, 0.0555, 0.0274, 0.0290, 0.0770, 0.0459, 0.0636, 0.0663], device='cuda:0'), in_proj_covar=tensor([0.0248, 0.0338, 0.0291, 0.0275, 0.0301, 0.0319, 0.0199, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-30 06:29:27,511 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5532, 4.8448, 4.6688, 4.6693, 4.3744, 4.3625, 4.2724, 4.9205], device='cuda:0'), covar=tensor([0.1016, 0.0824, 0.0855, 0.0656, 0.0758, 0.1126, 0.1077, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0570, 0.0706, 0.0573, 0.0505, 0.0444, 0.0456, 0.0583, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:30:08,504 INFO [train.py:904] (0/8) Epoch 15, batch 10100, loss[loss=0.1761, simple_loss=0.2634, pruned_loss=0.0444, over 16361.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.268, pruned_loss=0.03947, over 3048761.20 frames. ], batch size: 146, lr: 4.46e-03, grad_scale: 4.0 2023-04-30 06:30:13,047 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9806, 4.0185, 4.3170, 4.2860, 4.3094, 4.0868, 4.0722, 4.0500], device='cuda:0'), covar=tensor([0.0393, 0.1008, 0.0454, 0.0499, 0.0474, 0.0472, 0.0821, 0.0527], device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0361, 0.0354, 0.0342, 0.0402, 0.0376, 0.0458, 0.0301], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0002], device='cuda:0') 2023-04-30 06:30:28,192 INFO [optim.py:368] (0/8) 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 06:30:55,284 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:31:27,549 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-15.pt 2023-04-30 06:31:53,402 INFO [train.py:904] (0/8) Epoch 16, batch 0, loss[loss=0.2651, simple_loss=0.3164, pruned_loss=0.1068, over 16877.00 frames. ], tot_loss[loss=0.2651, simple_loss=0.3164, pruned_loss=0.1068, over 16877.00 frames. ], batch size: 116, lr: 4.32e-03, grad_scale: 8.0 2023-04-30 06:31:53,403 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 06:32:00,893 INFO [train.py:938] (0/8) Epoch 16, validation: loss=0.1492, simple_loss=0.2525, pruned_loss=0.02288, over 944034.00 frames. 2023-04-30 06:32:00,894 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 06:32:10,227 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 06:32:29,746 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:32:29,883 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3609, 3.6288, 4.0227, 2.2342, 3.0539, 2.5319, 3.7376, 3.7962], device='cuda:0'), covar=tensor([0.0323, 0.0767, 0.0471, 0.1839, 0.0847, 0.0901, 0.0690, 0.1009], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0145, 0.0157, 0.0145, 0.0137, 0.0124, 0.0137, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 06:32:48,516 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:33:09,970 INFO [train.py:904] (0/8) Epoch 16, batch 50, loss[loss=0.2101, simple_loss=0.2799, pruned_loss=0.07016, over 16764.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2803, pruned_loss=0.05863, over 750433.37 frames. ], batch size: 124, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:33:29,877 INFO [optim.py:368] (0/8) 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 06:33:53,597 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:07,525 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:34:18,733 INFO [train.py:904] (0/8) Epoch 16, batch 100, loss[loss=0.168, simple_loss=0.2504, pruned_loss=0.04284, over 15859.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2751, pruned_loss=0.0566, over 1313758.92 frames. ], batch size: 35, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:35:14,871 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:35:17,224 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6294, 2.3733, 1.8313, 2.1270, 2.8315, 2.5531, 2.8516, 2.8971], device='cuda:0'), covar=tensor([0.0172, 0.0343, 0.0496, 0.0454, 0.0192, 0.0282, 0.0199, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0219, 0.0212, 0.0213, 0.0217, 0.0220, 0.0217, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:35:21,882 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 06:35:26,683 INFO [train.py:904] (0/8) Epoch 16, batch 150, loss[loss=0.1471, simple_loss=0.2442, pruned_loss=0.02502, over 17313.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2719, pruned_loss=0.05309, over 1762594.02 frames. ], batch size: 52, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:35:48,056 INFO [optim.py:368] (0/8) 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 06:36:25,033 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:36:35,197 INFO [train.py:904] (0/8) Epoch 16, batch 200, loss[loss=0.2027, simple_loss=0.2865, pruned_loss=0.05949, over 17040.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2719, pruned_loss=0.05274, over 2107666.30 frames. ], batch size: 55, lr: 4.32e-03, grad_scale: 1.0 2023-04-30 06:36:42,913 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:44,241 INFO [train.py:904] (0/8) Epoch 16, batch 250, loss[loss=0.1692, simple_loss=0.2639, pruned_loss=0.03727, over 17021.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2686, pruned_loss=0.05118, over 2378586.46 frames. ], batch size: 53, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:37:48,059 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152505.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:37:49,480 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:38:05,713 INFO [optim.py:368] (0/8) 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 06:38:27,805 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4853, 2.5306, 2.1627, 2.3899, 2.9393, 2.7364, 3.1412, 3.1306], device='cuda:0'), covar=tensor([0.0126, 0.0380, 0.0473, 0.0419, 0.0256, 0.0334, 0.0294, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0221, 0.0213, 0.0214, 0.0219, 0.0221, 0.0220, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:38:53,037 INFO [train.py:904] (0/8) Epoch 16, batch 300, loss[loss=0.1811, simple_loss=0.2578, pruned_loss=0.0522, over 16112.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2661, pruned_loss=0.0494, over 2590973.98 frames. ], batch size: 164, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:39:33,837 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152581.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:34,995 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152582.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:43,038 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:39:51,043 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0852, 5.0444, 4.8939, 4.4911, 4.5361, 4.9900, 4.9549, 4.6213], device='cuda:0'), covar=tensor([0.0600, 0.0471, 0.0337, 0.0339, 0.1075, 0.0412, 0.0321, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0263, 0.0361, 0.0310, 0.0294, 0.0323, 0.0339, 0.0212, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:40:01,539 INFO [train.py:904] (0/8) Epoch 16, batch 350, loss[loss=0.1657, simple_loss=0.2528, pruned_loss=0.03929, over 17181.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2638, pruned_loss=0.04809, over 2751762.93 frames. ], batch size: 46, lr: 4.31e-03, grad_scale: 1.0 2023-04-30 06:40:20,733 INFO [optim.py:368] (0/8) 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 06:40:38,389 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:40:40,048 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 06:40:58,740 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:05,940 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 06:41:07,041 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152650.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:41:07,382 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-30 06:41:08,832 INFO [train.py:904] (0/8) Epoch 16, batch 400, loss[loss=0.1772, simple_loss=0.2539, pruned_loss=0.05029, over 12195.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2618, pruned_loss=0.04756, over 2874156.91 frames. ], batch size: 247, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:41:29,091 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-30 06:42:00,953 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 06:42:16,096 INFO [train.py:904] (0/8) Epoch 16, batch 450, loss[loss=0.1473, simple_loss=0.2305, pruned_loss=0.03205, over 16532.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2605, pruned_loss=0.04708, over 2982166.34 frames. ], batch size: 75, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:42:36,234 INFO [optim.py:368] (0/8) 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 06:42:49,873 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=152726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:43:25,830 INFO [train.py:904] (0/8) Epoch 16, batch 500, loss[loss=0.2078, simple_loss=0.2957, pruned_loss=0.05998, over 16676.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2587, pruned_loss=0.04503, over 3060698.25 frames. ], batch size: 57, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:43:33,065 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-04-30 06:43:58,678 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7914, 2.7370, 2.4244, 2.6911, 3.1731, 2.9767, 3.4548, 3.3634], device='cuda:0'), covar=tensor([0.0114, 0.0370, 0.0442, 0.0400, 0.0250, 0.0330, 0.0260, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0222, 0.0215, 0.0216, 0.0221, 0.0223, 0.0222, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:44:14,814 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=152787.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:33,459 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152801.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:44:34,225 INFO [train.py:904] (0/8) Epoch 16, batch 550, loss[loss=0.2514, simple_loss=0.3141, pruned_loss=0.09436, over 16478.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2584, pruned_loss=0.04529, over 3116786.31 frames. ], batch size: 146, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:44:55,530 INFO [optim.py:368] (0/8) 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 06:45:09,580 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-04-30 06:45:46,482 INFO [train.py:904] (0/8) Epoch 16, batch 600, loss[loss=0.1752, simple_loss=0.2466, pruned_loss=0.05191, over 16491.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2582, pruned_loss=0.04554, over 3152667.92 frames. ], batch size: 68, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:45:51,853 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.12 vs. limit=5.0 2023-04-30 06:46:24,490 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:46:53,570 INFO [train.py:904] (0/8) Epoch 16, batch 650, loss[loss=0.1706, simple_loss=0.2467, pruned_loss=0.04725, over 12222.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2572, pruned_loss=0.04546, over 3186322.10 frames. ], batch size: 246, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:47:14,342 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:30,709 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=152929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:43,479 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152938.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:47:52,084 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=152945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:48:00,889 INFO [train.py:904] (0/8) Epoch 16, batch 700, loss[loss=0.1644, simple_loss=0.251, pruned_loss=0.03889, over 17074.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2569, pruned_loss=0.0452, over 3221737.47 frames. ], batch size: 53, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:48:37,136 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=152977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:49:10,753 INFO [train.py:904] (0/8) Epoch 16, batch 750, loss[loss=0.1604, simple_loss=0.2555, pruned_loss=0.03261, over 17272.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2567, pruned_loss=0.04487, over 3248476.77 frames. ], batch size: 52, lr: 4.31e-03, grad_scale: 2.0 2023-04-30 06:49:31,101 INFO [optim.py:368] (0/8) 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 06:49:36,293 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7953, 2.8243, 2.6727, 5.0538, 4.1603, 4.4161, 1.4448, 3.2106], device='cuda:0'), covar=tensor([0.1348, 0.0741, 0.1201, 0.0193, 0.0238, 0.0362, 0.1673, 0.0732], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0164, 0.0186, 0.0166, 0.0195, 0.0210, 0.0190, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 06:49:53,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5084, 2.2155, 2.2803, 4.4126, 2.2096, 2.6441, 2.3024, 2.4278], device='cuda:0'), covar=tensor([0.1089, 0.3643, 0.2848, 0.0412, 0.3931, 0.2603, 0.3343, 0.3625], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0416, 0.0350, 0.0321, 0.0426, 0.0477, 0.0385, 0.0487], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:50:17,798 INFO [train.py:904] (0/8) Epoch 16, batch 800, loss[loss=0.1492, simple_loss=0.2363, pruned_loss=0.0311, over 16872.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2567, pruned_loss=0.04512, over 3267898.45 frames. ], batch size: 42, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:00,659 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:51:24,864 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 16, batch 850, loss[loss=0.1594, simple_loss=0.2431, pruned_loss=0.03792, over 15496.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2568, pruned_loss=0.04426, over 3282435.36 frames. ], batch size: 190, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:51:46,529 INFO [optim.py:368] (0/8) 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 06:51:46,980 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153117.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:32,211 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153149.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:52:35,403 INFO [train.py:904] (0/8) Epoch 16, batch 900, loss[loss=0.1887, simple_loss=0.2631, pruned_loss=0.05714, over 16891.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2562, pruned_loss=0.04345, over 3299153.66 frames. ], batch size: 116, lr: 4.31e-03, grad_scale: 4.0 2023-04-30 06:53:11,028 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:53:43,183 INFO [train.py:904] (0/8) Epoch 16, batch 950, loss[loss=0.1803, simple_loss=0.2509, pruned_loss=0.05486, over 16853.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2561, pruned_loss=0.04371, over 3305386.76 frames. ], batch size: 116, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:54:04,605 INFO [optim.py:368] (0/8) 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 06:54:19,441 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1762, 2.9262, 3.1247, 1.9677, 3.2632, 3.2981, 2.7348, 2.5711], device='cuda:0'), covar=tensor([0.0800, 0.0253, 0.0262, 0.1088, 0.0116, 0.0249, 0.0465, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0105, 0.0093, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 06:54:33,138 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153238.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:42,562 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153245.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:54:43,857 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9740, 4.0524, 4.2949, 4.2836, 4.2966, 4.0521, 4.0666, 4.0152], device='cuda:0'), covar=tensor([0.0368, 0.0740, 0.0424, 0.0419, 0.0535, 0.0447, 0.0794, 0.0530], device='cuda:0'), in_proj_covar=tensor([0.0375, 0.0406, 0.0394, 0.0378, 0.0446, 0.0419, 0.0511, 0.0333], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 06:54:53,052 INFO [train.py:904] (0/8) Epoch 16, batch 1000, loss[loss=0.1795, simple_loss=0.2795, pruned_loss=0.03968, over 17254.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2549, pruned_loss=0.04396, over 3316825.47 frames. ], batch size: 52, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:55:25,391 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 06:55:29,781 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153278.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:40,236 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153286.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:55:50,997 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:02,861 INFO [train.py:904] (0/8) Epoch 16, batch 1050, loss[loss=0.1825, simple_loss=0.2627, pruned_loss=0.05111, over 16735.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2545, pruned_loss=0.04339, over 3314190.50 frames. ], batch size: 134, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:56:24,685 INFO [optim.py:368] (0/8) 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 06:56:29,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3925, 3.4405, 2.2079, 3.6970, 2.7081, 3.6572, 2.2798, 2.8485], device='cuda:0'), covar=tensor([0.0260, 0.0405, 0.1519, 0.0279, 0.0732, 0.0789, 0.1365, 0.0643], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0173, 0.0194, 0.0152, 0.0175, 0.0215, 0.0203, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 06:56:50,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7780, 1.8980, 2.3463, 2.6054, 2.6721, 2.6252, 1.8787, 2.9293], device='cuda:0'), covar=tensor([0.0146, 0.0419, 0.0278, 0.0253, 0.0254, 0.0293, 0.0457, 0.0133], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0173, 0.0182, 0.0140, 0.0185, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:56:55,493 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153339.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:56:58,682 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 06:57:12,831 INFO [train.py:904] (0/8) Epoch 16, batch 1100, loss[loss=0.1642, simple_loss=0.2422, pruned_loss=0.04312, over 15638.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.254, pruned_loss=0.0433, over 3302392.76 frames. ], batch size: 191, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:57:53,790 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:21,563 INFO [train.py:904] (0/8) Epoch 16, batch 1150, loss[loss=0.1793, simple_loss=0.2527, pruned_loss=0.05292, over 12283.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2534, pruned_loss=0.04244, over 3316708.49 frames. ], batch size: 247, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 06:58:42,014 INFO [optim.py:368] (0/8) 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 06:58:46,809 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:58:58,554 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153430.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 06:59:16,712 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9546, 4.7165, 4.9945, 5.2065, 5.3734, 4.7259, 5.3647, 5.3666], device='cuda:0'), covar=tensor([0.1886, 0.1383, 0.1740, 0.0715, 0.0519, 0.0853, 0.0572, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0609, 0.0755, 0.0894, 0.0766, 0.0576, 0.0600, 0.0617, 0.0709], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 06:59:27,894 INFO [train.py:904] (0/8) Epoch 16, batch 1200, loss[loss=0.1624, simple_loss=0.2392, pruned_loss=0.04283, over 16305.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2528, pruned_loss=0.0419, over 3322824.85 frames. ], batch size: 165, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 06:59:34,535 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 06:59:48,339 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9710, 4.5748, 4.5362, 3.3616, 3.8432, 4.5050, 4.1872, 2.9307], device='cuda:0'), covar=tensor([0.0432, 0.0049, 0.0036, 0.0288, 0.0092, 0.0065, 0.0063, 0.0368], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0076, 0.0074, 0.0132, 0.0089, 0.0097, 0.0086, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 06:59:57,516 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153473.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:10,711 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:00:37,164 INFO [train.py:904] (0/8) Epoch 16, batch 1250, loss[loss=0.1648, simple_loss=0.2559, pruned_loss=0.03685, over 17202.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2531, pruned_loss=0.04256, over 3321453.53 frames. ], batch size: 46, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:00:57,398 INFO [optim.py:368] (0/8) 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 07:01:03,366 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 07:01:43,343 INFO [train.py:904] (0/8) Epoch 16, batch 1300, loss[loss=0.1629, simple_loss=0.2383, pruned_loss=0.04379, over 15989.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2528, pruned_loss=0.04309, over 3324356.27 frames. ], batch size: 35, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:02:52,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8644, 3.8040, 4.4177, 2.1660, 4.5637, 4.7449, 3.2822, 3.5518], device='cuda:0'), covar=tensor([0.0681, 0.0246, 0.0224, 0.1108, 0.0083, 0.0121, 0.0399, 0.0364], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0139, 0.0074, 0.0118, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 07:02:52,650 INFO [train.py:904] (0/8) Epoch 16, batch 1350, loss[loss=0.1819, simple_loss=0.2812, pruned_loss=0.04128, over 17099.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2532, pruned_loss=0.04287, over 3321266.19 frames. ], batch size: 47, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:03:12,846 INFO [optim.py:368] (0/8) 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 07:03:37,004 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:03:37,215 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1475, 2.5325, 2.0623, 2.3580, 2.9196, 2.6460, 3.0517, 3.0210], device='cuda:0'), covar=tensor([0.0247, 0.0354, 0.0499, 0.0403, 0.0238, 0.0335, 0.0292, 0.0261], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0227, 0.0219, 0.0220, 0.0228, 0.0229, 0.0232, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:04:02,081 INFO [train.py:904] (0/8) Epoch 16, batch 1400, loss[loss=0.1612, simple_loss=0.236, pruned_loss=0.04315, over 16353.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2533, pruned_loss=0.04313, over 3324395.81 frames. ], batch size: 165, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:04:53,166 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2152, 4.2009, 4.5585, 4.5587, 4.5976, 4.2803, 4.3302, 4.2095], device='cuda:0'), covar=tensor([0.0304, 0.0707, 0.0410, 0.0416, 0.0402, 0.0432, 0.0714, 0.0525], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0413, 0.0401, 0.0383, 0.0453, 0.0426, 0.0518, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 07:05:12,110 INFO [train.py:904] (0/8) Epoch 16, batch 1450, loss[loss=0.1657, simple_loss=0.2624, pruned_loss=0.03451, over 17030.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2531, pruned_loss=0.04326, over 3314024.71 frames. ], batch size: 50, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:05:34,103 INFO [optim.py:368] (0/8) 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 07:05:53,387 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2846, 4.2246, 4.6996, 2.7161, 4.8654, 4.9798, 3.5902, 3.9352], device='cuda:0'), covar=tensor([0.0625, 0.0165, 0.0168, 0.0893, 0.0076, 0.0139, 0.0348, 0.0309], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0104, 0.0092, 0.0138, 0.0073, 0.0117, 0.0124, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 07:06:22,466 INFO [train.py:904] (0/8) Epoch 16, batch 1500, loss[loss=0.1952, simple_loss=0.2688, pruned_loss=0.06078, over 16748.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2527, pruned_loss=0.04311, over 3310950.02 frames. ], batch size: 124, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:06:32,890 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-04-30 07:06:50,976 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:06:51,234 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.55 vs. limit=2.0 2023-04-30 07:06:55,374 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=153777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:07:30,678 INFO [train.py:904] (0/8) Epoch 16, batch 1550, loss[loss=0.1713, simple_loss=0.2605, pruned_loss=0.04103, over 16846.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2528, pruned_loss=0.04397, over 3307322.28 frames. ], batch size: 42, lr: 4.30e-03, grad_scale: 4.0 2023-04-30 07:07:53,744 INFO [optim.py:368] (0/8) 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 07:07:58,217 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153821.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:08:24,186 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3196, 3.3542, 3.5911, 2.4923, 3.2258, 3.6845, 3.4560, 2.1031], device='cuda:0'), covar=tensor([0.0440, 0.0110, 0.0044, 0.0332, 0.0102, 0.0074, 0.0073, 0.0411], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0076, 0.0074, 0.0131, 0.0089, 0.0098, 0.0086, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:08:40,387 INFO [train.py:904] (0/8) Epoch 16, batch 1600, loss[loss=0.2244, simple_loss=0.2999, pruned_loss=0.07441, over 11768.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2549, pruned_loss=0.04503, over 3305238.59 frames. ], batch size: 246, lr: 4.30e-03, grad_scale: 8.0 2023-04-30 07:09:47,775 INFO [train.py:904] (0/8) Epoch 16, batch 1650, loss[loss=0.2024, simple_loss=0.2774, pruned_loss=0.06371, over 16378.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2569, pruned_loss=0.0457, over 3313797.68 frames. ], batch size: 146, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:10:09,061 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.581e+02 2.549e+02 2.974e+02 3.703e+02 7.236e+02, threshold=5.949e+02, percent-clipped=5.0 2023-04-30 07:10:15,996 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=153923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:31,835 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=153934.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:10:56,044 INFO [train.py:904] (0/8) Epoch 16, batch 1700, loss[loss=0.1502, simple_loss=0.24, pruned_loss=0.03013, over 17222.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2584, pruned_loss=0.04599, over 3317343.30 frames. ], batch size: 45, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:11:38,411 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=153982.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:11:40,866 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=153984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:12:03,373 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-154000.pt 2023-04-30 07:12:09,298 INFO [train.py:904] (0/8) Epoch 16, batch 1750, loss[loss=0.1784, simple_loss=0.2723, pruned_loss=0.04219, over 17046.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2589, pruned_loss=0.04562, over 3328245.43 frames. ], batch size: 55, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:12:33,143 INFO [optim.py:368] (0/8) 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 07:13:07,916 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4614, 4.2775, 4.4991, 4.6376, 4.7323, 4.2712, 4.5926, 4.7579], device='cuda:0'), covar=tensor([0.1478, 0.1107, 0.1321, 0.0661, 0.0586, 0.1194, 0.2273, 0.0664], device='cuda:0'), in_proj_covar=tensor([0.0610, 0.0757, 0.0897, 0.0767, 0.0575, 0.0598, 0.0613, 0.0710], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:13:13,266 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6857, 2.3760, 1.9432, 2.1385, 2.8152, 2.5834, 2.8783, 2.9015], device='cuda:0'), covar=tensor([0.0198, 0.0336, 0.0439, 0.0427, 0.0199, 0.0285, 0.0223, 0.0233], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0225, 0.0217, 0.0218, 0.0226, 0.0228, 0.0230, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:13:15,082 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154049.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:13:18,934 INFO [train.py:904] (0/8) Epoch 16, batch 1800, loss[loss=0.1729, simple_loss=0.2535, pruned_loss=0.04611, over 16456.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2597, pruned_loss=0.04523, over 3337677.59 frames. ], batch size: 146, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:13:52,565 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:13:54,222 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:03,016 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:14:05,484 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 07:14:26,517 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-04-30 07:14:28,078 INFO [train.py:904] (0/8) Epoch 16, batch 1850, loss[loss=0.1912, simple_loss=0.2734, pruned_loss=0.05456, over 16431.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2603, pruned_loss=0.0445, over 3340604.29 frames. ], batch size: 146, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:14:38,630 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154110.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 07:14:50,230 INFO [optim.py:368] (0/8) 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 07:14:50,612 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4463, 5.3873, 5.1911, 4.6189, 5.2917, 2.1874, 5.0074, 5.2397], device='cuda:0'), covar=tensor([0.0072, 0.0076, 0.0177, 0.0364, 0.0078, 0.2201, 0.0115, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0140, 0.0187, 0.0172, 0.0160, 0.0200, 0.0176, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:14:59,484 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:13,785 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6166, 4.7895, 4.9118, 4.7567, 4.7407, 5.3575, 4.8847, 4.5485], device='cuda:0'), covar=tensor([0.1706, 0.1922, 0.2222, 0.2237, 0.2896, 0.1049, 0.1539, 0.2869], device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0560, 0.0608, 0.0468, 0.0627, 0.0641, 0.0476, 0.0623], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:15:15,185 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154137.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:15:22,724 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 07:15:25,785 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154145.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:15:33,374 INFO [train.py:904] (0/8) Epoch 16, batch 1900, loss[loss=0.2116, simple_loss=0.2891, pruned_loss=0.0671, over 11796.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2605, pruned_loss=0.04417, over 3333162.01 frames. ], batch size: 248, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:15:53,623 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.10 vs. limit=5.0 2023-04-30 07:15:54,585 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4394, 5.3651, 5.2783, 4.8207, 4.9570, 5.3577, 5.2467, 4.9918], device='cuda:0'), covar=tensor([0.0540, 0.0415, 0.0247, 0.0286, 0.0888, 0.0359, 0.0266, 0.0737], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0391, 0.0332, 0.0320, 0.0348, 0.0369, 0.0227, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:16:44,650 INFO [train.py:904] (0/8) Epoch 16, batch 1950, loss[loss=0.1827, simple_loss=0.2735, pruned_loss=0.04596, over 12211.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2607, pruned_loss=0.04442, over 3326465.80 frames. ], batch size: 247, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:17:04,524 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.551e+02 2.235e+02 2.551e+02 3.026e+02 6.742e+02, threshold=5.103e+02, percent-clipped=2.0 2023-04-30 07:17:09,885 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:17:22,222 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 07:17:51,480 INFO [train.py:904] (0/8) Epoch 16, batch 2000, loss[loss=0.1833, simple_loss=0.2653, pruned_loss=0.05059, over 16215.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2611, pruned_loss=0.04463, over 3317515.16 frames. ], batch size: 165, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:17:56,397 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 07:18:09,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9946, 2.1620, 2.6701, 3.0193, 2.7166, 3.4571, 2.4528, 3.4478], device='cuda:0'), covar=tensor([0.0246, 0.0426, 0.0284, 0.0283, 0.0307, 0.0185, 0.0423, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0186, 0.0172, 0.0176, 0.0186, 0.0142, 0.0186, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:18:27,792 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154279.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:32,760 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154282.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:18:38,792 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:18:59,383 INFO [train.py:904] (0/8) Epoch 16, batch 2050, loss[loss=0.1596, simple_loss=0.2517, pruned_loss=0.03377, over 16761.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2613, pruned_loss=0.04489, over 3321841.24 frames. ], batch size: 57, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:19:19,613 INFO [optim.py:368] (0/8) 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 07:19:38,467 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 07:19:46,274 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154337.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:20:00,347 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:06,435 INFO [train.py:904] (0/8) Epoch 16, batch 2100, loss[loss=0.2251, simple_loss=0.2951, pruned_loss=0.07755, over 16461.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2627, pruned_loss=0.04614, over 3323779.07 frames. ], batch size: 146, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:20:06,853 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154352.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:20:14,022 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9194, 4.8376, 4.7937, 4.4911, 4.4676, 4.8927, 4.6925, 4.5980], device='cuda:0'), covar=tensor([0.0664, 0.0718, 0.0289, 0.0287, 0.0874, 0.0480, 0.0449, 0.0648], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0397, 0.0337, 0.0325, 0.0355, 0.0374, 0.0231, 0.0406], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:21:09,623 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154398.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:21:14,702 INFO [train.py:904] (0/8) Epoch 16, batch 2150, loss[loss=0.2027, simple_loss=0.2788, pruned_loss=0.06324, over 16846.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2638, pruned_loss=0.04722, over 3312585.64 frames. ], batch size: 116, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:21:18,562 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154405.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:21:36,518 INFO [optim.py:368] (0/8) 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 07:21:37,020 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2753, 2.5182, 2.0953, 2.3592, 2.8664, 2.6656, 3.0943, 3.0848], device='cuda:0'), covar=tensor([0.0146, 0.0357, 0.0433, 0.0374, 0.0226, 0.0301, 0.0225, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0225, 0.0218, 0.0218, 0.0227, 0.0229, 0.0232, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:21:57,857 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154432.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:22:07,251 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154440.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:22:25,096 INFO [train.py:904] (0/8) Epoch 16, batch 2200, loss[loss=0.2235, simple_loss=0.3048, pruned_loss=0.07114, over 16678.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.264, pruned_loss=0.04728, over 3317878.14 frames. ], batch size: 68, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:22:55,691 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:23:32,255 INFO [train.py:904] (0/8) Epoch 16, batch 2250, loss[loss=0.169, simple_loss=0.2517, pruned_loss=0.04315, over 16813.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2644, pruned_loss=0.04749, over 3307360.41 frames. ], batch size: 96, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:23:55,134 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.236e+02 2.597e+02 3.012e+02 8.725e+02, threshold=5.194e+02, percent-clipped=2.0 2023-04-30 07:24:18,258 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154535.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:24:40,022 INFO [train.py:904] (0/8) Epoch 16, batch 2300, loss[loss=0.1772, simple_loss=0.2603, pruned_loss=0.04707, over 16509.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2646, pruned_loss=0.04764, over 3306225.47 frames. ], batch size: 68, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:24:48,750 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5759, 5.9592, 5.6612, 5.7478, 5.3777, 5.3103, 5.3309, 6.0489], device='cuda:0'), covar=tensor([0.1189, 0.0808, 0.0991, 0.0696, 0.0877, 0.0680, 0.1034, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0637, 0.0786, 0.0641, 0.0566, 0.0495, 0.0503, 0.0654, 0.0603], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:25:16,146 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154577.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:25:20,151 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154579.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:25:22,497 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8674, 3.0364, 2.7802, 4.9145, 3.8126, 4.4073, 1.8582, 3.1669], device='cuda:0'), covar=tensor([0.1330, 0.0855, 0.1225, 0.0206, 0.0269, 0.0401, 0.1555, 0.0816], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0167, 0.0187, 0.0172, 0.0200, 0.0214, 0.0190, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 07:25:50,715 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 07:25:51,254 INFO [train.py:904] (0/8) Epoch 16, batch 2350, loss[loss=0.1993, simple_loss=0.2836, pruned_loss=0.05753, over 16466.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2649, pruned_loss=0.04725, over 3314826.55 frames. ], batch size: 75, lr: 4.29e-03, grad_scale: 8.0 2023-04-30 07:25:59,124 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3850, 4.1227, 4.6183, 2.5333, 4.7918, 4.8966, 3.6099, 3.8704], device='cuda:0'), covar=tensor([0.0608, 0.0223, 0.0214, 0.1029, 0.0080, 0.0196, 0.0373, 0.0321], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0139, 0.0074, 0.0119, 0.0126, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 07:26:11,866 INFO [optim.py:368] (0/8) 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 07:26:25,332 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154627.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:26,743 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:45,762 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154642.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:52,306 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:26:57,946 INFO [train.py:904] (0/8) Epoch 16, batch 2400, loss[loss=0.199, simple_loss=0.2801, pruned_loss=0.05896, over 16768.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2661, pruned_loss=0.04748, over 3317657.86 frames. ], batch size: 102, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:27:50,402 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154689.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:27:55,468 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154693.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:27:56,799 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8799, 4.9668, 4.7550, 4.4725, 4.0110, 5.0190, 5.0263, 4.5734], device='cuda:0'), covar=tensor([0.0890, 0.0551, 0.0500, 0.0470, 0.2061, 0.0457, 0.0384, 0.0848], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0395, 0.0336, 0.0324, 0.0354, 0.0372, 0.0230, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:28:06,455 INFO [train.py:904] (0/8) Epoch 16, batch 2450, loss[loss=0.1693, simple_loss=0.2582, pruned_loss=0.04018, over 16969.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2652, pruned_loss=0.04624, over 3326437.93 frames. ], batch size: 55, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:28:11,560 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154705.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:28:28,921 INFO [optim.py:368] (0/8) 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 07:28:31,827 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:49,590 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:28:59,824 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:16,069 INFO [train.py:904] (0/8) Epoch 16, batch 2500, loss[loss=0.1498, simple_loss=0.2411, pruned_loss=0.02931, over 17170.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2649, pruned_loss=0.04592, over 3335266.96 frames. ], batch size: 46, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:29:18,087 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154753.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 07:29:53,787 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:29:55,069 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=154781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:04,430 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:30:24,144 INFO [train.py:904] (0/8) Epoch 16, batch 2550, loss[loss=0.1846, simple_loss=0.2623, pruned_loss=0.0535, over 16903.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2652, pruned_loss=0.04622, over 3336831.47 frames. ], batch size: 109, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:30:47,017 INFO [optim.py:368] (0/8) 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 07:30:48,641 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3138, 2.1301, 2.3936, 4.1682, 2.1972, 2.5524, 2.2038, 2.3542], device='cuda:0'), covar=tensor([0.1287, 0.4010, 0.2644, 0.0497, 0.3855, 0.2525, 0.3778, 0.3149], device='cuda:0'), in_proj_covar=tensor([0.0385, 0.0422, 0.0352, 0.0324, 0.0425, 0.0486, 0.0391, 0.0494], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:31:02,575 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154830.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:31:32,691 INFO [train.py:904] (0/8) Epoch 16, batch 2600, loss[loss=0.1655, simple_loss=0.2541, pruned_loss=0.03851, over 16723.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2638, pruned_loss=0.04546, over 3338545.71 frames. ], batch size: 83, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:32:07,963 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154877.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:32:36,963 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9223, 3.2490, 2.9733, 5.1251, 4.2304, 4.5476, 1.8058, 3.3423], device='cuda:0'), covar=tensor([0.1225, 0.0633, 0.1021, 0.0160, 0.0210, 0.0352, 0.1458, 0.0669], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0167, 0.0188, 0.0173, 0.0201, 0.0215, 0.0191, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 07:32:40,409 INFO [train.py:904] (0/8) Epoch 16, batch 2650, loss[loss=0.1483, simple_loss=0.2394, pruned_loss=0.02859, over 17034.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2641, pruned_loss=0.04516, over 3341713.95 frames. ], batch size: 41, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:33:01,394 INFO [optim.py:368] (0/8) 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 07:33:12,797 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:35,243 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:43,507 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154947.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:33:49,084 INFO [train.py:904] (0/8) Epoch 16, batch 2700, loss[loss=0.1624, simple_loss=0.2569, pruned_loss=0.03394, over 17134.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2646, pruned_loss=0.04501, over 3340781.15 frames. ], batch size: 47, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:34:21,624 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=154976.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:32,002 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=154984.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:41,399 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:45,814 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=154993.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:34:47,946 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=154995.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:34:56,800 INFO [train.py:904] (0/8) Epoch 16, batch 2750, loss[loss=0.2029, simple_loss=0.2869, pruned_loss=0.05942, over 16434.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2649, pruned_loss=0.0449, over 3333522.11 frames. ], batch size: 75, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:35:11,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9286, 4.6872, 4.9473, 5.1516, 5.3387, 4.6978, 5.3372, 5.3415], device='cuda:0'), covar=tensor([0.1632, 0.1333, 0.1656, 0.0688, 0.0542, 0.0913, 0.0536, 0.0596], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0771, 0.0914, 0.0782, 0.0586, 0.0617, 0.0620, 0.0725], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:35:18,260 INFO [optim.py:368] (0/8) 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 07:35:46,170 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155037.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:35:50,515 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155041.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 07:36:04,762 INFO [train.py:904] (0/8) Epoch 16, batch 2800, loss[loss=0.1718, simple_loss=0.2651, pruned_loss=0.03924, over 17064.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2652, pruned_loss=0.04504, over 3335590.07 frames. ], batch size: 55, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:36:27,729 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 07:36:28,407 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155069.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:38,311 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:36:41,703 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 07:37:15,043 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9929, 3.1507, 3.2758, 2.0237, 2.8640, 2.3298, 3.5293, 3.4549], device='cuda:0'), covar=tensor([0.0230, 0.0816, 0.0565, 0.1724, 0.0786, 0.0914, 0.0487, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0155, 0.0161, 0.0147, 0.0139, 0.0125, 0.0140, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 07:37:15,732 INFO [train.py:904] (0/8) Epoch 16, batch 2850, loss[loss=0.1742, simple_loss=0.2526, pruned_loss=0.04789, over 16780.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2646, pruned_loss=0.04509, over 3337100.61 frames. ], batch size: 124, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:37:18,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9781, 4.2449, 4.3544, 3.3084, 3.6419, 4.3057, 3.8852, 2.7358], device='cuda:0'), covar=tensor([0.0400, 0.0062, 0.0038, 0.0278, 0.0108, 0.0077, 0.0081, 0.0363], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0133, 0.0090, 0.0100, 0.0089, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:37:36,421 INFO [optim.py:368] (0/8) 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 07:37:54,436 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:37:54,516 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:38:15,004 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4519, 4.4985, 4.6687, 4.4431, 4.4774, 5.0970, 4.5999, 4.3298], device='cuda:0'), covar=tensor([0.1655, 0.2282, 0.2606, 0.2564, 0.3272, 0.1342, 0.1728, 0.2866], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0574, 0.0623, 0.0487, 0.0643, 0.0653, 0.0489, 0.0640], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:38:16,333 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7166, 3.7655, 4.0620, 2.3152, 4.1648, 4.1570, 3.3193, 3.1921], device='cuda:0'), covar=tensor([0.0741, 0.0180, 0.0139, 0.1004, 0.0063, 0.0147, 0.0330, 0.0414], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0105, 0.0092, 0.0137, 0.0074, 0.0118, 0.0125, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 07:38:23,953 INFO [train.py:904] (0/8) Epoch 16, batch 2900, loss[loss=0.1898, simple_loss=0.268, pruned_loss=0.05582, over 16475.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2637, pruned_loss=0.04595, over 3328166.57 frames. ], batch size: 68, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:00,499 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155178.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:39:04,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8711, 4.3574, 3.1453, 2.3962, 2.8411, 2.6335, 4.6659, 3.7972], device='cuda:0'), covar=tensor([0.2675, 0.0598, 0.1716, 0.2637, 0.2669, 0.1856, 0.0395, 0.1245], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0265, 0.0293, 0.0293, 0.0286, 0.0239, 0.0280, 0.0320], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:39:33,226 INFO [train.py:904] (0/8) Epoch 16, batch 2950, loss[loss=0.1803, simple_loss=0.26, pruned_loss=0.05026, over 16767.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.263, pruned_loss=0.0467, over 3327017.26 frames. ], batch size: 83, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:39:54,115 INFO [optim.py:368] (0/8) 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 07:40:01,022 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 07:40:12,062 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8652, 4.8342, 4.7092, 3.4884, 4.8247, 1.7948, 4.4208, 4.4701], device='cuda:0'), covar=tensor([0.0144, 0.0118, 0.0257, 0.0723, 0.0129, 0.3252, 0.0211, 0.0378], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0142, 0.0190, 0.0175, 0.0162, 0.0201, 0.0178, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:40:40,772 INFO [train.py:904] (0/8) Epoch 16, batch 3000, loss[loss=0.1943, simple_loss=0.2734, pruned_loss=0.05755, over 16397.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2631, pruned_loss=0.04706, over 3330444.69 frames. ], batch size: 146, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:40:40,773 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 07:40:49,856 INFO [train.py:938] (0/8) Epoch 16, validation: loss=0.1368, simple_loss=0.2429, pruned_loss=0.01541, over 944034.00 frames. 2023-04-30 07:40:49,857 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 07:41:34,128 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:41:59,662 INFO [train.py:904] (0/8) Epoch 16, batch 3050, loss[loss=0.1643, simple_loss=0.2516, pruned_loss=0.03855, over 16040.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2626, pruned_loss=0.04654, over 3328816.76 frames. ], batch size: 35, lr: 4.28e-03, grad_scale: 8.0 2023-04-30 07:42:21,046 INFO [optim.py:368] (0/8) 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 07:42:42,358 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:42,399 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155332.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:42:49,091 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2725, 4.0451, 4.2654, 4.3975, 4.4843, 4.0525, 4.2770, 4.4743], device='cuda:0'), covar=tensor([0.1254, 0.1101, 0.1165, 0.0628, 0.0618, 0.1411, 0.2034, 0.0694], device='cuda:0'), in_proj_covar=tensor([0.0619, 0.0769, 0.0905, 0.0781, 0.0581, 0.0611, 0.0614, 0.0724], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:42:51,584 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1914, 2.2278, 2.5511, 3.1253, 2.9478, 3.4884, 2.1472, 3.4206], device='cuda:0'), covar=tensor([0.0200, 0.0430, 0.0323, 0.0272, 0.0261, 0.0205, 0.0459, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0186, 0.0172, 0.0175, 0.0187, 0.0143, 0.0185, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:43:10,106 INFO [train.py:904] (0/8) Epoch 16, batch 3100, loss[loss=0.1975, simple_loss=0.2742, pruned_loss=0.0604, over 12080.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2625, pruned_loss=0.04693, over 3330004.98 frames. ], batch size: 246, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:43:42,917 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155376.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:17,694 INFO [train.py:904] (0/8) Epoch 16, batch 3150, loss[loss=0.1724, simple_loss=0.2526, pruned_loss=0.04617, over 16772.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2618, pruned_loss=0.04692, over 3322380.89 frames. ], batch size: 89, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:44:22,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2652, 2.4908, 2.0858, 2.1817, 2.8757, 2.5604, 3.0410, 3.0667], device='cuda:0'), covar=tensor([0.0200, 0.0381, 0.0497, 0.0467, 0.0227, 0.0379, 0.0273, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0227, 0.0219, 0.0220, 0.0229, 0.0230, 0.0236, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:44:39,891 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.743e+02 2.373e+02 2.704e+02 3.278e+02 4.749e+02, threshold=5.408e+02, percent-clipped=0.0 2023-04-30 07:44:47,893 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:44:49,096 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:45:27,257 INFO [train.py:904] (0/8) Epoch 16, batch 3200, loss[loss=0.1892, simple_loss=0.2647, pruned_loss=0.05683, over 16694.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.261, pruned_loss=0.04615, over 3328480.98 frames. ], batch size: 134, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:45:33,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155457.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:46:36,142 INFO [train.py:904] (0/8) Epoch 16, batch 3250, loss[loss=0.1738, simple_loss=0.2484, pruned_loss=0.04964, over 16902.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2609, pruned_loss=0.04605, over 3325479.72 frames. ], batch size: 109, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:46:52,176 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2199, 3.4344, 3.3619, 2.1897, 2.9864, 2.4346, 3.6743, 3.7353], device='cuda:0'), covar=tensor([0.0213, 0.0731, 0.0585, 0.1734, 0.0771, 0.0908, 0.0521, 0.0779], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0156, 0.0161, 0.0148, 0.0139, 0.0126, 0.0141, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 07:46:58,464 INFO [optim.py:368] (0/8) 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 07:46:58,920 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:47:34,929 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 07:47:36,954 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-04-30 07:47:45,924 INFO [train.py:904] (0/8) Epoch 16, batch 3300, loss[loss=0.1706, simple_loss=0.268, pruned_loss=0.0366, over 17123.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2626, pruned_loss=0.04691, over 3322548.32 frames. ], batch size: 48, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:48:34,848 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7696, 1.3265, 1.5992, 1.5550, 1.7012, 1.9855, 1.5451, 1.7533], device='cuda:0'), covar=tensor([0.0245, 0.0364, 0.0204, 0.0266, 0.0262, 0.0163, 0.0390, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0186, 0.0172, 0.0176, 0.0187, 0.0143, 0.0185, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:48:56,953 INFO [train.py:904] (0/8) Epoch 16, batch 3350, loss[loss=0.1379, simple_loss=0.2303, pruned_loss=0.02277, over 17229.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.263, pruned_loss=0.04663, over 3319799.69 frames. ], batch size: 44, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:49:01,205 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:49:19,486 INFO [optim.py:368] (0/8) 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 07:49:39,752 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:08,452 INFO [train.py:904] (0/8) Epoch 16, batch 3400, loss[loss=0.1845, simple_loss=0.2784, pruned_loss=0.04534, over 16766.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2627, pruned_loss=0.0461, over 3311356.51 frames. ], batch size: 57, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:50:16,616 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155657.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:27,994 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:44,108 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-04-30 07:50:48,698 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:50:56,859 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5921, 4.5341, 4.5340, 3.9839, 4.5237, 1.8191, 4.2469, 4.2694], device='cuda:0'), covar=tensor([0.0113, 0.0088, 0.0146, 0.0314, 0.0086, 0.2604, 0.0148, 0.0189], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0144, 0.0192, 0.0177, 0.0165, 0.0203, 0.0181, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:51:03,856 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6976, 5.0389, 5.4285, 5.3608, 5.4016, 5.0503, 4.6646, 4.7737], device='cuda:0'), covar=tensor([0.0608, 0.0698, 0.0491, 0.0646, 0.0673, 0.0515, 0.1381, 0.0579], device='cuda:0'), in_proj_covar=tensor([0.0395, 0.0424, 0.0412, 0.0392, 0.0466, 0.0441, 0.0536, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 07:51:19,270 INFO [train.py:904] (0/8) Epoch 16, batch 3450, loss[loss=0.1853, simple_loss=0.2753, pruned_loss=0.04771, over 16673.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2612, pruned_loss=0.04557, over 3319055.63 frames. ], batch size: 62, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:51:32,369 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9853, 4.8725, 4.8632, 4.5179, 4.5200, 4.9533, 4.8008, 4.6957], device='cuda:0'), covar=tensor([0.0676, 0.0774, 0.0332, 0.0342, 0.1039, 0.0534, 0.0398, 0.0712], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0403, 0.0341, 0.0331, 0.0359, 0.0382, 0.0234, 0.0412], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:51:41,377 INFO [optim.py:368] (0/8) 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 07:51:41,779 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155718.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:51:52,652 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=155725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:52:28,786 INFO [train.py:904] (0/8) Epoch 16, batch 3500, loss[loss=0.1756, simple_loss=0.2698, pruned_loss=0.04074, over 16680.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2608, pruned_loss=0.04529, over 3325713.06 frames. ], batch size: 62, lr: 4.27e-03, grad_scale: 16.0 2023-04-30 07:52:58,255 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=155773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:14,268 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=155784.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:53:40,577 INFO [train.py:904] (0/8) Epoch 16, batch 3550, loss[loss=0.1836, simple_loss=0.2607, pruned_loss=0.05321, over 16444.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.26, pruned_loss=0.04475, over 3320389.17 frames. ], batch size: 146, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:53:48,272 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0574, 4.7738, 5.0163, 5.2086, 5.4422, 4.7010, 5.4060, 5.4308], device='cuda:0'), covar=tensor([0.1616, 0.1308, 0.1688, 0.0756, 0.0465, 0.0972, 0.0488, 0.0556], device='cuda:0'), in_proj_covar=tensor([0.0632, 0.0783, 0.0924, 0.0797, 0.0593, 0.0623, 0.0629, 0.0738], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 07:53:56,152 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:54:04,795 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.310e+02 2.739e+02 3.224e+02 5.856e+02, threshold=5.477e+02, percent-clipped=2.0 2023-04-30 07:54:42,495 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=155845.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 07:54:52,023 INFO [train.py:904] (0/8) Epoch 16, batch 3600, loss[loss=0.2072, simple_loss=0.2743, pruned_loss=0.07009, over 11702.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2591, pruned_loss=0.04514, over 3300794.39 frames. ], batch size: 247, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:03,885 INFO [train.py:904] (0/8) Epoch 16, batch 3650, loss[loss=0.178, simple_loss=0.2464, pruned_loss=0.05485, over 16854.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2582, pruned_loss=0.04563, over 3311004.47 frames. ], batch size: 96, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:56:28,569 INFO [optim.py:368] (0/8) 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 07:57:09,359 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7192, 4.7731, 4.9507, 4.8005, 4.8405, 5.3992, 4.9611, 4.6767], device='cuda:0'), covar=tensor([0.1430, 0.2075, 0.1871, 0.2134, 0.2488, 0.1014, 0.1487, 0.2600], device='cuda:0'), in_proj_covar=tensor([0.0400, 0.0572, 0.0614, 0.0484, 0.0642, 0.0648, 0.0488, 0.0640], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 07:57:17,739 INFO [train.py:904] (0/8) Epoch 16, batch 3700, loss[loss=0.1781, simple_loss=0.2459, pruned_loss=0.05515, over 16768.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2575, pruned_loss=0.04737, over 3280541.84 frames. ], batch size: 124, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:57:32,042 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=155961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:57:38,398 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7498, 3.5901, 4.1313, 1.8819, 4.4218, 4.5193, 3.1252, 3.1804], device='cuda:0'), covar=tensor([0.0849, 0.0310, 0.0285, 0.1347, 0.0067, 0.0118, 0.0469, 0.0463], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0138, 0.0075, 0.0120, 0.0126, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 07:58:30,550 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-156000.pt 2023-04-30 07:58:36,932 INFO [train.py:904] (0/8) Epoch 16, batch 3750, loss[loss=0.2132, simple_loss=0.2908, pruned_loss=0.06778, over 16519.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.258, pruned_loss=0.04878, over 3282793.79 frames. ], batch size: 62, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 07:58:53,078 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 07:59:02,687 INFO [optim.py:368] (0/8) 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 07:59:51,249 INFO [train.py:904] (0/8) Epoch 16, batch 3800, loss[loss=0.2036, simple_loss=0.2858, pruned_loss=0.06064, over 16270.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2589, pruned_loss=0.04989, over 3288494.53 frames. ], batch size: 165, lr: 4.27e-03, grad_scale: 8.0 2023-04-30 08:00:17,813 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156070.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:03,109 INFO [train.py:904] (0/8) Epoch 16, batch 3850, loss[loss=0.1803, simple_loss=0.2539, pruned_loss=0.05339, over 16802.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2592, pruned_loss=0.05046, over 3283649.75 frames. ], batch size: 102, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:01:19,553 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156113.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:01:26,260 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8718, 4.7147, 4.8748, 5.0693, 5.2593, 4.6707, 5.1619, 5.2571], device='cuda:0'), covar=tensor([0.1565, 0.1114, 0.1576, 0.0683, 0.0458, 0.0790, 0.0635, 0.0558], device='cuda:0'), in_proj_covar=tensor([0.0611, 0.0756, 0.0893, 0.0767, 0.0572, 0.0602, 0.0606, 0.0712], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:01:28,707 INFO [optim.py:368] (0/8) 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 08:01:42,425 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 08:01:47,411 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156131.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:01:59,130 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156140.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 08:02:16,232 INFO [train.py:904] (0/8) Epoch 16, batch 3900, loss[loss=0.1948, simple_loss=0.2643, pruned_loss=0.06265, over 16784.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2579, pruned_loss=0.05051, over 3284566.05 frames. ], batch size: 124, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:02:28,770 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:26,088 INFO [train.py:904] (0/8) Epoch 16, batch 3950, loss[loss=0.1807, simple_loss=0.2485, pruned_loss=0.05642, over 16793.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.257, pruned_loss=0.05046, over 3292920.11 frames. ], batch size: 102, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:03:47,399 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:03:50,341 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.580e+02 2.349e+02 2.934e+02 3.426e+02 7.028e+02, threshold=5.868e+02, percent-clipped=1.0 2023-04-30 08:04:37,102 INFO [train.py:904] (0/8) Epoch 16, batch 4000, loss[loss=0.1543, simple_loss=0.2379, pruned_loss=0.0354, over 16908.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2568, pruned_loss=0.05103, over 3291989.75 frames. ], batch size: 90, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:04:49,149 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156261.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:12,192 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:22,656 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7249, 3.2307, 3.2073, 2.0869, 3.0030, 3.2705, 3.0910, 1.8891], device='cuda:0'), covar=tensor([0.0509, 0.0056, 0.0047, 0.0362, 0.0081, 0.0085, 0.0077, 0.0399], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0076, 0.0076, 0.0131, 0.0089, 0.0099, 0.0088, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 08:05:31,082 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:05:36,902 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6347, 3.7335, 2.2496, 4.1385, 2.8548, 4.2206, 2.3393, 2.9625], device='cuda:0'), covar=tensor([0.0249, 0.0327, 0.1582, 0.0129, 0.0704, 0.0286, 0.1464, 0.0703], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0177, 0.0194, 0.0155, 0.0174, 0.0217, 0.0202, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:05:49,241 INFO [train.py:904] (0/8) Epoch 16, batch 4050, loss[loss=0.1679, simple_loss=0.2583, pruned_loss=0.03874, over 16848.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2569, pruned_loss=0.04951, over 3295879.27 frames. ], batch size: 96, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:05:55,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4491, 3.4865, 2.1563, 3.8088, 2.6546, 3.8819, 2.1599, 2.7682], device='cuda:0'), covar=tensor([0.0250, 0.0333, 0.1530, 0.0132, 0.0755, 0.0346, 0.1521, 0.0743], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0176, 0.0193, 0.0155, 0.0173, 0.0216, 0.0202, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:05:59,642 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:04,378 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:06:14,233 INFO [optim.py:368] (0/8) 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 08:06:44,890 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0098, 5.2974, 5.0270, 5.0998, 4.7680, 4.6851, 4.7096, 5.4132], device='cuda:0'), covar=tensor([0.1124, 0.0782, 0.0893, 0.0726, 0.0724, 0.0845, 0.1021, 0.0792], device='cuda:0'), in_proj_covar=tensor([0.0633, 0.0785, 0.0640, 0.0568, 0.0492, 0.0504, 0.0654, 0.0602], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:07:01,319 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:01,987 INFO [train.py:904] (0/8) Epoch 16, batch 4100, loss[loss=0.1877, simple_loss=0.2727, pruned_loss=0.05134, over 16923.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2589, pruned_loss=0.04909, over 3286245.81 frames. ], batch size: 116, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:07:15,154 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:07:31,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7615, 2.4580, 2.2990, 3.2880, 2.3921, 3.4889, 1.4573, 2.6693], device='cuda:0'), covar=tensor([0.1327, 0.0751, 0.1233, 0.0185, 0.0196, 0.0407, 0.1692, 0.0829], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0169, 0.0188, 0.0174, 0.0204, 0.0214, 0.0191, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:08:15,975 INFO [train.py:904] (0/8) Epoch 16, batch 4150, loss[loss=0.279, simple_loss=0.3437, pruned_loss=0.1071, over 11329.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2663, pruned_loss=0.05176, over 3252099.14 frames. ], batch size: 248, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:08:21,468 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 08:08:40,350 INFO [optim.py:368] (0/8) 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 08:08:52,141 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156426.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:09:13,285 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156440.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:09:30,681 INFO [train.py:904] (0/8) Epoch 16, batch 4200, loss[loss=0.2045, simple_loss=0.2961, pruned_loss=0.05644, over 16898.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2736, pruned_loss=0.05369, over 3222923.87 frames. ], batch size: 109, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:10:24,401 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:10:27,413 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 08:10:27,601 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 08:10:43,947 INFO [train.py:904] (0/8) Epoch 16, batch 4250, loss[loss=0.1998, simple_loss=0.2808, pruned_loss=0.05935, over 12262.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2764, pruned_loss=0.05376, over 3187777.67 frames. ], batch size: 246, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:11:09,164 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.668e+02 2.212e+02 2.674e+02 3.089e+02 6.979e+02, threshold=5.347e+02, percent-clipped=2.0 2023-04-30 08:11:35,501 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3538, 3.4378, 1.6648, 3.7645, 2.4741, 3.7636, 1.9724, 2.5961], device='cuda:0'), covar=tensor([0.0265, 0.0344, 0.1989, 0.0153, 0.0837, 0.0439, 0.1656, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0173, 0.0191, 0.0150, 0.0173, 0.0213, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:11:56,446 INFO [train.py:904] (0/8) Epoch 16, batch 4300, loss[loss=0.2021, simple_loss=0.2908, pruned_loss=0.0567, over 16757.00 frames. ], tot_loss[loss=0.1919, simple_loss=0.2778, pruned_loss=0.05303, over 3173521.89 frames. ], batch size: 124, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:12:16,030 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0445, 2.7693, 2.7258, 1.9471, 2.5970, 1.9972, 2.7818, 2.8636], device='cuda:0'), covar=tensor([0.0264, 0.0669, 0.0571, 0.1732, 0.0862, 0.0930, 0.0495, 0.0553], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0156, 0.0161, 0.0148, 0.0139, 0.0126, 0.0140, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:12:19,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7773, 2.7283, 2.6386, 1.8804, 2.5631, 2.7063, 2.6182, 1.8802], device='cuda:0'), covar=tensor([0.0418, 0.0065, 0.0064, 0.0358, 0.0107, 0.0107, 0.0100, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0076, 0.0077, 0.0131, 0.0090, 0.0099, 0.0088, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 08:12:27,176 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:13:09,410 INFO [train.py:904] (0/8) Epoch 16, batch 4350, loss[loss=0.2161, simple_loss=0.2983, pruned_loss=0.06692, over 17248.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2809, pruned_loss=0.05396, over 3174831.11 frames. ], batch size: 52, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:13:34,597 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.777e+02 2.270e+02 2.623e+02 3.147e+02 7.539e+02, threshold=5.245e+02, percent-clipped=2.0 2023-04-30 08:14:14,568 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=156646.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:14:22,117 INFO [train.py:904] (0/8) Epoch 16, batch 4400, loss[loss=0.2117, simple_loss=0.2998, pruned_loss=0.06182, over 17103.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2835, pruned_loss=0.0552, over 3174341.02 frames. ], batch size: 47, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:06,779 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8963, 1.9166, 2.4882, 2.8876, 2.7252, 3.2843, 1.9692, 3.1937], device='cuda:0'), covar=tensor([0.0160, 0.0412, 0.0247, 0.0224, 0.0236, 0.0120, 0.0448, 0.0110], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0174, 0.0185, 0.0140, 0.0183, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:15:32,110 INFO [train.py:904] (0/8) Epoch 16, batch 4450, loss[loss=0.1972, simple_loss=0.2965, pruned_loss=0.04896, over 16792.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2864, pruned_loss=0.05628, over 3192881.13 frames. ], batch size: 102, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:15:57,565 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.385e+02 2.034e+02 2.328e+02 2.907e+02 4.699e+02, threshold=4.656e+02, percent-clipped=0.0 2023-04-30 08:16:04,250 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7222, 3.7162, 2.2843, 4.3770, 2.8582, 4.3212, 2.4748, 2.9869], device='cuda:0'), covar=tensor([0.0252, 0.0362, 0.1623, 0.0134, 0.0835, 0.0389, 0.1398, 0.0746], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0190, 0.0149, 0.0172, 0.0212, 0.0199, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:16:09,716 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156726.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:16:41,423 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:16:45,643 INFO [train.py:904] (0/8) Epoch 16, batch 4500, loss[loss=0.2032, simple_loss=0.2934, pruned_loss=0.05652, over 16718.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2872, pruned_loss=0.05678, over 3198723.54 frames. ], batch size: 89, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:16:47,449 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0298, 2.3296, 1.9348, 2.1251, 2.7215, 2.3139, 2.8100, 2.8916], device='cuda:0'), covar=tensor([0.0105, 0.0331, 0.0457, 0.0402, 0.0193, 0.0352, 0.0162, 0.0213], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0217, 0.0210, 0.0211, 0.0220, 0.0221, 0.0224, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:16:49,040 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-04-30 08:17:16,167 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:17:19,696 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 08:17:54,730 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6504, 2.5363, 2.4634, 3.5281, 2.6955, 3.7301, 1.5246, 2.8236], device='cuda:0'), covar=tensor([0.1298, 0.0714, 0.1083, 0.0137, 0.0199, 0.0332, 0.1571, 0.0744], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0168, 0.0187, 0.0172, 0.0203, 0.0213, 0.0190, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:17:56,547 INFO [train.py:904] (0/8) Epoch 16, batch 4550, loss[loss=0.2161, simple_loss=0.3011, pruned_loss=0.06553, over 16552.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2881, pruned_loss=0.05753, over 3214492.48 frames. ], batch size: 62, lr: 4.26e-03, grad_scale: 8.0 2023-04-30 08:18:07,300 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156810.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:18:16,491 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4749, 3.6031, 2.6596, 2.1537, 2.3580, 2.1658, 3.8300, 3.2693], device='cuda:0'), covar=tensor([0.2993, 0.0642, 0.1776, 0.2548, 0.2573, 0.2113, 0.0479, 0.1177], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0264, 0.0295, 0.0295, 0.0291, 0.0240, 0.0282, 0.0320], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 08:18:19,417 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.005e+02 2.274e+02 2.770e+02 4.941e+02, threshold=4.548e+02, percent-clipped=1.0 2023-04-30 08:19:06,335 INFO [train.py:904] (0/8) Epoch 16, batch 4600, loss[loss=0.2155, simple_loss=0.2934, pruned_loss=0.06883, over 11818.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2891, pruned_loss=0.05778, over 3226102.53 frames. ], batch size: 246, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:19:36,314 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:19:55,725 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=156885.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:20:19,290 INFO [train.py:904] (0/8) Epoch 16, batch 4650, loss[loss=0.1867, simple_loss=0.282, pruned_loss=0.04572, over 15413.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2878, pruned_loss=0.05767, over 3214482.22 frames. ], batch size: 191, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:20:37,610 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-04-30 08:20:39,060 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 08:20:45,040 INFO [optim.py:368] (0/8) 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 08:20:46,970 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=156920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:20:48,343 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6837, 3.6593, 4.0733, 1.9461, 4.3109, 4.3255, 3.0779, 3.1228], device='cuda:0'), covar=tensor([0.0836, 0.0251, 0.0231, 0.1266, 0.0067, 0.0102, 0.0472, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0106, 0.0093, 0.0139, 0.0074, 0.0119, 0.0126, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 08:21:12,296 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-04-30 08:21:25,723 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:25,812 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=156946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:21:33,081 INFO [train.py:904] (0/8) Epoch 16, batch 4700, loss[loss=0.1749, simple_loss=0.2618, pruned_loss=0.04403, over 16507.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.285, pruned_loss=0.05668, over 3207547.70 frames. ], batch size: 146, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:36,261 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 16, batch 4750, loss[loss=0.162, simple_loss=0.252, pruned_loss=0.03598, over 16403.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.2812, pruned_loss=0.05445, over 3205675.53 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:22:59,380 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157010.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:23:11,421 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 1.885e+02 2.255e+02 2.719e+02 5.243e+02, threshold=4.511e+02, percent-clipped=2.0 2023-04-30 08:23:59,608 INFO [train.py:904] (0/8) Epoch 16, batch 4800, loss[loss=0.186, simple_loss=0.2703, pruned_loss=0.05079, over 16478.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2774, pruned_loss=0.05221, over 3207859.37 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:24:28,453 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 16, batch 4850, loss[loss=0.1987, simple_loss=0.2934, pruned_loss=0.05198, over 15371.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2772, pruned_loss=0.05108, over 3201525.92 frames. ], batch size: 190, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:25:20,106 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:25:36,205 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 08:25:41,924 INFO [optim.py:368] (0/8) 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 08:26:16,198 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:26:30,916 INFO [train.py:904] (0/8) Epoch 16, batch 4900, loss[loss=0.1697, simple_loss=0.2632, pruned_loss=0.03811, over 16877.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2767, pruned_loss=0.04988, over 3194590.24 frames. ], batch size: 96, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:45,027 INFO [train.py:904] (0/8) Epoch 16, batch 4950, loss[loss=0.1916, simple_loss=0.2817, pruned_loss=0.05076, over 16653.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2765, pruned_loss=0.04942, over 3188171.90 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:27:46,713 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157203.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:27:59,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5313, 4.5729, 4.3736, 4.1050, 4.0445, 4.4775, 4.2529, 4.1734], device='cuda:0'), covar=tensor([0.0525, 0.0471, 0.0280, 0.0268, 0.0967, 0.0467, 0.0536, 0.0637], device='cuda:0'), in_proj_covar=tensor([0.0264, 0.0367, 0.0313, 0.0302, 0.0329, 0.0349, 0.0214, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:28:08,813 INFO [optim.py:368] (0/8) 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 08:28:40,300 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157241.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:28:57,717 INFO [train.py:904] (0/8) Epoch 16, batch 5000, loss[loss=0.1808, simple_loss=0.2688, pruned_loss=0.04636, over 16453.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.278, pruned_loss=0.04937, over 3209344.74 frames. ], batch size: 68, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:10,088 INFO [train.py:904] (0/8) Epoch 16, batch 5050, loss[loss=0.2198, simple_loss=0.3043, pruned_loss=0.06763, over 12006.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2785, pruned_loss=0.04939, over 3209998.63 frames. ], batch size: 246, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:30:33,955 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.188e+02 2.518e+02 3.068e+02 6.843e+02, threshold=5.037e+02, percent-clipped=3.0 2023-04-30 08:31:22,547 INFO [train.py:904] (0/8) Epoch 16, batch 5100, loss[loss=0.1513, simple_loss=0.2419, pruned_loss=0.03035, over 16756.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2776, pruned_loss=0.04918, over 3198489.79 frames. ], batch size: 57, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:31:28,565 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157356.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:31:40,598 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1482, 5.0735, 5.0677, 4.2581, 5.0737, 1.9016, 4.7577, 4.9155], device='cuda:0'), covar=tensor([0.0076, 0.0075, 0.0123, 0.0500, 0.0083, 0.2497, 0.0111, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0136, 0.0182, 0.0170, 0.0155, 0.0194, 0.0170, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:31:44,283 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:07,474 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:32:38,192 INFO [train.py:904] (0/8) Epoch 16, batch 5150, loss[loss=0.1813, simple_loss=0.2651, pruned_loss=0.04875, over 16409.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2775, pruned_loss=0.04864, over 3193905.88 frames. ], batch size: 35, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:32:42,189 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157404.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:32:43,382 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157405.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:02,223 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157417.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:04,664 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.981e+02 2.319e+02 2.667e+02 4.266e+02, threshold=4.638e+02, percent-clipped=0.0 2023-04-30 08:33:37,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7520, 3.8492, 2.8707, 2.3100, 2.6320, 2.4278, 3.9919, 3.4792], device='cuda:0'), covar=tensor([0.2410, 0.0584, 0.1716, 0.2270, 0.2275, 0.1758, 0.0477, 0.1088], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0261, 0.0292, 0.0293, 0.0286, 0.0236, 0.0279, 0.0315], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 08:33:40,847 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157443.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:42,518 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-04-30 08:33:53,097 INFO [train.py:904] (0/8) Epoch 16, batch 5200, loss[loss=0.2186, simple_loss=0.2983, pruned_loss=0.06941, over 16257.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2763, pruned_loss=0.04817, over 3193537.17 frames. ], batch size: 165, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:33:54,779 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:33:56,477 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 08:34:12,719 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157465.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 08:35:01,953 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:35:08,190 INFO [train.py:904] (0/8) Epoch 16, batch 5250, loss[loss=0.1764, simple_loss=0.2689, pruned_loss=0.04197, over 16225.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2734, pruned_loss=0.04756, over 3203911.20 frames. ], batch size: 165, lr: 4.25e-03, grad_scale: 8.0 2023-04-30 08:35:13,320 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157506.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:35:32,959 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.377e+02 2.109e+02 2.306e+02 2.689e+02 3.834e+02, threshold=4.611e+02, percent-clipped=0.0 2023-04-30 08:36:05,784 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157541.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:36:21,522 INFO [train.py:904] (0/8) Epoch 16, batch 5300, loss[loss=0.1773, simple_loss=0.2624, pruned_loss=0.04612, over 16540.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2703, pruned_loss=0.04638, over 3218055.69 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:36:45,061 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157567.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:36:46,110 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:36:46,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 08:36:49,791 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3634, 2.2414, 2.3299, 4.1500, 2.1834, 2.6591, 2.3777, 2.4464], device='cuda:0'), covar=tensor([0.1116, 0.3437, 0.2640, 0.0411, 0.3722, 0.2328, 0.3354, 0.2887], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0421, 0.0349, 0.0319, 0.0423, 0.0485, 0.0388, 0.0490], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:37:13,788 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7206, 3.0039, 3.1429, 1.7723, 2.8278, 2.1184, 3.2724, 3.2487], device='cuda:0'), covar=tensor([0.0243, 0.0724, 0.0626, 0.2037, 0.0826, 0.1019, 0.0567, 0.0821], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0148, 0.0139, 0.0125, 0.0139, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:37:17,189 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:37:35,571 INFO [train.py:904] (0/8) Epoch 16, batch 5350, loss[loss=0.1859, simple_loss=0.2768, pruned_loss=0.04756, over 17140.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2687, pruned_loss=0.04576, over 3211486.82 frames. ], batch size: 47, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:00,212 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.392e+02 1.990e+02 2.338e+02 2.844e+02 4.721e+02, threshold=4.676e+02, percent-clipped=1.0 2023-04-30 08:38:13,590 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2739, 5.5854, 5.2969, 5.3014, 5.0298, 4.9396, 4.9386, 5.6474], device='cuda:0'), covar=tensor([0.1057, 0.0704, 0.0792, 0.0748, 0.0727, 0.0692, 0.0968, 0.0792], device='cuda:0'), in_proj_covar=tensor([0.0612, 0.0759, 0.0620, 0.0550, 0.0477, 0.0482, 0.0627, 0.0581], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:38:16,623 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:38:49,049 INFO [train.py:904] (0/8) Epoch 16, batch 5400, loss[loss=0.2203, simple_loss=0.3026, pruned_loss=0.06903, over 12122.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.272, pruned_loss=0.04672, over 3206080.75 frames. ], batch size: 246, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:38:52,822 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0102, 2.0225, 2.1801, 3.4610, 1.9849, 2.3021, 2.1467, 2.1562], device='cuda:0'), covar=tensor([0.1233, 0.3397, 0.2548, 0.0568, 0.3974, 0.2497, 0.3434, 0.3187], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0421, 0.0350, 0.0320, 0.0424, 0.0486, 0.0389, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:39:09,641 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157666.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:04,948 INFO [train.py:904] (0/8) Epoch 16, batch 5450, loss[loss=0.1919, simple_loss=0.2823, pruned_loss=0.05071, over 16864.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2746, pruned_loss=0.04806, over 3196454.13 frames. ], batch size: 116, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:40:08,943 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1722, 2.1038, 2.2358, 3.7532, 2.0139, 2.4957, 2.2400, 2.2766], device='cuda:0'), covar=tensor([0.1180, 0.3381, 0.2577, 0.0521, 0.3947, 0.2326, 0.3178, 0.3273], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0421, 0.0350, 0.0320, 0.0423, 0.0485, 0.0388, 0.0490], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:40:19,880 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157712.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:23,495 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157714.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:40:29,847 INFO [optim.py:368] (0/8) 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 08:41:00,209 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:41:16,514 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.51 vs. limit=5.0 2023-04-30 08:41:20,985 INFO [train.py:904] (0/8) Epoch 16, batch 5500, loss[loss=0.2112, simple_loss=0.3018, pruned_loss=0.06026, over 16291.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2826, pruned_loss=0.05317, over 3167972.59 frames. ], batch size: 165, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:41:33,659 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157760.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:42:24,727 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4019, 4.6901, 4.4672, 4.4700, 4.2491, 4.1888, 4.2453, 4.7241], device='cuda:0'), covar=tensor([0.1099, 0.0846, 0.0951, 0.0818, 0.0766, 0.1407, 0.0974, 0.0894], device='cuda:0'), in_proj_covar=tensor([0.0607, 0.0754, 0.0616, 0.0546, 0.0473, 0.0480, 0.0622, 0.0575], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:42:30,918 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=157798.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:42:36,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5099, 3.0645, 2.9583, 1.9486, 2.7322, 2.1537, 3.0734, 3.2583], device='cuda:0'), covar=tensor([0.0281, 0.0659, 0.0578, 0.1805, 0.0784, 0.0966, 0.0604, 0.0760], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0155, 0.0162, 0.0148, 0.0139, 0.0126, 0.0140, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:42:36,601 INFO [train.py:904] (0/8) Epoch 16, batch 5550, loss[loss=0.23, simple_loss=0.312, pruned_loss=0.07398, over 15417.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2895, pruned_loss=0.05759, over 3180374.76 frames. ], batch size: 190, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:04,450 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.012e+02 3.202e+02 3.935e+02 4.988e+02 8.755e+02, threshold=7.870e+02, percent-clipped=18.0 2023-04-30 08:43:47,879 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=157846.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:43:51,460 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8756, 2.6438, 2.5602, 1.9493, 2.4588, 2.6459, 2.5402, 1.9087], device='cuda:0'), covar=tensor([0.0397, 0.0071, 0.0078, 0.0344, 0.0115, 0.0117, 0.0100, 0.0338], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0073, 0.0075, 0.0128, 0.0088, 0.0098, 0.0086, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 08:43:57,057 INFO [train.py:904] (0/8) Epoch 16, batch 5600, loss[loss=0.2507, simple_loss=0.327, pruned_loss=0.08722, over 15487.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2949, pruned_loss=0.06257, over 3123043.58 frames. ], batch size: 191, lr: 4.24e-03, grad_scale: 16.0 2023-04-30 08:43:59,532 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=157853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:44:01,307 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7103, 1.7526, 2.3188, 2.6666, 2.6174, 3.0088, 1.9289, 2.9631], device='cuda:0'), covar=tensor([0.0164, 0.0456, 0.0274, 0.0220, 0.0268, 0.0165, 0.0429, 0.0119], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0185, 0.0171, 0.0174, 0.0185, 0.0141, 0.0184, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:44:13,642 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157862.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:45:21,491 INFO [train.py:904] (0/8) Epoch 16, batch 5650, loss[loss=0.2529, simple_loss=0.3358, pruned_loss=0.08499, over 16573.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2996, pruned_loss=0.06598, over 3111667.01 frames. ], batch size: 75, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:45:40,966 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=157914.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:45:50,675 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.188e+02 3.549e+02 4.142e+02 4.836e+02 1.033e+03, threshold=8.283e+02, percent-clipped=2.0 2023-04-30 08:45:53,639 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4629, 3.5318, 3.3000, 2.9072, 3.1422, 3.4420, 3.2583, 3.1782], device='cuda:0'), covar=tensor([0.0573, 0.0491, 0.0250, 0.0234, 0.0469, 0.0378, 0.1131, 0.0424], device='cuda:0'), in_proj_covar=tensor([0.0269, 0.0375, 0.0316, 0.0306, 0.0331, 0.0355, 0.0217, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:45:56,956 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=157924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:46:08,851 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 08:46:39,308 INFO [train.py:904] (0/8) Epoch 16, batch 5700, loss[loss=0.2777, simple_loss=0.3364, pruned_loss=0.1095, over 11275.00 frames. ], tot_loss[loss=0.2183, simple_loss=0.301, pruned_loss=0.06777, over 3082921.64 frames. ], batch size: 248, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:47:08,805 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 08:47:11,038 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 08:47:44,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8916, 4.1338, 3.9561, 3.9952, 3.6799, 3.7301, 3.8270, 4.1412], device='cuda:0'), covar=tensor([0.1061, 0.0829, 0.1032, 0.0787, 0.0762, 0.1663, 0.0902, 0.0894], device='cuda:0'), in_proj_covar=tensor([0.0599, 0.0742, 0.0608, 0.0540, 0.0465, 0.0474, 0.0615, 0.0565], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:47:56,009 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-158000.pt 2023-04-30 08:48:01,942 INFO [train.py:904] (0/8) Epoch 16, batch 5750, loss[loss=0.1983, simple_loss=0.292, pruned_loss=0.05226, over 16873.00 frames. ], tot_loss[loss=0.2227, simple_loss=0.304, pruned_loss=0.07063, over 3021018.84 frames. ], batch size: 96, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:48:16,709 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158012.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:48:30,524 INFO [optim.py:368] (0/8) 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 08:48:42,754 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5655, 3.8709, 2.9762, 2.2012, 2.6958, 2.4794, 4.0634, 3.5274], device='cuda:0'), covar=tensor([0.2931, 0.0601, 0.1716, 0.2925, 0.2221, 0.1850, 0.0507, 0.1127], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0261, 0.0294, 0.0294, 0.0286, 0.0238, 0.0279, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 08:48:52,821 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3099, 4.1415, 4.3666, 4.5240, 4.6603, 4.2724, 4.5946, 4.6631], device='cuda:0'), covar=tensor([0.1927, 0.1281, 0.1586, 0.0742, 0.0635, 0.1100, 0.0741, 0.0639], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0723, 0.0851, 0.0727, 0.0549, 0.0578, 0.0581, 0.0679], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:49:00,624 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:22,786 INFO [train.py:904] (0/8) Epoch 16, batch 5800, loss[loss=0.231, simple_loss=0.3104, pruned_loss=0.07584, over 15611.00 frames. ], tot_loss[loss=0.2205, simple_loss=0.303, pruned_loss=0.06894, over 3036267.82 frames. ], batch size: 192, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:49:35,955 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:49:36,009 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158060.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:50:16,573 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158086.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:50:28,400 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6695, 2.8170, 2.5632, 4.3703, 3.2327, 4.1158, 1.5900, 2.9254], device='cuda:0'), covar=tensor([0.1339, 0.0714, 0.1173, 0.0143, 0.0274, 0.0366, 0.1572, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0168, 0.0189, 0.0171, 0.0202, 0.0213, 0.0192, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:50:41,310 INFO [train.py:904] (0/8) Epoch 16, batch 5850, loss[loss=0.2255, simple_loss=0.3121, pruned_loss=0.06947, over 16594.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.3003, pruned_loss=0.0666, over 3058988.95 frames. ], batch size: 76, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:50:51,622 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158108.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:51:08,531 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.851e+02 3.457e+02 4.318e+02 8.019e+02, threshold=6.915e+02, percent-clipped=1.0 2023-04-30 08:51:13,000 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:52:03,756 INFO [train.py:904] (0/8) Epoch 16, batch 5900, loss[loss=0.1813, simple_loss=0.2713, pruned_loss=0.04566, over 16843.00 frames. ], tot_loss[loss=0.2165, simple_loss=0.2999, pruned_loss=0.06651, over 3068963.41 frames. ], batch size: 116, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:52:23,707 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158162.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 08:52:37,578 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 08:52:56,019 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:25,270 INFO [train.py:904] (0/8) Epoch 16, batch 5950, loss[loss=0.2036, simple_loss=0.294, pruned_loss=0.05664, over 16556.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.3004, pruned_loss=0.06502, over 3080837.02 frames. ], batch size: 68, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:53:36,986 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:53:38,265 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158210.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 08:53:55,270 INFO [optim.py:368] (0/8) 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 08:54:01,919 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:54:45,450 INFO [train.py:904] (0/8) Epoch 16, batch 6000, loss[loss=0.2021, simple_loss=0.29, pruned_loss=0.05712, over 16732.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2992, pruned_loss=0.06451, over 3088566.88 frames. ], batch size: 83, lr: 4.24e-03, grad_scale: 8.0 2023-04-30 08:54:45,451 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 08:54:56,482 INFO [train.py:938] (0/8) Epoch 16, validation: loss=0.1553, simple_loss=0.2682, pruned_loss=0.0212, over 944034.00 frames. 2023-04-30 08:54:56,483 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 08:54:57,318 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 08:55:27,052 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:56:13,224 INFO [train.py:904] (0/8) Epoch 16, batch 6050, loss[loss=0.2927, simple_loss=0.3461, pruned_loss=0.1196, over 11618.00 frames. ], tot_loss[loss=0.2133, simple_loss=0.2979, pruned_loss=0.0643, over 3091844.64 frames. ], batch size: 246, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:56:40,250 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.935e+02 2.678e+02 3.324e+02 4.253e+02 8.310e+02, threshold=6.647e+02, percent-clipped=4.0 2023-04-30 08:56:57,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8908, 4.1799, 3.9697, 4.0391, 3.7195, 3.8095, 3.8289, 4.1630], device='cuda:0'), covar=tensor([0.1181, 0.0830, 0.1000, 0.0800, 0.0767, 0.1612, 0.0948, 0.1060], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0744, 0.0610, 0.0544, 0.0465, 0.0475, 0.0618, 0.0569], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 08:57:24,943 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 08:57:32,001 INFO [train.py:904] (0/8) Epoch 16, batch 6100, loss[loss=0.2086, simple_loss=0.2978, pruned_loss=0.05969, over 16204.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2977, pruned_loss=0.06321, over 3103952.23 frames. ], batch size: 165, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:58:24,042 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 08:58:37,185 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 08:58:46,001 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7762, 3.8579, 3.9158, 3.7059, 3.8529, 4.2602, 3.8876, 3.6334], device='cuda:0'), covar=tensor([0.2218, 0.2006, 0.2280, 0.2387, 0.2657, 0.1592, 0.1561, 0.2592], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0547, 0.0596, 0.0461, 0.0618, 0.0624, 0.0470, 0.0621], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 08:58:49,711 INFO [train.py:904] (0/8) Epoch 16, batch 6150, loss[loss=0.194, simple_loss=0.2871, pruned_loss=0.05039, over 16419.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2962, pruned_loss=0.06287, over 3101059.66 frames. ], batch size: 146, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 08:59:11,310 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4537, 3.0105, 2.9387, 1.9077, 2.7165, 2.1337, 3.0464, 3.1548], device='cuda:0'), covar=tensor([0.0279, 0.0667, 0.0590, 0.1951, 0.0812, 0.0955, 0.0644, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0155, 0.0161, 0.0148, 0.0139, 0.0125, 0.0140, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 08:59:18,318 INFO [optim.py:368] (0/8) 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 08:59:57,086 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158445.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:00:08,006 INFO [train.py:904] (0/8) Epoch 16, batch 6200, loss[loss=0.2071, simple_loss=0.2853, pruned_loss=0.06442, over 15440.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.294, pruned_loss=0.06242, over 3105296.53 frames. ], batch size: 191, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:00:47,991 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158478.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:12,340 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 09:01:21,949 INFO [train.py:904] (0/8) Epoch 16, batch 6250, loss[loss=0.239, simple_loss=0.3061, pruned_loss=0.0859, over 11811.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2934, pruned_loss=0.06198, over 3101211.71 frames. ], batch size: 248, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:01:34,535 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:01:44,126 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9459, 2.9701, 2.6393, 4.9772, 3.6390, 4.3369, 2.0262, 2.8292], device='cuda:0'), covar=tensor([0.1198, 0.0718, 0.1167, 0.0150, 0.0413, 0.0459, 0.1334, 0.0937], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0166, 0.0187, 0.0170, 0.0201, 0.0212, 0.0190, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 09:01:50,904 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.774e+02 2.618e+02 3.109e+02 4.242e+02 1.066e+03, threshold=6.218e+02, percent-clipped=4.0 2023-04-30 09:02:20,178 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4492, 4.4715, 4.8419, 4.7933, 4.8159, 4.4939, 4.5047, 4.3366], device='cuda:0'), covar=tensor([0.0292, 0.0528, 0.0365, 0.0436, 0.0419, 0.0372, 0.0883, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0405, 0.0395, 0.0376, 0.0447, 0.0419, 0.0519, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 09:02:24,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8195, 4.8362, 4.6818, 4.3633, 4.3088, 4.7819, 4.6081, 4.4473], device='cuda:0'), covar=tensor([0.0687, 0.0646, 0.0292, 0.0318, 0.1019, 0.0508, 0.0441, 0.0764], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0370, 0.0312, 0.0300, 0.0326, 0.0350, 0.0215, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:02:38,740 INFO [train.py:904] (0/8) Epoch 16, batch 6300, loss[loss=0.2189, simple_loss=0.3044, pruned_loss=0.06668, over 16839.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2936, pruned_loss=0.06137, over 3106221.70 frames. ], batch size: 39, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:02:45,561 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:03:28,146 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 09:03:55,413 INFO [train.py:904] (0/8) Epoch 16, batch 6350, loss[loss=0.2246, simple_loss=0.3091, pruned_loss=0.0701, over 16949.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2946, pruned_loss=0.06261, over 3101593.32 frames. ], batch size: 109, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:04:16,121 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7439, 2.6105, 2.0156, 2.3231, 3.0975, 2.6248, 3.3938, 3.3859], device='cuda:0'), covar=tensor([0.0087, 0.0392, 0.0584, 0.0481, 0.0239, 0.0415, 0.0205, 0.0206], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0217, 0.0211, 0.0210, 0.0218, 0.0219, 0.0221, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:04:24,045 INFO [optim.py:368] (0/8) 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 09:04:46,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6094, 4.7046, 4.8867, 4.6920, 4.7347, 5.3088, 4.8072, 4.5615], device='cuda:0'), covar=tensor([0.1151, 0.1864, 0.2008, 0.1986, 0.2443, 0.0965, 0.1475, 0.2519], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0550, 0.0601, 0.0464, 0.0622, 0.0630, 0.0474, 0.0624], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:05:11,923 INFO [train.py:904] (0/8) Epoch 16, batch 6400, loss[loss=0.2887, simple_loss=0.3522, pruned_loss=0.1126, over 11305.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2954, pruned_loss=0.06412, over 3075859.95 frames. ], batch size: 247, lr: 4.23e-03, grad_scale: 8.0 2023-04-30 09:06:15,363 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0596, 2.9718, 1.8609, 3.2701, 2.3345, 3.3057, 2.1171, 2.6150], device='cuda:0'), covar=tensor([0.0358, 0.0648, 0.1884, 0.0297, 0.0901, 0.0917, 0.1503, 0.0779], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0172, 0.0195, 0.0149, 0.0174, 0.0212, 0.0202, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 09:06:27,734 INFO [train.py:904] (0/8) Epoch 16, batch 6450, loss[loss=0.1805, simple_loss=0.2772, pruned_loss=0.04194, over 16806.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2945, pruned_loss=0.06264, over 3089335.62 frames. ], batch size: 96, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:06:56,389 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.821e+02 3.663e+02 4.597e+02 9.189e+02, threshold=7.326e+02, percent-clipped=6.0 2023-04-30 09:07:25,933 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=158740.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:07:43,545 INFO [train.py:904] (0/8) Epoch 16, batch 6500, loss[loss=0.22, simple_loss=0.2869, pruned_loss=0.07657, over 11824.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2923, pruned_loss=0.06152, over 3103054.91 frames. ], batch size: 246, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:08:10,771 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 09:08:21,095 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:22,823 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=158778.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:08:28,674 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 09:08:31,623 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:09:01,982 INFO [train.py:904] (0/8) Epoch 16, batch 6550, loss[loss=0.2212, simple_loss=0.3226, pruned_loss=0.05986, over 16361.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2954, pruned_loss=0.06279, over 3096579.78 frames. ], batch size: 146, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:09:33,175 INFO [optim.py:368] (0/8) 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 09:09:39,593 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=158826.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:09:52,693 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=158835.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:09:57,651 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158838.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:06,322 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158844.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:10:18,021 INFO [train.py:904] (0/8) Epoch 16, batch 6600, loss[loss=0.2253, simple_loss=0.3112, pruned_loss=0.06974, over 16745.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2978, pruned_loss=0.06298, over 3118063.49 frames. ], batch size: 83, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:11:10,968 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 09:11:26,053 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=158896.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:11:31,970 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-04-30 09:11:34,223 INFO [train.py:904] (0/8) Epoch 16, batch 6650, loss[loss=0.2053, simple_loss=0.2936, pruned_loss=0.0585, over 16306.00 frames. ], tot_loss[loss=0.2126, simple_loss=0.2979, pruned_loss=0.06368, over 3109881.34 frames. ], batch size: 165, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:11:40,577 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4312, 4.1068, 3.8784, 2.6325, 3.6603, 3.9667, 3.7070, 1.9254], device='cuda:0'), covar=tensor([0.0554, 0.0043, 0.0069, 0.0449, 0.0108, 0.0158, 0.0109, 0.0578], device='cuda:0'), in_proj_covar=tensor([0.0130, 0.0074, 0.0076, 0.0129, 0.0088, 0.0098, 0.0086, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:12:04,650 INFO [optim.py:368] (0/8) 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 09:12:50,521 INFO [train.py:904] (0/8) Epoch 16, batch 6700, loss[loss=0.2061, simple_loss=0.2843, pruned_loss=0.06398, over 15495.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2962, pruned_loss=0.06394, over 3100621.36 frames. ], batch size: 191, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:13:57,255 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3853, 3.6339, 3.6711, 2.1958, 3.2124, 2.2794, 3.8394, 3.7962], device='cuda:0'), covar=tensor([0.0208, 0.0656, 0.0572, 0.1847, 0.0774, 0.0982, 0.0518, 0.0788], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0157, 0.0163, 0.0149, 0.0140, 0.0126, 0.0141, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 09:14:07,737 INFO [train.py:904] (0/8) Epoch 16, batch 6750, loss[loss=0.2565, simple_loss=0.3209, pruned_loss=0.0961, over 12045.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2959, pruned_loss=0.06448, over 3082053.53 frames. ], batch size: 248, lr: 4.23e-03, grad_scale: 4.0 2023-04-30 09:14:33,092 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3221, 4.3823, 4.6154, 4.4366, 4.5019, 5.0215, 4.5398, 4.2998], device='cuda:0'), covar=tensor([0.1520, 0.1947, 0.2213, 0.1968, 0.2465, 0.1028, 0.1673, 0.2699], device='cuda:0'), in_proj_covar=tensor([0.0383, 0.0547, 0.0599, 0.0461, 0.0612, 0.0629, 0.0471, 0.0619], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:14:37,804 INFO [optim.py:368] (0/8) 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 09:15:05,326 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 16, batch 6800, loss[loss=0.2015, simple_loss=0.2837, pruned_loss=0.0596, over 16754.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2954, pruned_loss=0.06394, over 3099931.67 frames. ], batch size: 57, lr: 4.22e-03, grad_scale: 8.0 2023-04-30 09:15:50,682 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:16:29,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3909, 3.6920, 3.7009, 1.9943, 3.0589, 2.4868, 3.8311, 3.7586], device='cuda:0'), covar=tensor([0.0262, 0.0705, 0.0580, 0.2077, 0.0890, 0.0957, 0.0606, 0.0909], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0158, 0.0164, 0.0150, 0.0141, 0.0127, 0.0142, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 09:16:40,970 INFO [train.py:904] (0/8) Epoch 16, batch 6850, loss[loss=0.2312, simple_loss=0.3172, pruned_loss=0.07259, over 15329.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2969, pruned_loss=0.06447, over 3096796.96 frames. ], batch size: 190, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:17:12,964 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.918e+02 2.851e+02 3.389e+02 4.147e+02 9.414e+02, threshold=6.778e+02, percent-clipped=4.0 2023-04-30 09:17:22,476 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159130.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:26,686 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159133.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:35,365 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:17:40,224 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 09:17:54,585 INFO [train.py:904] (0/8) Epoch 16, batch 6900, loss[loss=0.2342, simple_loss=0.3171, pruned_loss=0.07572, over 17041.00 frames. ], tot_loss[loss=0.2137, simple_loss=0.2989, pruned_loss=0.06425, over 3094116.37 frames. ], batch size: 53, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:18:05,938 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4719, 1.7160, 2.0864, 2.4025, 2.4425, 2.6997, 1.7597, 2.6466], device='cuda:0'), covar=tensor([0.0170, 0.0426, 0.0289, 0.0276, 0.0265, 0.0176, 0.0497, 0.0115], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0187, 0.0172, 0.0175, 0.0187, 0.0143, 0.0186, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:18:07,176 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:18:29,613 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0824, 4.3135, 4.5748, 4.5035, 4.5163, 4.2072, 3.9711, 4.1676], device='cuda:0'), covar=tensor([0.0568, 0.0732, 0.0570, 0.0656, 0.0685, 0.0653, 0.1604, 0.0595], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0407, 0.0398, 0.0378, 0.0448, 0.0423, 0.0520, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 09:18:52,304 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 09:18:56,503 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159191.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 09:19:13,877 INFO [train.py:904] (0/8) Epoch 16, batch 6950, loss[loss=0.2793, simple_loss=0.3377, pruned_loss=0.1104, over 11476.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.3001, pruned_loss=0.06569, over 3080531.70 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:19:42,077 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:19:48,723 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.671e+02 3.353e+02 4.093e+02 5.038e+02 1.287e+03, threshold=8.186e+02, percent-clipped=11.0 2023-04-30 09:20:29,903 INFO [train.py:904] (0/8) Epoch 16, batch 7000, loss[loss=0.1924, simple_loss=0.2837, pruned_loss=0.0505, over 17041.00 frames. ], tot_loss[loss=0.216, simple_loss=0.3006, pruned_loss=0.06575, over 3077599.18 frames. ], batch size: 53, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:21:17,232 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 09:21:17,468 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 09:21:27,384 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:21:45,221 INFO [train.py:904] (0/8) Epoch 16, batch 7050, loss[loss=0.2653, simple_loss=0.3277, pruned_loss=0.1014, over 11512.00 frames. ], tot_loss[loss=0.2159, simple_loss=0.3012, pruned_loss=0.06531, over 3084999.80 frames. ], batch size: 246, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:22:18,897 INFO [optim.py:368] (0/8) 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 09:23:00,669 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159351.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:23:01,359 INFO [train.py:904] (0/8) Epoch 16, batch 7100, loss[loss=0.2331, simple_loss=0.2975, pruned_loss=0.08428, over 11931.00 frames. ], tot_loss[loss=0.2153, simple_loss=0.3, pruned_loss=0.0653, over 3071784.99 frames. ], batch size: 248, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:24:17,488 INFO [train.py:904] (0/8) Epoch 16, batch 7150, loss[loss=0.2258, simple_loss=0.3048, pruned_loss=0.07341, over 16387.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.2988, pruned_loss=0.06578, over 3064154.56 frames. ], batch size: 146, lr: 4.22e-03, grad_scale: 2.0 2023-04-30 09:24:49,380 INFO [optim.py:368] (0/8) 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 09:24:51,036 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159425.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:02,362 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:11,753 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:25:30,670 INFO [train.py:904] (0/8) Epoch 16, batch 7200, loss[loss=0.1828, simple_loss=0.2709, pruned_loss=0.04734, over 16899.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2958, pruned_loss=0.06341, over 3079429.07 frames. ], batch size: 109, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:26:15,847 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:26,899 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159487.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:26:33,348 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159491.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:26:50,674 INFO [train.py:904] (0/8) Epoch 16, batch 7250, loss[loss=0.2304, simple_loss=0.3135, pruned_loss=0.07368, over 16254.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2928, pruned_loss=0.06174, over 3085268.38 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:26:51,206 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5067, 3.5004, 3.4634, 2.7604, 3.3567, 2.0621, 3.1046, 2.7108], device='cuda:0'), covar=tensor([0.0150, 0.0127, 0.0163, 0.0219, 0.0098, 0.2209, 0.0128, 0.0201], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0135, 0.0182, 0.0168, 0.0154, 0.0194, 0.0168, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:27:10,910 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:27:21,048 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6752, 1.8207, 1.5772, 1.5789, 1.9536, 1.6431, 1.6363, 1.8895], device='cuda:0'), covar=tensor([0.0149, 0.0224, 0.0353, 0.0324, 0.0177, 0.0234, 0.0180, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0217, 0.0211, 0.0211, 0.0217, 0.0217, 0.0220, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:27:23,393 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.928e+02 2.845e+02 3.560e+02 4.449e+02 7.164e+02, threshold=7.119e+02, percent-clipped=0.0 2023-04-30 09:27:28,927 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5601, 3.5603, 2.7206, 2.1576, 2.4884, 2.2985, 3.7680, 3.2861], device='cuda:0'), covar=tensor([0.2760, 0.0680, 0.1755, 0.2650, 0.2488, 0.1975, 0.0489, 0.1178], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0263, 0.0296, 0.0298, 0.0289, 0.0240, 0.0281, 0.0318], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:27:40,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6892, 2.2356, 1.7388, 2.0066, 2.5424, 2.2129, 2.5237, 2.7343], device='cuda:0'), covar=tensor([0.0174, 0.0368, 0.0534, 0.0450, 0.0253, 0.0388, 0.0207, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0216, 0.0210, 0.0211, 0.0217, 0.0217, 0.0219, 0.0212], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:27:46,675 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159539.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 09:28:06,279 INFO [train.py:904] (0/8) Epoch 16, batch 7300, loss[loss=0.2416, simple_loss=0.3155, pruned_loss=0.08384, over 11406.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2926, pruned_loss=0.06221, over 3058876.12 frames. ], batch size: 247, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:22,397 INFO [train.py:904] (0/8) Epoch 16, batch 7350, loss[loss=0.2276, simple_loss=0.3095, pruned_loss=0.07288, over 16411.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2932, pruned_loss=0.06236, over 3074171.07 frames. ], batch size: 146, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:29:33,061 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5051, 3.5456, 3.2995, 2.9986, 3.1612, 3.4408, 3.3071, 3.2539], device='cuda:0'), covar=tensor([0.0546, 0.0547, 0.0252, 0.0246, 0.0484, 0.0390, 0.1092, 0.0458], device='cuda:0'), in_proj_covar=tensor([0.0261, 0.0364, 0.0305, 0.0294, 0.0318, 0.0342, 0.0211, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:29:56,736 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.741e+02 3.169e+02 3.768e+02 4.450e+02 9.358e+02, threshold=7.536e+02, percent-clipped=3.0 2023-04-30 09:30:06,108 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 09:30:07,715 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:27,883 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4539, 3.3234, 2.7206, 2.1452, 2.4345, 2.3336, 3.4631, 3.2270], device='cuda:0'), covar=tensor([0.2904, 0.0798, 0.1736, 0.2684, 0.2165, 0.1921, 0.0597, 0.1216], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0264, 0.0297, 0.0298, 0.0290, 0.0241, 0.0282, 0.0319], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:30:30,771 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159646.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:30:39,224 INFO [train.py:904] (0/8) Epoch 16, batch 7400, loss[loss=0.1949, simple_loss=0.2863, pruned_loss=0.0517, over 16565.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2947, pruned_loss=0.06291, over 3094002.84 frames. ], batch size: 68, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:31:41,318 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:31:57,491 INFO [train.py:904] (0/8) Epoch 16, batch 7450, loss[loss=0.2126, simple_loss=0.3092, pruned_loss=0.05803, over 16211.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2958, pruned_loss=0.06407, over 3085866.63 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:32:33,423 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 3.124e+02 3.664e+02 4.558e+02 8.463e+02, threshold=7.327e+02, percent-clipped=4.0 2023-04-30 09:32:35,928 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:33:17,744 INFO [train.py:904] (0/8) Epoch 16, batch 7500, loss[loss=0.2013, simple_loss=0.2872, pruned_loss=0.0577, over 16278.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2965, pruned_loss=0.06363, over 3077988.73 frames. ], batch size: 165, lr: 4.22e-03, grad_scale: 4.0 2023-04-30 09:33:50,301 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159773.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:33:52,454 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8507, 1.7508, 2.3459, 2.7952, 2.6691, 3.1306, 2.0071, 3.1000], device='cuda:0'), covar=tensor([0.0198, 0.0499, 0.0329, 0.0269, 0.0282, 0.0167, 0.0487, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0184, 0.0168, 0.0173, 0.0183, 0.0140, 0.0183, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:33:54,982 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4337, 4.4047, 4.2964, 3.0869, 4.3357, 1.4220, 3.9627, 3.9183], device='cuda:0'), covar=tensor([0.0178, 0.0114, 0.0233, 0.0635, 0.0144, 0.3532, 0.0183, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0133, 0.0180, 0.0165, 0.0152, 0.0191, 0.0166, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:34:37,019 INFO [train.py:904] (0/8) Epoch 16, batch 7550, loss[loss=0.1824, simple_loss=0.2719, pruned_loss=0.04647, over 16720.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2957, pruned_loss=0.06378, over 3077083.75 frames. ], batch size: 83, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:34:38,819 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7976, 1.3071, 1.6902, 1.5967, 1.7296, 1.8609, 1.5755, 1.7808], device='cuda:0'), covar=tensor([0.0214, 0.0331, 0.0170, 0.0234, 0.0221, 0.0145, 0.0370, 0.0102], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0184, 0.0168, 0.0172, 0.0183, 0.0140, 0.0183, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:34:58,288 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:35:10,146 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.970e+02 2.719e+02 3.178e+02 3.820e+02 7.109e+02, threshold=6.356e+02, percent-clipped=0.0 2023-04-30 09:35:21,301 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 09:35:53,838 INFO [train.py:904] (0/8) Epoch 16, batch 7600, loss[loss=0.2471, simple_loss=0.3203, pruned_loss=0.08692, over 11418.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2945, pruned_loss=0.06313, over 3099494.71 frames. ], batch size: 246, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:35:59,120 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:36:10,415 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:09,867 INFO [train.py:904] (0/8) Epoch 16, batch 7650, loss[loss=0.257, simple_loss=0.332, pruned_loss=0.09103, over 15376.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.295, pruned_loss=0.06392, over 3091579.26 frames. ], batch size: 190, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:37:21,643 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=159910.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:30,425 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159916.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:37:43,266 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.780e+02 3.207e+02 4.068e+02 5.310e+02 2.281e+03, threshold=8.135e+02, percent-clipped=15.0 2023-04-30 09:38:13,633 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=159946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:38:22,091 INFO [train.py:904] (0/8) Epoch 16, batch 7700, loss[loss=0.2098, simple_loss=0.3006, pruned_loss=0.05952, over 16656.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2949, pruned_loss=0.06428, over 3087074.80 frames. ], batch size: 134, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:38:41,783 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0694, 2.4485, 2.3536, 2.7787, 2.0830, 3.2338, 1.8399, 2.7571], device='cuda:0'), covar=tensor([0.1057, 0.0516, 0.1000, 0.0148, 0.0114, 0.0394, 0.1290, 0.0628], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0166, 0.0188, 0.0169, 0.0203, 0.0212, 0.0193, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 09:38:50,554 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=159971.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:15,442 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=159987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:26,412 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=159994.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:39:35,063 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-160000.pt 2023-04-30 09:39:39,466 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9047, 5.1615, 5.4135, 5.1933, 5.2327, 5.7847, 5.2500, 5.0549], device='cuda:0'), covar=tensor([0.0864, 0.1922, 0.2161, 0.1722, 0.2174, 0.0842, 0.1485, 0.2119], device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0551, 0.0601, 0.0463, 0.0613, 0.0630, 0.0473, 0.0621], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:39:41,049 INFO [train.py:904] (0/8) Epoch 16, batch 7750, loss[loss=0.1979, simple_loss=0.2826, pruned_loss=0.05667, over 16553.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2946, pruned_loss=0.0635, over 3101122.67 frames. ], batch size: 75, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:40:13,575 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.949e+02 3.102e+02 3.551e+02 4.185e+02 8.391e+02, threshold=7.102e+02, percent-clipped=1.0 2023-04-30 09:40:33,216 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6200, 4.6391, 5.0475, 5.0152, 5.0141, 4.6942, 4.6704, 4.4834], device='cuda:0'), covar=tensor([0.0303, 0.0581, 0.0346, 0.0397, 0.0426, 0.0350, 0.0906, 0.0481], device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0408, 0.0397, 0.0377, 0.0449, 0.0422, 0.0516, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 09:40:41,474 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 09:40:53,485 INFO [train.py:904] (0/8) Epoch 16, batch 7800, loss[loss=0.2337, simple_loss=0.3131, pruned_loss=0.0771, over 16225.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2959, pruned_loss=0.0642, over 3110538.03 frames. ], batch size: 165, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:08,841 INFO [train.py:904] (0/8) Epoch 16, batch 7850, loss[loss=0.199, simple_loss=0.2915, pruned_loss=0.05326, over 15436.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2966, pruned_loss=0.06385, over 3112675.39 frames. ], batch size: 190, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:42:43,407 INFO [optim.py:368] (0/8) 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 09:42:46,883 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3092, 1.5544, 1.9804, 2.1917, 2.3097, 2.5269, 1.6916, 2.4172], device='cuda:0'), covar=tensor([0.0199, 0.0460, 0.0248, 0.0317, 0.0268, 0.0194, 0.0474, 0.0156], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0183, 0.0167, 0.0172, 0.0182, 0.0140, 0.0183, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:43:25,079 INFO [train.py:904] (0/8) Epoch 16, batch 7900, loss[loss=0.2119, simple_loss=0.2992, pruned_loss=0.06236, over 16726.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.295, pruned_loss=0.06307, over 3108032.89 frames. ], batch size: 124, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:43:54,605 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7835, 3.6838, 3.9093, 3.7046, 3.8499, 4.2482, 3.8935, 3.6344], device='cuda:0'), covar=tensor([0.2152, 0.2528, 0.2569, 0.2538, 0.2649, 0.1760, 0.1657, 0.2627], device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0556, 0.0605, 0.0466, 0.0619, 0.0636, 0.0477, 0.0628], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:44:43,671 INFO [train.py:904] (0/8) Epoch 16, batch 7950, loss[loss=0.2211, simple_loss=0.3015, pruned_loss=0.07032, over 15432.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2956, pruned_loss=0.06405, over 3090993.77 frames. ], batch size: 191, lr: 4.21e-03, grad_scale: 4.0 2023-04-30 09:44:56,870 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:45:16,544 INFO [optim.py:368] (0/8) 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 09:45:56,371 INFO [train.py:904] (0/8) Epoch 16, batch 8000, loss[loss=0.2804, simple_loss=0.339, pruned_loss=0.111, over 11265.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2963, pruned_loss=0.06466, over 3083910.49 frames. ], batch size: 246, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:46:17,936 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:46:49,589 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:47:12,593 INFO [train.py:904] (0/8) Epoch 16, batch 8050, loss[loss=0.2194, simple_loss=0.3064, pruned_loss=0.06623, over 16647.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2962, pruned_loss=0.06447, over 3082861.59 frames. ], batch size: 124, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:47:20,349 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160307.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:47:47,810 INFO [optim.py:368] (0/8) 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 09:48:04,049 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160335.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:48:29,615 INFO [train.py:904] (0/8) Epoch 16, batch 8100, loss[loss=0.2403, simple_loss=0.3174, pruned_loss=0.08161, over 16712.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2958, pruned_loss=0.06373, over 3096623.04 frames. ], batch size: 124, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:48:54,936 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160368.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 09:48:56,040 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9055, 3.1937, 3.2181, 2.1439, 2.9822, 3.1537, 3.0564, 1.8739], device='cuda:0'), covar=tensor([0.0525, 0.0058, 0.0056, 0.0387, 0.0092, 0.0111, 0.0085, 0.0440], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0074, 0.0075, 0.0130, 0.0088, 0.0098, 0.0086, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:49:37,350 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4433, 3.3888, 3.8312, 1.6673, 3.9393, 3.9647, 2.9349, 2.8152], device='cuda:0'), covar=tensor([0.0835, 0.0266, 0.0178, 0.1330, 0.0061, 0.0146, 0.0420, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0106, 0.0092, 0.0138, 0.0074, 0.0117, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 09:49:46,015 INFO [train.py:904] (0/8) Epoch 16, batch 8150, loss[loss=0.2382, simple_loss=0.3041, pruned_loss=0.08618, over 11485.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2945, pruned_loss=0.06384, over 3065892.28 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:49:55,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3756, 4.7179, 4.2830, 4.5297, 4.2599, 4.2081, 4.2448, 4.7436], device='cuda:0'), covar=tensor([0.2064, 0.1369, 0.2209, 0.1342, 0.1507, 0.2100, 0.2268, 0.1658], device='cuda:0'), in_proj_covar=tensor([0.0602, 0.0739, 0.0605, 0.0543, 0.0464, 0.0479, 0.0614, 0.0566], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:50:21,739 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.821e+02 3.128e+02 3.748e+02 4.647e+02 9.176e+02, threshold=7.497e+02, percent-clipped=1.0 2023-04-30 09:50:49,923 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 09:51:05,080 INFO [train.py:904] (0/8) Epoch 16, batch 8200, loss[loss=0.2329, simple_loss=0.3031, pruned_loss=0.08139, over 11728.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2923, pruned_loss=0.06329, over 3076259.61 frames. ], batch size: 247, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:51:24,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9220, 2.4089, 2.4673, 2.9634, 2.3988, 3.1519, 1.7011, 2.8192], device='cuda:0'), covar=tensor([0.1233, 0.0595, 0.0938, 0.0223, 0.0145, 0.0353, 0.1441, 0.0617], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0171, 0.0204, 0.0212, 0.0193, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 09:52:27,321 INFO [train.py:904] (0/8) Epoch 16, batch 8250, loss[loss=0.1974, simple_loss=0.2915, pruned_loss=0.05167, over 16880.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2904, pruned_loss=0.06044, over 3056219.65 frames. ], batch size: 116, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:52:42,367 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:53:00,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2941, 1.6413, 1.9000, 2.3095, 2.3033, 2.5999, 1.7564, 2.4531], device='cuda:0'), covar=tensor([0.0215, 0.0468, 0.0338, 0.0287, 0.0309, 0.0184, 0.0476, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0184, 0.0169, 0.0173, 0.0183, 0.0141, 0.0184, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:53:04,607 INFO [optim.py:368] (0/8) 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 09:53:16,934 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.84 vs. limit=5.0 2023-04-30 09:53:49,473 INFO [train.py:904] (0/8) Epoch 16, batch 8300, loss[loss=0.1767, simple_loss=0.254, pruned_loss=0.04975, over 12252.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2875, pruned_loss=0.05738, over 3043092.41 frames. ], batch size: 248, lr: 4.21e-03, grad_scale: 8.0 2023-04-30 09:54:00,905 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:13,053 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:54:30,369 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4495, 3.2967, 2.7028, 2.1287, 2.1259, 2.3152, 3.4234, 3.0448], device='cuda:0'), covar=tensor([0.2818, 0.0706, 0.1675, 0.2862, 0.2616, 0.2011, 0.0422, 0.1229], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0259, 0.0292, 0.0295, 0.0285, 0.0237, 0.0277, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 09:54:40,664 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 09:55:10,572 INFO [train.py:904] (0/8) Epoch 16, batch 8350, loss[loss=0.2036, simple_loss=0.2813, pruned_loss=0.06298, over 12401.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2869, pruned_loss=0.05539, over 3058533.63 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:55:30,915 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=160614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 09:55:32,866 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0549, 3.9234, 4.1333, 4.2471, 4.3828, 3.9684, 4.3219, 4.3987], device='cuda:0'), covar=tensor([0.1608, 0.1210, 0.1370, 0.0656, 0.0508, 0.1278, 0.0656, 0.0630], device='cuda:0'), in_proj_covar=tensor([0.0565, 0.0701, 0.0822, 0.0706, 0.0535, 0.0558, 0.0570, 0.0666], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:55:48,786 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.217e+02 2.608e+02 3.326e+02 6.898e+02, threshold=5.216e+02, percent-clipped=1.0 2023-04-30 09:55:57,541 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2391, 5.2361, 5.6541, 5.6316, 5.6275, 5.3790, 5.2241, 5.0997], device='cuda:0'), covar=tensor([0.0323, 0.0738, 0.0367, 0.0385, 0.0443, 0.0330, 0.0992, 0.0402], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0405, 0.0395, 0.0374, 0.0443, 0.0418, 0.0512, 0.0332], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 09:56:33,071 INFO [train.py:904] (0/8) Epoch 16, batch 8400, loss[loss=0.1779, simple_loss=0.2816, pruned_loss=0.03714, over 16798.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2839, pruned_loss=0.05298, over 3051341.03 frames. ], batch size: 102, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:56:50,589 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=160663.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 09:57:09,610 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2821, 2.1206, 2.2135, 3.9954, 1.9898, 2.5167, 2.2186, 2.2513], device='cuda:0'), covar=tensor([0.1069, 0.3902, 0.2734, 0.0419, 0.4263, 0.2419, 0.3562, 0.3576], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0410, 0.0339, 0.0308, 0.0415, 0.0469, 0.0378, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 09:57:54,563 INFO [train.py:904] (0/8) Epoch 16, batch 8450, loss[loss=0.1767, simple_loss=0.272, pruned_loss=0.04065, over 16457.00 frames. ], tot_loss[loss=0.1924, simple_loss=0.282, pruned_loss=0.05142, over 3055840.73 frames. ], batch size: 68, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:58:31,821 INFO [optim.py:368] (0/8) 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 09:59:15,557 INFO [train.py:904] (0/8) Epoch 16, batch 8500, loss[loss=0.1789, simple_loss=0.2795, pruned_loss=0.03918, over 16812.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2786, pruned_loss=0.04894, over 3072100.44 frames. ], batch size: 102, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 09:59:46,710 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5464, 4.8494, 4.6411, 4.6564, 4.4112, 4.4017, 4.3158, 4.9148], device='cuda:0'), covar=tensor([0.1137, 0.0856, 0.0971, 0.0734, 0.0764, 0.1114, 0.1086, 0.0821], device='cuda:0'), in_proj_covar=tensor([0.0595, 0.0724, 0.0592, 0.0533, 0.0455, 0.0473, 0.0605, 0.0557], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:00:06,628 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3290, 4.3015, 4.1497, 3.5454, 4.2187, 1.7577, 3.9780, 3.9597], device='cuda:0'), covar=tensor([0.0096, 0.0097, 0.0186, 0.0315, 0.0107, 0.2512, 0.0143, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0131, 0.0177, 0.0162, 0.0149, 0.0189, 0.0164, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:00:42,455 INFO [train.py:904] (0/8) Epoch 16, batch 8550, loss[loss=0.2094, simple_loss=0.3088, pruned_loss=0.05501, over 16739.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2765, pruned_loss=0.04775, over 3064412.81 frames. ], batch size: 124, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:01:27,141 INFO [optim.py:368] (0/8) 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 10:01:44,425 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7284, 3.9145, 3.9640, 2.8196, 3.4471, 3.8933, 3.6774, 2.2152], device='cuda:0'), covar=tensor([0.0379, 0.0041, 0.0034, 0.0298, 0.0087, 0.0077, 0.0058, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0074, 0.0075, 0.0129, 0.0088, 0.0098, 0.0087, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 10:01:56,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5097, 2.1017, 1.7790, 1.7898, 2.3870, 2.0496, 2.1028, 2.5029], device='cuda:0'), covar=tensor([0.0168, 0.0367, 0.0493, 0.0467, 0.0266, 0.0365, 0.0185, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0213, 0.0207, 0.0207, 0.0211, 0.0213, 0.0215, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:02:19,531 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1986, 3.2079, 1.9520, 3.4574, 2.3697, 3.4546, 2.0271, 2.6399], device='cuda:0'), covar=tensor([0.0286, 0.0376, 0.1605, 0.0243, 0.0905, 0.0672, 0.1605, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0143, 0.0168, 0.0204, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 10:02:22,627 INFO [train.py:904] (0/8) Epoch 16, batch 8600, loss[loss=0.1729, simple_loss=0.2566, pruned_loss=0.04464, over 12428.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2763, pruned_loss=0.04649, over 3047639.96 frames. ], batch size: 250, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:02:25,232 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4475, 3.3580, 3.4673, 3.5321, 3.5893, 3.2957, 3.5549, 3.6072], device='cuda:0'), covar=tensor([0.1021, 0.0842, 0.0944, 0.0578, 0.0539, 0.2413, 0.0712, 0.0663], device='cuda:0'), in_proj_covar=tensor([0.0557, 0.0690, 0.0810, 0.0696, 0.0526, 0.0550, 0.0560, 0.0658], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:04:02,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160901.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:04:03,219 INFO [train.py:904] (0/8) Epoch 16, batch 8650, loss[loss=0.1657, simple_loss=0.2649, pruned_loss=0.03325, over 15143.00 frames. ], tot_loss[loss=0.1823, simple_loss=0.2744, pruned_loss=0.04509, over 3053887.40 frames. ], batch size: 190, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:04:56,269 INFO [optim.py:368] (0/8) 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 10:05:52,477 INFO [train.py:904] (0/8) Epoch 16, batch 8700, loss[loss=0.1711, simple_loss=0.266, pruned_loss=0.03815, over 16412.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2714, pruned_loss=0.04374, over 3062648.01 frames. ], batch size: 146, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:06:13,105 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=160962.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:06:15,204 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=160963.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 10:06:18,529 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=160965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:07:29,345 INFO [train.py:904] (0/8) Epoch 16, batch 8750, loss[loss=0.1664, simple_loss=0.2696, pruned_loss=0.0316, over 16735.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2714, pruned_loss=0.04314, over 3079520.25 frames. ], batch size: 83, lr: 4.20e-03, grad_scale: 4.0 2023-04-30 10:07:53,446 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161011.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 10:07:56,115 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-30 10:08:14,985 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1940, 3.5144, 3.5477, 2.2934, 3.1722, 3.5141, 3.3187, 2.0764], device='cuda:0'), covar=tensor([0.0451, 0.0039, 0.0040, 0.0387, 0.0085, 0.0070, 0.0067, 0.0445], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0073, 0.0075, 0.0130, 0.0088, 0.0097, 0.0086, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 10:08:27,684 INFO [optim.py:368] (0/8) 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 10:08:28,277 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:09:20,786 INFO [train.py:904] (0/8) Epoch 16, batch 8800, loss[loss=0.1857, simple_loss=0.2679, pruned_loss=0.05171, over 12620.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2693, pruned_loss=0.042, over 3064758.77 frames. ], batch size: 247, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:05,651 INFO [train.py:904] (0/8) Epoch 16, batch 8850, loss[loss=0.181, simple_loss=0.2799, pruned_loss=0.04108, over 17036.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2721, pruned_loss=0.04142, over 3064893.26 frames. ], batch size: 55, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:11:55,947 INFO [optim.py:368] (0/8) 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 10:12:52,576 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-30 10:12:53,430 INFO [train.py:904] (0/8) Epoch 16, batch 8900, loss[loss=0.1674, simple_loss=0.2563, pruned_loss=0.03926, over 12598.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2719, pruned_loss=0.04092, over 3036521.65 frames. ], batch size: 248, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:13:01,553 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8782, 3.9920, 2.4259, 4.4863, 2.9855, 4.4151, 2.6241, 3.1944], device='cuda:0'), covar=tensor([0.0227, 0.0266, 0.1502, 0.0186, 0.0736, 0.0403, 0.1391, 0.0630], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0141, 0.0167, 0.0200, 0.0195, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 10:14:59,661 INFO [train.py:904] (0/8) Epoch 16, batch 8950, loss[loss=0.177, simple_loss=0.2624, pruned_loss=0.04584, over 12960.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2714, pruned_loss=0.04099, over 3056396.09 frames. ], batch size: 250, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:15:07,279 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4623, 3.4706, 2.1723, 3.8985, 2.6692, 3.8394, 2.1675, 2.7466], device='cuda:0'), covar=tensor([0.0312, 0.0395, 0.1624, 0.0152, 0.0861, 0.0563, 0.1734, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0163, 0.0185, 0.0140, 0.0167, 0.0200, 0.0195, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 10:15:29,185 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:15:47,471 INFO [optim.py:368] (0/8) 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 10:16:01,833 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:16:04,793 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1360, 1.5101, 1.8405, 2.0818, 2.2185, 2.2694, 1.6772, 2.2848], device='cuda:0'), covar=tensor([0.0207, 0.0434, 0.0271, 0.0256, 0.0279, 0.0178, 0.0432, 0.0116], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0181, 0.0167, 0.0168, 0.0181, 0.0137, 0.0182, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:16:46,889 INFO [train.py:904] (0/8) Epoch 16, batch 9000, loss[loss=0.1563, simple_loss=0.2529, pruned_loss=0.02983, over 15463.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2681, pruned_loss=0.03943, over 3072580.91 frames. ], batch size: 191, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:16:46,890 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 10:16:56,910 INFO [train.py:938] (0/8) Epoch 16, validation: loss=0.1491, simple_loss=0.2531, pruned_loss=0.02259, over 944034.00 frames. 2023-04-30 10:16:56,911 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 10:17:07,653 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:17:33,496 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8415, 2.1225, 1.7103, 1.9109, 2.4868, 2.1545, 2.4661, 2.6777], device='cuda:0'), covar=tensor([0.0136, 0.0422, 0.0560, 0.0502, 0.0298, 0.0427, 0.0224, 0.0272], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0215, 0.0209, 0.0209, 0.0214, 0.0214, 0.0216, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:17:48,799 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:17:54,327 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8955, 1.9367, 2.2448, 3.2218, 2.0813, 2.1371, 2.1482, 1.9774], device='cuda:0'), covar=tensor([0.1179, 0.3836, 0.2564, 0.0631, 0.4324, 0.2860, 0.3532, 0.4171], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0412, 0.0345, 0.0310, 0.0418, 0.0472, 0.0382, 0.0479], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:17:56,223 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 10:17:56,414 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 10:18:20,223 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:18:40,782 INFO [train.py:904] (0/8) Epoch 16, batch 9050, loss[loss=0.1535, simple_loss=0.2505, pruned_loss=0.0283, over 16772.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2686, pruned_loss=0.03977, over 3073737.70 frames. ], batch size: 90, lr: 4.20e-03, grad_scale: 8.0 2023-04-30 10:19:19,371 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161321.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:19:27,694 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.317e+02 2.661e+02 3.212e+02 6.846e+02, threshold=5.322e+02, percent-clipped=5.0 2023-04-30 10:20:14,387 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 10:20:22,817 INFO [train.py:904] (0/8) Epoch 16, batch 9100, loss[loss=0.1855, simple_loss=0.2757, pruned_loss=0.04764, over 16629.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2688, pruned_loss=0.04061, over 3072796.63 frames. ], batch size: 62, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:22:19,311 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:22:19,986 INFO [train.py:904] (0/8) Epoch 16, batch 9150, loss[loss=0.1793, simple_loss=0.2662, pruned_loss=0.04618, over 16818.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2692, pruned_loss=0.04024, over 3071938.96 frames. ], batch size: 124, lr: 4.19e-03, grad_scale: 4.0 2023-04-30 10:23:13,976 INFO [optim.py:368] (0/8) 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 10:24:04,524 INFO [train.py:904] (0/8) Epoch 16, batch 9200, loss[loss=0.1564, simple_loss=0.2445, pruned_loss=0.03412, over 16710.00 frames. ], tot_loss[loss=0.172, simple_loss=0.265, pruned_loss=0.03953, over 3075548.10 frames. ], batch size: 39, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:24:20,251 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:24:24,456 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161462.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:25:42,701 INFO [train.py:904] (0/8) Epoch 16, batch 9250, loss[loss=0.1603, simple_loss=0.2554, pruned_loss=0.03267, over 16304.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2648, pruned_loss=0.03959, over 3073181.18 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:26:15,820 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6564, 2.7469, 2.4296, 2.5331, 3.0834, 2.8219, 3.2914, 3.2939], device='cuda:0'), covar=tensor([0.0095, 0.0337, 0.0422, 0.0421, 0.0230, 0.0336, 0.0192, 0.0219], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0214, 0.0208, 0.0208, 0.0213, 0.0214, 0.0214, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:26:22,890 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161521.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:26:36,797 INFO [optim.py:368] (0/8) 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 10:27:18,895 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-04-30 10:27:34,802 INFO [train.py:904] (0/8) Epoch 16, batch 9300, loss[loss=0.1549, simple_loss=0.2371, pruned_loss=0.03631, over 12172.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2634, pruned_loss=0.03906, over 3068705.44 frames. ], batch size: 248, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:27:45,424 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161557.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:22,588 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161572.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:28:54,025 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:29:21,692 INFO [train.py:904] (0/8) Epoch 16, batch 9350, loss[loss=0.2029, simple_loss=0.2909, pruned_loss=0.05741, over 16267.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2627, pruned_loss=0.03871, over 3063808.51 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:29:28,432 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161605.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:02,109 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:30:12,563 INFO [optim.py:368] (0/8) 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 10:30:52,714 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2902, 4.3836, 4.2057, 3.9023, 3.8858, 4.2941, 3.9932, 4.0429], device='cuda:0'), covar=tensor([0.0621, 0.0587, 0.0312, 0.0311, 0.0798, 0.0559, 0.0731, 0.0621], device='cuda:0'), in_proj_covar=tensor([0.0257, 0.0358, 0.0299, 0.0289, 0.0309, 0.0336, 0.0208, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:31:04,000 INFO [train.py:904] (0/8) Epoch 16, batch 9400, loss[loss=0.1799, simple_loss=0.2769, pruned_loss=0.04143, over 15501.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2628, pruned_loss=0.03845, over 3051697.76 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:31:38,660 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161669.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:31:48,776 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4470, 3.4995, 2.1361, 3.8997, 2.5009, 3.8845, 2.2526, 2.8483], device='cuda:0'), covar=tensor([0.0278, 0.0399, 0.1643, 0.0253, 0.0956, 0.0568, 0.1495, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0164, 0.0184, 0.0140, 0.0166, 0.0199, 0.0194, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 10:31:54,651 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7437, 1.3095, 1.6414, 1.6442, 1.8079, 1.8772, 1.5386, 1.8168], device='cuda:0'), covar=tensor([0.0251, 0.0367, 0.0199, 0.0291, 0.0261, 0.0181, 0.0394, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0178, 0.0164, 0.0166, 0.0178, 0.0135, 0.0179, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:32:44,156 INFO [train.py:904] (0/8) Epoch 16, batch 9450, loss[loss=0.1752, simple_loss=0.2579, pruned_loss=0.04628, over 12419.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2647, pruned_loss=0.03863, over 3054339.64 frames. ], batch size: 247, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:33:33,863 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.598e+02 2.284e+02 2.576e+02 3.195e+02 6.155e+02, threshold=5.152e+02, percent-clipped=3.0 2023-04-30 10:33:50,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9051, 2.0533, 2.3557, 2.8617, 2.7561, 3.1944, 2.1959, 3.1858], device='cuda:0'), covar=tensor([0.0184, 0.0438, 0.0342, 0.0293, 0.0293, 0.0148, 0.0430, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0166, 0.0167, 0.0179, 0.0136, 0.0180, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:34:23,910 INFO [train.py:904] (0/8) Epoch 16, batch 9500, loss[loss=0.1741, simple_loss=0.2726, pruned_loss=0.03778, over 16329.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2641, pruned_loss=0.03846, over 3051380.49 frames. ], batch size: 165, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:34:37,274 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161757.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:34:47,319 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5935, 3.5970, 2.8900, 2.1981, 2.2608, 2.3278, 3.8015, 3.2137], device='cuda:0'), covar=tensor([0.2659, 0.0554, 0.1566, 0.2708, 0.2779, 0.2067, 0.0389, 0.1157], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0254, 0.0286, 0.0288, 0.0273, 0.0233, 0.0273, 0.0306], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:35:08,765 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3558, 2.1741, 2.2156, 4.1816, 2.2320, 2.5705, 2.3219, 2.3852], device='cuda:0'), covar=tensor([0.1136, 0.3755, 0.2922, 0.0404, 0.4037, 0.2602, 0.3529, 0.3347], device='cuda:0'), in_proj_covar=tensor([0.0368, 0.0408, 0.0341, 0.0308, 0.0415, 0.0466, 0.0377, 0.0472], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:35:10,971 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 10:36:08,228 INFO [train.py:904] (0/8) Epoch 16, batch 9550, loss[loss=0.1804, simple_loss=0.2805, pruned_loss=0.04013, over 16789.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2633, pruned_loss=0.0383, over 3051673.83 frames. ], batch size: 83, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:36:38,178 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=161815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:40,244 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=161816.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:36:40,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8601, 2.2124, 1.9230, 2.0075, 2.5811, 2.2195, 2.4093, 2.6852], device='cuda:0'), covar=tensor([0.0131, 0.0362, 0.0469, 0.0447, 0.0237, 0.0367, 0.0210, 0.0242], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0216, 0.0209, 0.0208, 0.0214, 0.0216, 0.0215, 0.0206], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:37:00,501 INFO [optim.py:368] (0/8) 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 10:37:51,445 INFO [train.py:904] (0/8) Epoch 16, batch 9600, loss[loss=0.1892, simple_loss=0.2718, pruned_loss=0.05333, over 12324.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2643, pruned_loss=0.03921, over 3027828.69 frames. ], batch size: 248, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:38:29,833 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161872.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:38:38,484 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=161876.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:02,474 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=161887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:39:37,502 INFO [train.py:904] (0/8) Epoch 16, batch 9650, loss[loss=0.1794, simple_loss=0.2756, pruned_loss=0.04164, over 15561.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2663, pruned_loss=0.04001, over 3002255.32 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:40:21,560 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161920.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:27,877 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5107, 3.5045, 3.4882, 2.8414, 3.4232, 1.9485, 3.3102, 2.9768], device='cuda:0'), covar=tensor([0.0123, 0.0106, 0.0147, 0.0177, 0.0097, 0.2252, 0.0121, 0.0217], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0128, 0.0173, 0.0156, 0.0147, 0.0187, 0.0160, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:40:36,073 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.311e+02 2.824e+02 3.460e+02 5.730e+02, threshold=5.647e+02, percent-clipped=1.0 2023-04-30 10:40:49,471 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=161935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:40:50,005 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 10:40:53,244 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3681, 2.1861, 2.2636, 4.2419, 2.1675, 2.4918, 2.3196, 2.4397], device='cuda:0'), covar=tensor([0.1121, 0.3681, 0.2863, 0.0376, 0.4188, 0.2549, 0.3448, 0.3195], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0409, 0.0343, 0.0309, 0.0417, 0.0467, 0.0379, 0.0474], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:41:23,674 INFO [train.py:904] (0/8) Epoch 16, batch 9700, loss[loss=0.1713, simple_loss=0.2633, pruned_loss=0.0396, over 15260.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2652, pruned_loss=0.03946, over 3011120.49 frames. ], batch size: 191, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:41:50,320 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 10:43:02,478 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-162000.pt 2023-04-30 10:43:08,507 INFO [train.py:904] (0/8) Epoch 16, batch 9750, loss[loss=0.1737, simple_loss=0.2635, pruned_loss=0.04192, over 16955.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2641, pruned_loss=0.03981, over 3021245.93 frames. ], batch size: 109, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:43:51,756 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 10:43:58,495 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.156e+02 2.615e+02 3.105e+02 5.589e+02, threshold=5.230e+02, percent-clipped=0.0 2023-04-30 10:44:46,291 INFO [train.py:904] (0/8) Epoch 16, batch 9800, loss[loss=0.1756, simple_loss=0.2651, pruned_loss=0.04302, over 12165.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2645, pruned_loss=0.03901, over 3031239.50 frames. ], batch size: 247, lr: 4.19e-03, grad_scale: 8.0 2023-04-30 10:44:56,591 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162057.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:45:18,485 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6743, 2.3975, 2.2435, 3.4938, 2.0709, 3.6025, 1.3913, 2.7845], device='cuda:0'), covar=tensor([0.1453, 0.0776, 0.1276, 0.0147, 0.0098, 0.0381, 0.1808, 0.0757], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0161, 0.0184, 0.0162, 0.0189, 0.0204, 0.0189, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0003, 0.0003], device='cuda:0') 2023-04-30 10:45:34,291 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:31,265 INFO [train.py:904] (0/8) Epoch 16, batch 9850, loss[loss=0.1714, simple_loss=0.2675, pruned_loss=0.03762, over 16861.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2661, pruned_loss=0.03899, over 3034302.72 frames. ], batch size: 116, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:46:38,615 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162105.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:46:48,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9636, 4.2474, 4.0926, 4.0983, 3.7706, 3.8251, 3.9133, 4.2184], device='cuda:0'), covar=tensor([0.1247, 0.0838, 0.0929, 0.0750, 0.0895, 0.1678, 0.0937, 0.0996], device='cuda:0'), in_proj_covar=tensor([0.0583, 0.0719, 0.0580, 0.0526, 0.0452, 0.0467, 0.0596, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:46:59,797 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:47:23,361 INFO [optim.py:368] (0/8) 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 10:47:51,523 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162138.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:48:22,262 INFO [train.py:904] (0/8) Epoch 16, batch 9900, loss[loss=0.1632, simple_loss=0.2667, pruned_loss=0.02984, over 16888.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2663, pruned_loss=0.03873, over 3032166.97 frames. ], batch size: 96, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:48:52,697 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162164.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:49:08,588 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162171.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:50:22,158 INFO [train.py:904] (0/8) Epoch 16, batch 9950, loss[loss=0.1802, simple_loss=0.2684, pruned_loss=0.04597, over 17186.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2683, pruned_loss=0.03913, over 3034776.10 frames. ], batch size: 44, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:51:26,526 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.161e+02 2.630e+02 3.300e+02 8.357e+02, threshold=5.260e+02, percent-clipped=1.0 2023-04-30 10:51:36,407 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162231.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:52:23,907 INFO [train.py:904] (0/8) Epoch 16, batch 10000, loss[loss=0.1555, simple_loss=0.2606, pruned_loss=0.02519, over 16835.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2666, pruned_loss=0.0385, over 3066963.92 frames. ], batch size: 76, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:53:41,588 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0696, 2.0319, 2.0663, 3.6433, 1.9824, 2.3531, 2.2064, 2.1765], device='cuda:0'), covar=tensor([0.1112, 0.3514, 0.2829, 0.0476, 0.4087, 0.2422, 0.3318, 0.3078], device='cuda:0'), in_proj_covar=tensor([0.0365, 0.0403, 0.0339, 0.0305, 0.0411, 0.0462, 0.0374, 0.0468], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:53:46,661 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:54:04,084 INFO [train.py:904] (0/8) Epoch 16, batch 10050, loss[loss=0.1784, simple_loss=0.271, pruned_loss=0.04289, over 16940.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2666, pruned_loss=0.03846, over 3071926.72 frames. ], batch size: 109, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:54:54,661 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.225e+02 2.711e+02 3.158e+02 7.314e+02, threshold=5.421e+02, percent-clipped=6.0 2023-04-30 10:55:38,464 INFO [train.py:904] (0/8) Epoch 16, batch 10100, loss[loss=0.1557, simple_loss=0.2437, pruned_loss=0.0338, over 12420.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.267, pruned_loss=0.03859, over 3061005.86 frames. ], batch size: 248, lr: 4.18e-03, grad_scale: 8.0 2023-04-30 10:56:59,050 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-16.pt 2023-04-30 10:57:23,036 INFO [train.py:904] (0/8) Epoch 17, batch 0, loss[loss=0.2194, simple_loss=0.2817, pruned_loss=0.07853, over 16764.00 frames. ], tot_loss[loss=0.2194, simple_loss=0.2817, pruned_loss=0.07853, over 16764.00 frames. ], batch size: 83, lr: 4.05e-03, grad_scale: 8.0 2023-04-30 10:57:23,037 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 10:57:30,750 INFO [train.py:938] (0/8) Epoch 17, validation: loss=0.1481, simple_loss=0.2518, pruned_loss=0.02217, over 944034.00 frames. 2023-04-30 10:57:30,751 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 10:58:09,505 INFO [optim.py:368] (0/8) 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 10:58:14,145 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:58:39,861 INFO [train.py:904] (0/8) Epoch 17, batch 50, loss[loss=0.2056, simple_loss=0.2887, pruned_loss=0.06122, over 17064.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2761, pruned_loss=0.05478, over 746543.71 frames. ], batch size: 53, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 10:59:06,427 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 10:59:13,250 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4707, 2.2304, 2.4628, 4.3424, 2.2085, 2.6784, 2.3422, 2.4461], device='cuda:0'), covar=tensor([0.1063, 0.3638, 0.2636, 0.0410, 0.3949, 0.2349, 0.3403, 0.3287], device='cuda:0'), in_proj_covar=tensor([0.0372, 0.0410, 0.0345, 0.0312, 0.0419, 0.0470, 0.0382, 0.0476], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:59:35,895 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6828, 2.3169, 1.8899, 2.1615, 2.7951, 2.4994, 2.7934, 2.8595], device='cuda:0'), covar=tensor([0.0201, 0.0378, 0.0519, 0.0405, 0.0215, 0.0311, 0.0208, 0.0236], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0219, 0.0213, 0.0212, 0.0217, 0.0218, 0.0219, 0.0210], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 10:59:47,961 INFO [train.py:904] (0/8) Epoch 17, batch 100, loss[loss=0.1524, simple_loss=0.2396, pruned_loss=0.03258, over 16831.00 frames. ], tot_loss[loss=0.1853, simple_loss=0.2703, pruned_loss=0.05008, over 1314800.16 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:00:00,032 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:00:26,390 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.355e+02 2.682e+02 3.198e+02 6.473e+02, threshold=5.364e+02, percent-clipped=1.0 2023-04-30 11:00:36,869 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2335, 5.2321, 5.0018, 4.5296, 5.0308, 1.8827, 4.8160, 5.0119], device='cuda:0'), covar=tensor([0.0100, 0.0086, 0.0199, 0.0353, 0.0112, 0.2557, 0.0136, 0.0178], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0131, 0.0176, 0.0159, 0.0151, 0.0192, 0.0164, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:00:56,556 INFO [train.py:904] (0/8) Epoch 17, batch 150, loss[loss=0.2344, simple_loss=0.3114, pruned_loss=0.07867, over 11923.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2683, pruned_loss=0.04922, over 1755981.62 frames. ], batch size: 246, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:01:23,603 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162572.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 11:01:44,471 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:02:05,938 INFO [train.py:904] (0/8) Epoch 17, batch 200, loss[loss=0.177, simple_loss=0.2735, pruned_loss=0.04022, over 17269.00 frames. ], tot_loss[loss=0.1843, simple_loss=0.2691, pruned_loss=0.04982, over 2091991.07 frames. ], batch size: 52, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:02:42,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7297, 4.5381, 4.7765, 4.9460, 5.0855, 4.5537, 5.0720, 5.0890], device='cuda:0'), covar=tensor([0.1697, 0.1238, 0.1607, 0.0714, 0.0564, 0.0895, 0.0664, 0.0659], device='cuda:0'), in_proj_covar=tensor([0.0571, 0.0706, 0.0833, 0.0712, 0.0539, 0.0563, 0.0578, 0.0669], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:02:43,592 INFO [optim.py:368] (0/8) 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 11:03:12,323 INFO [train.py:904] (0/8) Epoch 17, batch 250, loss[loss=0.1955, simple_loss=0.2649, pruned_loss=0.06304, over 16900.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2675, pruned_loss=0.04952, over 2375079.43 frames. ], batch size: 109, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:03:22,619 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 11:03:39,269 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8500, 4.1880, 3.2051, 2.3644, 2.7267, 2.5843, 4.4149, 3.6197], device='cuda:0'), covar=tensor([0.2673, 0.0546, 0.1524, 0.2658, 0.2664, 0.1937, 0.0381, 0.1227], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0255, 0.0289, 0.0292, 0.0276, 0.0236, 0.0276, 0.0311], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:04:20,336 INFO [train.py:904] (0/8) Epoch 17, batch 300, loss[loss=0.1834, simple_loss=0.2763, pruned_loss=0.04522, over 17123.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2649, pruned_loss=0.04829, over 2571611.57 frames. ], batch size: 53, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:04:27,593 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-04-30 11:04:57,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8631, 2.8714, 2.8451, 4.8036, 3.8418, 4.3915, 1.6738, 3.1394], device='cuda:0'), covar=tensor([0.1352, 0.0755, 0.1125, 0.0258, 0.0243, 0.0411, 0.1611, 0.0804], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0164, 0.0186, 0.0167, 0.0194, 0.0209, 0.0193, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:04:59,738 INFO [optim.py:368] (0/8) 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 11:05:02,689 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8476, 3.9988, 3.0394, 2.3017, 2.4984, 2.3897, 4.1314, 3.4698], device='cuda:0'), covar=tensor([0.2562, 0.0550, 0.1622, 0.2731, 0.2778, 0.2023, 0.0452, 0.1295], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0257, 0.0290, 0.0293, 0.0278, 0.0237, 0.0277, 0.0313], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:05:03,702 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162733.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:05:29,589 INFO [train.py:904] (0/8) Epoch 17, batch 350, loss[loss=0.1627, simple_loss=0.2475, pruned_loss=0.03894, over 16510.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2625, pruned_loss=0.04681, over 2741117.76 frames. ], batch size: 68, lr: 4.05e-03, grad_scale: 2.0 2023-04-30 11:06:07,974 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162781.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:06:36,770 INFO [train.py:904] (0/8) Epoch 17, batch 400, loss[loss=0.1596, simple_loss=0.2539, pruned_loss=0.03269, over 17180.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2612, pruned_loss=0.04636, over 2874604.12 frames. ], batch size: 46, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:07:11,017 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=162827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:07:15,901 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.145e+02 2.685e+02 3.328e+02 1.079e+03, threshold=5.370e+02, percent-clipped=6.0 2023-04-30 11:07:46,583 INFO [train.py:904] (0/8) Epoch 17, batch 450, loss[loss=0.1589, simple_loss=0.2509, pruned_loss=0.03347, over 16842.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2588, pruned_loss=0.04554, over 2982793.64 frames. ], batch size: 42, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:08:06,783 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=162867.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:08:34,227 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=162887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:35,573 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=162888.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:08:55,344 INFO [train.py:904] (0/8) Epoch 17, batch 500, loss[loss=0.1562, simple_loss=0.2437, pruned_loss=0.03436, over 17175.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2574, pruned_loss=0.04438, over 3051606.69 frames. ], batch size: 44, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:09:32,596 INFO [optim.py:368] (0/8) 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 11:09:39,211 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=162935.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:10:01,805 INFO [train.py:904] (0/8) Epoch 17, batch 550, loss[loss=0.1624, simple_loss=0.2611, pruned_loss=0.03183, over 17122.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2573, pruned_loss=0.04427, over 3114505.35 frames. ], batch size: 49, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:10,230 INFO [train.py:904] (0/8) Epoch 17, batch 600, loss[loss=0.1792, simple_loss=0.2444, pruned_loss=0.057, over 16898.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2569, pruned_loss=0.04399, over 3170059.31 frames. ], batch size: 116, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:11:47,298 INFO [optim.py:368] (0/8) 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 11:12:09,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0705, 5.6527, 5.8017, 5.5078, 5.6655, 6.2268, 5.7267, 5.4558], device='cuda:0'), covar=tensor([0.1006, 0.2274, 0.2371, 0.2135, 0.2764, 0.0963, 0.1493, 0.2274], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0564, 0.0619, 0.0474, 0.0632, 0.0652, 0.0489, 0.0634], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:12:16,985 INFO [train.py:904] (0/8) Epoch 17, batch 650, loss[loss=0.1657, simple_loss=0.2646, pruned_loss=0.03346, over 17138.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.256, pruned_loss=0.04322, over 3207450.08 frames. ], batch size: 47, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:13:09,344 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:13:20,296 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 11:13:25,545 INFO [train.py:904] (0/8) Epoch 17, batch 700, loss[loss=0.1744, simple_loss=0.25, pruned_loss=0.04946, over 16868.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2554, pruned_loss=0.04308, over 3238992.34 frames. ], batch size: 96, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:13:35,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4777, 2.2353, 2.3474, 4.3171, 2.2268, 2.6091, 2.3445, 2.4399], device='cuda:0'), covar=tensor([0.1125, 0.3633, 0.2862, 0.0490, 0.4026, 0.2552, 0.3336, 0.3547], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0420, 0.0354, 0.0322, 0.0427, 0.0483, 0.0392, 0.0492], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:13:36,856 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7997, 2.7318, 2.4374, 2.7316, 3.1331, 2.9009, 3.5712, 3.3304], device='cuda:0'), covar=tensor([0.0122, 0.0430, 0.0457, 0.0396, 0.0262, 0.0350, 0.0193, 0.0224], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0228, 0.0220, 0.0219, 0.0227, 0.0228, 0.0231, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:14:04,489 INFO [optim.py:368] (0/8) 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 11:14:34,362 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 17, batch 750, loss[loss=0.151, simple_loss=0.2458, pruned_loss=0.02808, over 17286.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.256, pruned_loss=0.04331, over 3260005.48 frames. ], batch size: 52, lr: 4.05e-03, grad_scale: 4.0 2023-04-30 11:14:57,140 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163167.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:15:05,452 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 11:15:18,073 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163183.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:15:36,488 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-04-30 11:15:44,401 INFO [train.py:904] (0/8) Epoch 17, batch 800, loss[loss=0.1885, simple_loss=0.2671, pruned_loss=0.05492, over 15502.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2561, pruned_loss=0.04303, over 3278033.56 frames. ], batch size: 190, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:16:03,292 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163215.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:23,808 INFO [optim.py:368] (0/8) 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 11:16:36,516 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=163240.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:16:36,776 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.52 vs. limit=5.0 2023-04-30 11:16:53,803 INFO [train.py:904] (0/8) Epoch 17, batch 850, loss[loss=0.182, simple_loss=0.2611, pruned_loss=0.05146, over 16342.00 frames. ], tot_loss[loss=0.17, simple_loss=0.255, pruned_loss=0.04245, over 3294400.69 frames. ], batch size: 165, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:17:01,572 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1323, 5.7584, 5.8310, 5.6395, 5.7380, 6.2502, 5.8113, 5.5689], device='cuda:0'), covar=tensor([0.0911, 0.2197, 0.2224, 0.2114, 0.2471, 0.0950, 0.1330, 0.2242], device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0562, 0.0621, 0.0473, 0.0631, 0.0655, 0.0490, 0.0634], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:18:01,109 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 17, batch 900, loss[loss=0.1772, simple_loss=0.2539, pruned_loss=0.05029, over 16734.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2533, pruned_loss=0.04215, over 3305080.80 frames. ], batch size: 124, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:18:26,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7970, 3.9171, 4.1217, 4.1007, 4.1241, 3.8869, 3.9313, 3.8823], device='cuda:0'), covar=tensor([0.0390, 0.0604, 0.0401, 0.0406, 0.0472, 0.0445, 0.0705, 0.0509], device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0418, 0.0408, 0.0383, 0.0456, 0.0431, 0.0525, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 11:18:40,400 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.550e+02 2.201e+02 2.666e+02 3.227e+02 4.456e+02, threshold=5.332e+02, percent-clipped=0.0 2023-04-30 11:18:42,619 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4298, 2.2129, 1.7367, 2.0398, 2.6360, 2.3921, 2.5974, 2.7178], device='cuda:0'), covar=tensor([0.0213, 0.0365, 0.0507, 0.0425, 0.0209, 0.0332, 0.0205, 0.0258], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0230, 0.0221, 0.0220, 0.0228, 0.0230, 0.0234, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:18:52,660 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 11:19:09,573 INFO [train.py:904] (0/8) Epoch 17, batch 950, loss[loss=0.1687, simple_loss=0.2444, pruned_loss=0.04646, over 12633.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2538, pruned_loss=0.04255, over 3303638.64 frames. ], batch size: 246, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:19:33,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2741, 4.2729, 4.6587, 4.6564, 4.6662, 4.3693, 4.4181, 4.2338], device='cuda:0'), covar=tensor([0.0440, 0.0880, 0.0531, 0.0648, 0.0583, 0.0618, 0.1024, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0419, 0.0411, 0.0385, 0.0459, 0.0432, 0.0528, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 11:20:00,163 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7874, 3.9137, 2.4644, 4.5639, 3.0500, 4.5475, 2.5875, 3.1789], device='cuda:0'), covar=tensor([0.0265, 0.0359, 0.1514, 0.0251, 0.0814, 0.0482, 0.1451, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0171, 0.0192, 0.0152, 0.0173, 0.0212, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:20:13,034 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 11:20:17,962 INFO [train.py:904] (0/8) Epoch 17, batch 1000, loss[loss=0.1593, simple_loss=0.236, pruned_loss=0.04125, over 16387.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2526, pruned_loss=0.04278, over 3312996.88 frames. ], batch size: 165, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:20:54,827 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.252e+02 2.741e+02 3.139e+02 5.661e+02, threshold=5.483e+02, percent-clipped=2.0 2023-04-30 11:20:59,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5628, 3.3076, 2.7540, 2.2356, 2.2763, 2.2799, 3.4169, 3.0616], device='cuda:0'), covar=tensor([0.2486, 0.0706, 0.1566, 0.2545, 0.2616, 0.1961, 0.0505, 0.1370], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0262, 0.0295, 0.0298, 0.0286, 0.0241, 0.0281, 0.0320], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:21:18,946 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:21:26,459 INFO [train.py:904] (0/8) Epoch 17, batch 1050, loss[loss=0.1481, simple_loss=0.2336, pruned_loss=0.03133, over 17232.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2518, pruned_loss=0.04272, over 3305185.50 frames. ], batch size: 45, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:21:31,419 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5625, 1.8622, 2.2467, 2.3899, 2.5663, 2.5069, 1.8776, 2.5847], device='cuda:0'), covar=tensor([0.0180, 0.0388, 0.0251, 0.0268, 0.0240, 0.0234, 0.0452, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0187, 0.0174, 0.0178, 0.0188, 0.0145, 0.0188, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:22:10,671 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163483.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:22:36,966 INFO [train.py:904] (0/8) Epoch 17, batch 1100, loss[loss=0.1795, simple_loss=0.2663, pruned_loss=0.04632, over 17035.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2514, pruned_loss=0.04226, over 3312917.38 frames. ], batch size: 53, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:23:14,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1545, 2.1348, 2.3011, 3.9318, 2.1198, 2.4875, 2.1982, 2.2594], device='cuda:0'), covar=tensor([0.1472, 0.3875, 0.2939, 0.0610, 0.4065, 0.2586, 0.3962, 0.3048], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0425, 0.0357, 0.0325, 0.0430, 0.0489, 0.0396, 0.0497], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:23:14,826 INFO [optim.py:368] (0/8) 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 11:23:16,189 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:23:43,933 INFO [train.py:904] (0/8) Epoch 17, batch 1150, loss[loss=0.184, simple_loss=0.2538, pruned_loss=0.05714, over 16890.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2508, pruned_loss=0.04192, over 3308026.34 frames. ], batch size: 109, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:24:27,955 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6275, 2.5759, 2.1007, 2.4688, 2.9389, 2.7626, 3.2876, 3.1913], device='cuda:0'), covar=tensor([0.0148, 0.0410, 0.0513, 0.0433, 0.0279, 0.0352, 0.0256, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0230, 0.0222, 0.0220, 0.0229, 0.0230, 0.0234, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:24:43,118 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=163596.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:24:52,268 INFO [train.py:904] (0/8) Epoch 17, batch 1200, loss[loss=0.1489, simple_loss=0.2298, pruned_loss=0.03397, over 16829.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2507, pruned_loss=0.04157, over 3312217.89 frames. ], batch size: 102, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:25:06,980 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5450, 3.7896, 4.0677, 2.3105, 3.2678, 2.5870, 3.9975, 3.8635], device='cuda:0'), covar=tensor([0.0258, 0.0797, 0.0466, 0.1804, 0.0753, 0.0889, 0.0605, 0.1055], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0150, 0.0141, 0.0127, 0.0142, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:25:29,957 INFO [optim.py:368] (0/8) 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 11:25:58,590 INFO [train.py:904] (0/8) Epoch 17, batch 1250, loss[loss=0.1565, simple_loss=0.2324, pruned_loss=0.04026, over 16381.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.251, pruned_loss=0.04214, over 3310908.43 frames. ], batch size: 146, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:06,152 INFO [train.py:904] (0/8) Epoch 17, batch 1300, loss[loss=0.1874, simple_loss=0.2625, pruned_loss=0.05615, over 16874.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.251, pruned_loss=0.0423, over 3308689.69 frames. ], batch size: 116, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:27:44,997 INFO [optim.py:368] (0/8) 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 11:28:08,401 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:28:16,452 INFO [train.py:904] (0/8) Epoch 17, batch 1350, loss[loss=0.1648, simple_loss=0.2454, pruned_loss=0.04208, over 16853.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2515, pruned_loss=0.04234, over 3314050.82 frames. ], batch size: 102, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:28:48,639 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 11:29:15,150 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:29:25,103 INFO [train.py:904] (0/8) Epoch 17, batch 1400, loss[loss=0.1441, simple_loss=0.2257, pruned_loss=0.03129, over 16754.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.252, pruned_loss=0.04245, over 3314407.92 frames. ], batch size: 39, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:29:52,538 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7824, 3.9986, 3.0406, 2.2828, 2.5755, 2.4462, 4.1431, 3.4914], device='cuda:0'), covar=tensor([0.2571, 0.0531, 0.1582, 0.2900, 0.2842, 0.1952, 0.0434, 0.1308], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0263, 0.0296, 0.0298, 0.0287, 0.0243, 0.0282, 0.0322], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:30:05,129 INFO [optim.py:368] (0/8) 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 11:30:36,601 INFO [train.py:904] (0/8) Epoch 17, batch 1450, loss[loss=0.1503, simple_loss=0.2466, pruned_loss=0.027, over 17118.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2509, pruned_loss=0.0422, over 3309273.23 frames. ], batch size: 47, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:31:26,373 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4648, 4.3594, 4.3843, 4.1100, 4.0708, 4.4247, 4.1793, 4.1800], device='cuda:0'), covar=tensor([0.0587, 0.0740, 0.0315, 0.0292, 0.0926, 0.0456, 0.0651, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0395, 0.0331, 0.0322, 0.0344, 0.0373, 0.0228, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:31:38,103 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=163896.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:31:45,965 INFO [train.py:904] (0/8) Epoch 17, batch 1500, loss[loss=0.1785, simple_loss=0.2767, pruned_loss=0.04012, over 16604.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2512, pruned_loss=0.04239, over 3310168.05 frames. ], batch size: 57, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:32:24,621 INFO [optim.py:368] (0/8) 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 11:32:45,104 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=163944.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:32:56,212 INFO [train.py:904] (0/8) Epoch 17, batch 1550, loss[loss=0.1703, simple_loss=0.256, pruned_loss=0.04227, over 17163.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2533, pruned_loss=0.04382, over 3290162.95 frames. ], batch size: 46, lr: 4.04e-03, grad_scale: 8.0 2023-04-30 11:33:00,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4378, 2.2931, 2.3293, 4.3309, 2.3041, 2.6638, 2.3618, 2.4309], device='cuda:0'), covar=tensor([0.1196, 0.3276, 0.2841, 0.0470, 0.3903, 0.2526, 0.3229, 0.3469], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0426, 0.0360, 0.0328, 0.0432, 0.0493, 0.0397, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:33:14,914 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 11:34:01,115 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-164000.pt 2023-04-30 11:34:07,076 INFO [train.py:904] (0/8) Epoch 17, batch 1600, loss[loss=0.1538, simple_loss=0.2381, pruned_loss=0.03478, over 16321.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2554, pruned_loss=0.04412, over 3297662.43 frames. ], batch size: 36, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:34:45,108 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.680e+02 2.274e+02 2.815e+02 3.289e+02 5.501e+02, threshold=5.629e+02, percent-clipped=1.0 2023-04-30 11:35:10,271 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5829, 2.2470, 1.6104, 1.9155, 2.6779, 2.4334, 2.8026, 2.7399], device='cuda:0'), covar=tensor([0.0235, 0.0472, 0.0702, 0.0601, 0.0295, 0.0404, 0.0282, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0232, 0.0222, 0.0222, 0.0231, 0.0232, 0.0237, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:35:15,608 INFO [train.py:904] (0/8) Epoch 17, batch 1650, loss[loss=0.1507, simple_loss=0.2318, pruned_loss=0.03475, over 16785.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2568, pruned_loss=0.04471, over 3298584.78 frames. ], batch size: 39, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:35:39,287 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8788, 4.4103, 4.5047, 3.2155, 3.7690, 4.4223, 3.9978, 2.6688], device='cuda:0'), covar=tensor([0.0447, 0.0064, 0.0032, 0.0314, 0.0109, 0.0077, 0.0073, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0079, 0.0078, 0.0133, 0.0091, 0.0102, 0.0090, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:36:08,299 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:36:24,919 INFO [train.py:904] (0/8) Epoch 17, batch 1700, loss[loss=0.1746, simple_loss=0.274, pruned_loss=0.03756, over 17066.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2584, pruned_loss=0.04493, over 3301618.97 frames. ], batch size: 50, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:36:40,701 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8747, 4.9759, 5.4549, 5.3513, 5.3734, 5.0110, 4.9491, 4.7435], device='cuda:0'), covar=tensor([0.0357, 0.0505, 0.0377, 0.0498, 0.0510, 0.0393, 0.0935, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0425, 0.0415, 0.0390, 0.0462, 0.0438, 0.0535, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 11:36:50,163 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3184, 5.1952, 5.1423, 4.6015, 4.6785, 5.1609, 5.2017, 4.7790], device='cuda:0'), covar=tensor([0.0561, 0.0532, 0.0322, 0.0366, 0.1290, 0.0485, 0.0301, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0400, 0.0335, 0.0328, 0.0350, 0.0378, 0.0230, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:37:01,956 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.646e+02 2.209e+02 2.796e+02 3.472e+02 6.333e+02, threshold=5.591e+02, percent-clipped=2.0 2023-04-30 11:37:02,658 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3596, 1.9998, 2.1970, 4.0240, 1.9447, 2.3184, 2.1111, 2.1645], device='cuda:0'), covar=tensor([0.1365, 0.4342, 0.3154, 0.0626, 0.5350, 0.3444, 0.3924, 0.4649], device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0427, 0.0360, 0.0327, 0.0433, 0.0493, 0.0397, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:37:06,476 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3033, 3.2601, 3.4843, 1.8768, 3.5873, 3.5812, 2.8848, 2.7219], device='cuda:0'), covar=tensor([0.0770, 0.0217, 0.0161, 0.1105, 0.0084, 0.0197, 0.0417, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0107, 0.0094, 0.0139, 0.0075, 0.0121, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 11:37:13,413 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9905, 4.1515, 4.4462, 2.1473, 4.6299, 4.7185, 3.3581, 3.6279], device='cuda:0'), covar=tensor([0.0734, 0.0209, 0.0197, 0.1210, 0.0073, 0.0159, 0.0408, 0.0369], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0107, 0.0094, 0.0139, 0.0075, 0.0121, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 11:37:31,637 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:37:32,935 INFO [train.py:904] (0/8) Epoch 17, batch 1750, loss[loss=0.1571, simple_loss=0.2392, pruned_loss=0.03747, over 16918.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2585, pruned_loss=0.04422, over 3306794.97 frames. ], batch size: 96, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:37:36,764 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:38:05,317 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 11:38:37,779 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6000, 6.0127, 5.7171, 5.8206, 5.4138, 5.3055, 5.4678, 6.0920], device='cuda:0'), covar=tensor([0.1331, 0.0902, 0.1010, 0.0830, 0.0846, 0.0730, 0.1268, 0.0910], device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0792, 0.0647, 0.0588, 0.0503, 0.0506, 0.0660, 0.0613], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:38:41,947 INFO [train.py:904] (0/8) Epoch 17, batch 1800, loss[loss=0.1544, simple_loss=0.2429, pruned_loss=0.03299, over 17215.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2595, pruned_loss=0.04366, over 3315777.75 frames. ], batch size: 44, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:38:57,647 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7172, 3.8650, 3.9568, 2.7660, 3.5492, 3.9735, 3.7256, 2.0361], device='cuda:0'), covar=tensor([0.0469, 0.0261, 0.0067, 0.0403, 0.0119, 0.0131, 0.0103, 0.0588], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0078, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:39:00,467 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5270, 3.6374, 2.1591, 3.7611, 2.7044, 3.7481, 2.1851, 2.8452], device='cuda:0'), covar=tensor([0.0235, 0.0357, 0.1442, 0.0305, 0.0724, 0.0766, 0.1328, 0.0664], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0157, 0.0175, 0.0218, 0.0204, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:39:01,637 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:39:19,769 INFO [optim.py:368] (0/8) 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 11:39:49,967 INFO [train.py:904] (0/8) Epoch 17, batch 1850, loss[loss=0.1878, simple_loss=0.2745, pruned_loss=0.05055, over 11863.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2601, pruned_loss=0.04363, over 3317277.29 frames. ], batch size: 246, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:01,367 INFO [train.py:904] (0/8) Epoch 17, batch 1900, loss[loss=0.1606, simple_loss=0.2435, pruned_loss=0.0388, over 16745.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2595, pruned_loss=0.04338, over 3313942.44 frames. ], batch size: 89, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:41:06,167 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 11:41:26,307 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8551, 2.7961, 2.3967, 2.7191, 3.1728, 2.9126, 3.5142, 3.3426], device='cuda:0'), covar=tensor([0.0106, 0.0350, 0.0431, 0.0376, 0.0223, 0.0318, 0.0212, 0.0224], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0231, 0.0222, 0.0221, 0.0230, 0.0232, 0.0236, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:41:39,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6619, 2.5572, 2.0385, 2.4472, 2.9892, 2.7094, 3.2848, 3.1923], device='cuda:0'), covar=tensor([0.0129, 0.0439, 0.0548, 0.0473, 0.0269, 0.0399, 0.0269, 0.0262], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0230, 0.0221, 0.0221, 0.0230, 0.0231, 0.0236, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:41:41,209 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.647e+02 2.257e+02 2.647e+02 3.150e+02 4.785e+02, threshold=5.294e+02, percent-clipped=0.0 2023-04-30 11:42:08,489 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9932, 2.5342, 2.0551, 2.2711, 2.9792, 2.6841, 3.0118, 3.0088], device='cuda:0'), covar=tensor([0.0204, 0.0336, 0.0453, 0.0394, 0.0193, 0.0300, 0.0218, 0.0240], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0230, 0.0221, 0.0221, 0.0230, 0.0231, 0.0236, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:42:12,320 INFO [train.py:904] (0/8) Epoch 17, batch 1950, loss[loss=0.1727, simple_loss=0.2599, pruned_loss=0.04274, over 17220.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2603, pruned_loss=0.04357, over 3315821.11 frames. ], batch size: 45, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:42:24,719 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:42:47,944 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-04-30 11:42:56,731 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2561, 3.0918, 3.3331, 1.7993, 3.4705, 3.4298, 2.8313, 2.5983], device='cuda:0'), covar=tensor([0.0769, 0.0224, 0.0208, 0.1111, 0.0105, 0.0223, 0.0442, 0.0427], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0108, 0.0094, 0.0140, 0.0075, 0.0121, 0.0127, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 11:43:23,962 INFO [train.py:904] (0/8) Epoch 17, batch 2000, loss[loss=0.195, simple_loss=0.2857, pruned_loss=0.05215, over 11954.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2591, pruned_loss=0.04307, over 3316575.67 frames. ], batch size: 249, lr: 4.03e-03, grad_scale: 8.0 2023-04-30 11:43:51,006 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:44:02,065 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.498e+02 2.950e+02 3.427e+02 6.648e+02, threshold=5.900e+02, percent-clipped=3.0 2023-04-30 11:44:25,528 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164446.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:44:32,479 INFO [train.py:904] (0/8) Epoch 17, batch 2050, loss[loss=0.1703, simple_loss=0.2516, pruned_loss=0.0445, over 16797.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2595, pruned_loss=0.04338, over 3321896.76 frames. ], batch size: 42, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:44:47,282 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7587, 3.9366, 2.5836, 4.6687, 3.0376, 4.6119, 2.7630, 3.2959], device='cuda:0'), covar=tensor([0.0309, 0.0385, 0.1483, 0.0224, 0.0822, 0.0415, 0.1271, 0.0687], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0157, 0.0176, 0.0218, 0.0204, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:44:52,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6327, 2.3460, 2.2451, 4.6356, 2.3169, 2.7527, 2.4045, 2.5509], device='cuda:0'), covar=tensor([0.1100, 0.3669, 0.3055, 0.0380, 0.4027, 0.2528, 0.3382, 0.3505], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0428, 0.0358, 0.0326, 0.0431, 0.0493, 0.0397, 0.0500], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:45:06,988 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-04-30 11:45:41,560 INFO [train.py:904] (0/8) Epoch 17, batch 2100, loss[loss=0.1945, simple_loss=0.2707, pruned_loss=0.05921, over 16689.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2607, pruned_loss=0.04473, over 3308979.34 frames. ], batch size: 134, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:45:54,972 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164511.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:46:13,724 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6316, 4.6577, 4.8353, 4.6829, 4.6629, 5.2978, 4.8100, 4.5315], device='cuda:0'), covar=tensor([0.1526, 0.1984, 0.2453, 0.2074, 0.2926, 0.1084, 0.1697, 0.2656], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0570, 0.0630, 0.0483, 0.0651, 0.0665, 0.0497, 0.0646], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:46:20,656 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.320e+02 2.783e+02 3.356e+02 6.919e+02, threshold=5.567e+02, percent-clipped=1.0 2023-04-30 11:46:50,972 INFO [train.py:904] (0/8) Epoch 17, batch 2150, loss[loss=0.1588, simple_loss=0.2453, pruned_loss=0.03621, over 17214.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2619, pruned_loss=0.04576, over 3312880.84 frames. ], batch size: 43, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:47:13,588 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164568.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:27,025 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:47:58,410 INFO [train.py:904] (0/8) Epoch 17, batch 2200, loss[loss=0.1478, simple_loss=0.2386, pruned_loss=0.02851, over 17191.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2618, pruned_loss=0.04598, over 3311902.64 frames. ], batch size: 44, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:48:36,608 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164629.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:48:37,256 INFO [optim.py:368] (0/8) 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 11:48:44,211 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0903, 3.1680, 3.3493, 2.4174, 3.0452, 3.3962, 3.1291, 1.9494], device='cuda:0'), covar=tensor([0.0481, 0.0114, 0.0058, 0.0342, 0.0100, 0.0093, 0.0096, 0.0449], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0077, 0.0077, 0.0132, 0.0091, 0.0101, 0.0089, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 11:48:44,445 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.91 vs. limit=5.0 2023-04-30 11:48:50,038 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164639.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:48:51,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.54 vs. limit=2.0 2023-04-30 11:48:57,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9783, 3.9562, 4.4363, 2.0707, 4.6431, 4.6903, 3.1764, 3.5601], device='cuda:0'), covar=tensor([0.0703, 0.0227, 0.0192, 0.1178, 0.0055, 0.0151, 0.0439, 0.0369], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0108, 0.0095, 0.0139, 0.0075, 0.0122, 0.0127, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 11:49:02,801 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 11:49:06,813 INFO [train.py:904] (0/8) Epoch 17, batch 2250, loss[loss=0.1727, simple_loss=0.2535, pruned_loss=0.04597, over 16901.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2625, pruned_loss=0.04609, over 3317155.67 frames. ], batch size: 116, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:49:15,059 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164658.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:15,838 INFO [train.py:904] (0/8) Epoch 17, batch 2300, loss[loss=0.156, simple_loss=0.2459, pruned_loss=0.03306, over 16152.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2622, pruned_loss=0.04567, over 3326583.26 frames. ], batch size: 35, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:50:34,578 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164716.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:38,839 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:50:53,239 INFO [optim.py:368] (0/8) 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 11:51:16,723 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:16,862 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=164746.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:51:23,998 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5692, 2.3037, 2.2693, 4.5393, 2.2110, 2.7546, 2.3926, 2.4468], device='cuda:0'), covar=tensor([0.1199, 0.3774, 0.3127, 0.0423, 0.4168, 0.2639, 0.3482, 0.3697], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0429, 0.0359, 0.0327, 0.0431, 0.0496, 0.0398, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:51:24,582 INFO [train.py:904] (0/8) Epoch 17, batch 2350, loss[loss=0.1627, simple_loss=0.2516, pruned_loss=0.03691, over 17135.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2619, pruned_loss=0.04582, over 3328839.10 frames. ], batch size: 48, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:23,522 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164794.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:24,331 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 11:52:34,383 INFO [train.py:904] (0/8) Epoch 17, batch 2400, loss[loss=0.1762, simple_loss=0.2541, pruned_loss=0.04921, over 16468.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2632, pruned_loss=0.04605, over 3332977.59 frames. ], batch size: 75, lr: 4.03e-03, grad_scale: 16.0 2023-04-30 11:52:41,593 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=164807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:52:46,687 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=164811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:53:12,675 INFO [optim.py:368] (0/8) 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 11:53:15,544 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8247, 2.9737, 2.8463, 5.0457, 4.1585, 4.4144, 1.6098, 3.2986], device='cuda:0'), covar=tensor([0.1320, 0.0732, 0.1085, 0.0191, 0.0224, 0.0377, 0.1572, 0.0712], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0164, 0.0185, 0.0172, 0.0198, 0.0210, 0.0190, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:53:19,585 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9372, 4.0616, 2.4861, 4.7257, 3.2794, 4.7661, 2.7682, 3.4699], device='cuda:0'), covar=tensor([0.0299, 0.0337, 0.1518, 0.0252, 0.0744, 0.0362, 0.1297, 0.0627], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0174, 0.0193, 0.0157, 0.0173, 0.0217, 0.0202, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:53:41,596 INFO [train.py:904] (0/8) Epoch 17, batch 2450, loss[loss=0.1636, simple_loss=0.2514, pruned_loss=0.03788, over 17211.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.264, pruned_loss=0.04616, over 3326942.40 frames. ], batch size: 44, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:53:46,647 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1188, 5.4661, 5.2122, 5.2318, 4.9220, 4.8121, 4.8879, 5.5691], device='cuda:0'), covar=tensor([0.1131, 0.0809, 0.0934, 0.0791, 0.0833, 0.0949, 0.1037, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0797, 0.0649, 0.0593, 0.0506, 0.0509, 0.0662, 0.0618], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:53:51,213 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=164859.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:54:46,730 INFO [train.py:904] (0/8) Epoch 17, batch 2500, loss[loss=0.2065, simple_loss=0.2905, pruned_loss=0.06124, over 15384.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2629, pruned_loss=0.04549, over 3326444.53 frames. ], batch size: 190, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:55:11,525 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0963, 2.1387, 2.3035, 3.8437, 2.1297, 2.4281, 2.2088, 2.2629], device='cuda:0'), covar=tensor([0.1358, 0.3644, 0.2736, 0.0596, 0.3829, 0.2652, 0.3652, 0.3264], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0427, 0.0357, 0.0326, 0.0429, 0.0494, 0.0396, 0.0500], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:55:17,587 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:55:26,643 INFO [optim.py:368] (0/8) 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 11:55:29,664 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 11:55:30,306 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=164934.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 11:55:55,243 INFO [train.py:904] (0/8) Epoch 17, batch 2550, loss[loss=0.2022, simple_loss=0.3023, pruned_loss=0.05106, over 16630.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2634, pruned_loss=0.04547, over 3327791.39 frames. ], batch size: 62, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:02,095 INFO [train.py:904] (0/8) Epoch 17, batch 2600, loss[loss=0.1776, simple_loss=0.2645, pruned_loss=0.04533, over 16801.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.263, pruned_loss=0.04514, over 3330973.87 frames. ], batch size: 83, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:57:17,261 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165014.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:20,590 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165016.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:57:31,714 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9761, 2.2926, 2.2962, 2.7389, 2.1526, 3.2009, 1.7336, 2.7116], device='cuda:0'), covar=tensor([0.1097, 0.0659, 0.1041, 0.0177, 0.0158, 0.0391, 0.1357, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0166, 0.0187, 0.0174, 0.0199, 0.0212, 0.0192, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 11:57:41,410 INFO [optim.py:368] (0/8) 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 11:57:43,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5285, 5.9421, 5.5935, 5.7274, 5.2747, 5.2600, 5.3285, 6.0313], device='cuda:0'), covar=tensor([0.1391, 0.0963, 0.1118, 0.0851, 0.1012, 0.0770, 0.1191, 0.0982], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0800, 0.0651, 0.0594, 0.0508, 0.0509, 0.0663, 0.0618], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 11:58:08,471 INFO [train.py:904] (0/8) Epoch 17, batch 2650, loss[loss=0.1565, simple_loss=0.2406, pruned_loss=0.03617, over 16966.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2639, pruned_loss=0.04503, over 3332040.23 frames. ], batch size: 41, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:58:26,259 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:58:29,408 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-04-30 11:59:18,037 INFO [train.py:904] (0/8) Epoch 17, batch 2700, loss[loss=0.1634, simple_loss=0.2596, pruned_loss=0.03361, over 17168.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2646, pruned_loss=0.04423, over 3336287.20 frames. ], batch size: 46, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 11:59:18,381 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:47,071 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 11:59:47,970 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 11:59:59,035 INFO [optim.py:368] (0/8) 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 12:00:28,484 INFO [train.py:904] (0/8) Epoch 17, batch 2750, loss[loss=0.1901, simple_loss=0.2858, pruned_loss=0.04722, over 17104.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2639, pruned_loss=0.04369, over 3330415.98 frames. ], batch size: 53, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:01:12,642 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165184.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:01:38,401 INFO [train.py:904] (0/8) Epoch 17, batch 2800, loss[loss=0.1723, simple_loss=0.2563, pruned_loss=0.04408, over 16405.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2634, pruned_loss=0.04411, over 3329332.00 frames. ], batch size: 146, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:02:09,589 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165224.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:02:19,322 INFO [optim.py:368] (0/8) 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 12:02:24,065 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165234.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:02:48,651 INFO [train.py:904] (0/8) Epoch 17, batch 2850, loss[loss=0.1809, simple_loss=0.2544, pruned_loss=0.05369, over 16906.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2638, pruned_loss=0.04487, over 3323298.94 frames. ], batch size: 109, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:03:18,274 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165272.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:31,433 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:03:58,715 INFO [train.py:904] (0/8) Epoch 17, batch 2900, loss[loss=0.1471, simple_loss=0.2418, pruned_loss=0.02622, over 17179.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2624, pruned_loss=0.04495, over 3326476.71 frames. ], batch size: 46, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:04:16,925 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165314.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:04:39,536 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.311e+02 2.666e+02 3.214e+02 5.469e+02, threshold=5.331e+02, percent-clipped=1.0 2023-04-30 12:04:55,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.06 vs. limit=5.0 2023-04-30 12:05:09,252 INFO [train.py:904] (0/8) Epoch 17, batch 2950, loss[loss=0.1919, simple_loss=0.2997, pruned_loss=0.04208, over 17105.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2627, pruned_loss=0.04553, over 3314753.07 frames. ], batch size: 49, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:05:23,907 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165362.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:02,017 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8394, 3.8772, 4.2902, 2.0057, 4.5380, 4.5543, 3.2045, 3.4241], device='cuda:0'), covar=tensor([0.0758, 0.0232, 0.0231, 0.1174, 0.0057, 0.0164, 0.0433, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0137, 0.0075, 0.0121, 0.0126, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 12:06:20,626 INFO [train.py:904] (0/8) Epoch 17, batch 3000, loss[loss=0.1504, simple_loss=0.2388, pruned_loss=0.03104, over 17189.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2623, pruned_loss=0.04561, over 3315088.31 frames. ], batch size: 44, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:06:20,627 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 12:06:29,129 INFO [train.py:938] (0/8) Epoch 17, validation: loss=0.1364, simple_loss=0.2422, pruned_loss=0.0153, over 944034.00 frames. 2023-04-30 12:06:29,130 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 12:06:29,407 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165402.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:06:46,734 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1660, 4.4699, 3.4148, 2.6197, 3.2137, 2.8975, 4.9264, 3.9305], device='cuda:0'), covar=tensor([0.2488, 0.0598, 0.1598, 0.2522, 0.2472, 0.1797, 0.0360, 0.1147], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0263, 0.0296, 0.0298, 0.0289, 0.0242, 0.0282, 0.0323], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:06:50,976 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.67 vs. limit=5.0 2023-04-30 12:07:09,776 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.715e+02 2.395e+02 2.828e+02 3.269e+02 8.319e+02, threshold=5.656e+02, percent-clipped=3.0 2023-04-30 12:07:36,653 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165450.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:07:38,659 INFO [train.py:904] (0/8) Epoch 17, batch 3050, loss[loss=0.1887, simple_loss=0.2842, pruned_loss=0.04663, over 17072.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2617, pruned_loss=0.04561, over 3311151.24 frames. ], batch size: 53, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:07:59,189 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1906, 5.6124, 5.7251, 5.4813, 5.5209, 6.1365, 5.6832, 5.4184], device='cuda:0'), covar=tensor([0.0879, 0.2337, 0.2579, 0.1999, 0.2950, 0.1042, 0.1418, 0.2590], device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0578, 0.0637, 0.0489, 0.0657, 0.0671, 0.0503, 0.0655], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:08:12,763 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.59 vs. limit=2.0 2023-04-30 12:08:15,884 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165479.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:08:45,829 INFO [train.py:904] (0/8) Epoch 17, batch 3100, loss[loss=0.1622, simple_loss=0.2361, pruned_loss=0.0442, over 16760.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2604, pruned_loss=0.0448, over 3325229.31 frames. ], batch size: 83, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:09:27,661 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 2.216e+02 2.589e+02 3.162e+02 6.656e+02, threshold=5.179e+02, percent-clipped=4.0 2023-04-30 12:09:55,023 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165551.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:09:55,713 INFO [train.py:904] (0/8) Epoch 17, batch 3150, loss[loss=0.1647, simple_loss=0.2455, pruned_loss=0.04193, over 16740.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.259, pruned_loss=0.04456, over 3321805.68 frames. ], batch size: 83, lr: 4.02e-03, grad_scale: 4.0 2023-04-30 12:10:22,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2550, 3.4983, 3.7450, 2.5842, 3.4276, 3.8043, 3.6572, 2.1500], device='cuda:0'), covar=tensor([0.0497, 0.0189, 0.0062, 0.0371, 0.0104, 0.0101, 0.0069, 0.0467], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0077, 0.0077, 0.0130, 0.0091, 0.0101, 0.0088, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:10:24,799 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1786, 5.2250, 5.7105, 5.6490, 5.6638, 5.2766, 5.2502, 4.9701], device='cuda:0'), covar=tensor([0.0301, 0.0529, 0.0286, 0.0387, 0.0503, 0.0351, 0.0948, 0.0465], device='cuda:0'), in_proj_covar=tensor([0.0395, 0.0431, 0.0418, 0.0395, 0.0468, 0.0445, 0.0546, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 12:11:06,287 INFO [train.py:904] (0/8) Epoch 17, batch 3200, loss[loss=0.1691, simple_loss=0.2706, pruned_loss=0.03382, over 17136.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2583, pruned_loss=0.04438, over 3321931.74 frames. ], batch size: 48, lr: 4.02e-03, grad_scale: 8.0 2023-04-30 12:11:20,403 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165612.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:11:49,642 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.164e+02 2.494e+02 3.096e+02 4.677e+02, threshold=4.988e+02, percent-clipped=0.0 2023-04-30 12:11:50,122 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3476, 5.3269, 5.1021, 4.4968, 5.1729, 2.3461, 4.8594, 5.1496], device='cuda:0'), covar=tensor([0.0081, 0.0073, 0.0171, 0.0393, 0.0089, 0.2190, 0.0127, 0.0156], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0143, 0.0191, 0.0174, 0.0164, 0.0199, 0.0179, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:12:15,440 INFO [train.py:904] (0/8) Epoch 17, batch 3250, loss[loss=0.2136, simple_loss=0.2833, pruned_loss=0.07188, over 16185.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2586, pruned_loss=0.04468, over 3317666.65 frames. ], batch size: 165, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:12:29,808 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9902, 4.5394, 3.2941, 2.4401, 2.9881, 2.6414, 4.7434, 3.9307], device='cuda:0'), covar=tensor([0.2737, 0.0521, 0.1666, 0.2827, 0.2721, 0.2018, 0.0376, 0.1210], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0263, 0.0296, 0.0297, 0.0289, 0.0242, 0.0282, 0.0323], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:13:23,337 INFO [train.py:904] (0/8) Epoch 17, batch 3300, loss[loss=0.1658, simple_loss=0.2641, pruned_loss=0.03374, over 17127.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2596, pruned_loss=0.0448, over 3316161.19 frames. ], batch size: 48, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:14:06,779 INFO [optim.py:368] (0/8) 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 12:14:30,166 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165750.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:14:32,760 INFO [train.py:904] (0/8) Epoch 17, batch 3350, loss[loss=0.1445, simple_loss=0.2305, pruned_loss=0.02926, over 16976.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2603, pruned_loss=0.04464, over 3314644.71 frames. ], batch size: 41, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:11,009 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=165779.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:12,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8776, 1.9316, 2.3907, 2.7499, 2.7732, 2.8954, 2.0074, 3.0629], device='cuda:0'), covar=tensor([0.0164, 0.0431, 0.0302, 0.0222, 0.0259, 0.0223, 0.0457, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0189, 0.0175, 0.0180, 0.0189, 0.0147, 0.0190, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:15:42,361 INFO [train.py:904] (0/8) Epoch 17, batch 3400, loss[loss=0.1629, simple_loss=0.2545, pruned_loss=0.03566, over 17158.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2596, pruned_loss=0.04445, over 3317672.97 frames. ], batch size: 46, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:15:44,563 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=165803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:15:56,237 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165811.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:18,383 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=165827.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:16:27,739 INFO [optim.py:368] (0/8) 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 12:16:28,299 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3279, 3.7238, 3.8015, 2.1294, 3.2206, 2.5406, 3.8775, 3.8923], device='cuda:0'), covar=tensor([0.0311, 0.0820, 0.0532, 0.1946, 0.0772, 0.0963, 0.0636, 0.1032], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0151, 0.0142, 0.0128, 0.0142, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 12:16:54,372 INFO [train.py:904] (0/8) Epoch 17, batch 3450, loss[loss=0.1703, simple_loss=0.2502, pruned_loss=0.0452, over 16584.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.259, pruned_loss=0.04408, over 3312063.65 frames. ], batch size: 68, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:17:03,105 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6302, 2.9170, 2.9725, 1.9743, 2.7103, 2.1579, 3.2582, 3.2489], device='cuda:0'), covar=tensor([0.0278, 0.0926, 0.0591, 0.1974, 0.0846, 0.1031, 0.0617, 0.0898], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0160, 0.0164, 0.0150, 0.0141, 0.0128, 0.0141, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 12:17:11,541 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=165864.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:17:29,024 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8736, 3.9984, 2.6162, 4.7430, 3.2177, 4.6129, 2.7741, 3.2710], device='cuda:0'), covar=tensor([0.0347, 0.0401, 0.1490, 0.0194, 0.0781, 0.0495, 0.1378, 0.0737], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0176, 0.0194, 0.0159, 0.0174, 0.0218, 0.0202, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 12:17:33,055 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3243, 2.0266, 2.1638, 4.1985, 1.9998, 2.3199, 2.1243, 2.1743], device='cuda:0'), covar=tensor([0.1364, 0.4316, 0.3070, 0.0540, 0.5019, 0.3269, 0.3962, 0.4196], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0427, 0.0356, 0.0326, 0.0429, 0.0495, 0.0397, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:18:05,607 INFO [train.py:904] (0/8) Epoch 17, batch 3500, loss[loss=0.1923, simple_loss=0.2709, pruned_loss=0.05685, over 11713.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2575, pruned_loss=0.04401, over 3316494.99 frames. ], batch size: 247, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:18:13,132 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=165907.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:18:33,778 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 12:18:49,909 INFO [optim.py:368] (0/8) 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 12:19:16,046 INFO [train.py:904] (0/8) Epoch 17, batch 3550, loss[loss=0.1522, simple_loss=0.2403, pruned_loss=0.03207, over 17215.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2569, pruned_loss=0.04359, over 3322695.73 frames. ], batch size: 45, lr: 4.01e-03, grad_scale: 4.0 2023-04-30 12:20:22,861 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-166000.pt 2023-04-30 12:20:27,934 INFO [train.py:904] (0/8) Epoch 17, batch 3600, loss[loss=0.1719, simple_loss=0.2662, pruned_loss=0.03885, over 16686.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2558, pruned_loss=0.04318, over 3316045.95 frames. ], batch size: 57, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:20:32,763 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1357, 5.6109, 5.7077, 5.4269, 5.4812, 6.1361, 5.6325, 5.3880], device='cuda:0'), covar=tensor([0.0879, 0.2102, 0.2464, 0.2180, 0.3152, 0.1038, 0.1343, 0.2149], device='cuda:0'), in_proj_covar=tensor([0.0400, 0.0579, 0.0636, 0.0489, 0.0656, 0.0669, 0.0500, 0.0655], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:21:08,710 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-04-30 12:21:12,107 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.087e+02 2.490e+02 2.982e+02 5.501e+02, threshold=4.979e+02, percent-clipped=1.0 2023-04-30 12:21:40,977 INFO [train.py:904] (0/8) Epoch 17, batch 3650, loss[loss=0.1836, simple_loss=0.2556, pruned_loss=0.0558, over 15342.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2544, pruned_loss=0.04329, over 3315050.49 frames. ], batch size: 190, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:21:43,860 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166054.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:21:54,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2920, 4.1053, 4.3158, 4.4290, 4.5438, 4.0966, 4.3096, 4.5249], device='cuda:0'), covar=tensor([0.1538, 0.1203, 0.1343, 0.0744, 0.0634, 0.1377, 0.2850, 0.0742], device='cuda:0'), in_proj_covar=tensor([0.0630, 0.0778, 0.0928, 0.0793, 0.0593, 0.0631, 0.0631, 0.0736], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:22:55,841 INFO [train.py:904] (0/8) Epoch 17, batch 3700, loss[loss=0.1681, simple_loss=0.2386, pruned_loss=0.04877, over 16760.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2532, pruned_loss=0.04485, over 3300165.75 frames. ], batch size: 83, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:23:02,188 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:23:15,493 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166115.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:23:42,389 INFO [optim.py:368] (0/8) 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 12:24:03,612 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166147.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:24:10,336 INFO [train.py:904] (0/8) Epoch 17, batch 3750, loss[loss=0.1827, simple_loss=0.2774, pruned_loss=0.04402, over 17103.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2536, pruned_loss=0.04602, over 3296499.99 frames. ], batch size: 49, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:24:21,443 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:24,426 INFO [train.py:904] (0/8) Epoch 17, batch 3800, loss[loss=0.1798, simple_loss=0.2625, pruned_loss=0.04857, over 16688.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2553, pruned_loss=0.04748, over 3295693.36 frames. ], batch size: 134, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:25:32,334 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:25:33,765 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166208.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:26:10,619 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.250e+02 2.656e+02 3.137e+02 5.830e+02, threshold=5.312e+02, percent-clipped=1.0 2023-04-30 12:26:33,952 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 12:26:38,289 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-04-30 12:26:38,515 INFO [train.py:904] (0/8) Epoch 17, batch 3850, loss[loss=0.1542, simple_loss=0.2358, pruned_loss=0.03631, over 16879.00 frames. ], tot_loss[loss=0.1761, simple_loss=0.2557, pruned_loss=0.04826, over 3292087.71 frames. ], batch size: 96, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:26:44,004 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166255.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:27:02,368 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4744, 3.5702, 2.2120, 3.7239, 2.8276, 3.7728, 2.1933, 2.7362], device='cuda:0'), covar=tensor([0.0264, 0.0331, 0.1425, 0.0249, 0.0637, 0.0573, 0.1342, 0.0699], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0175, 0.0193, 0.0159, 0.0173, 0.0218, 0.0202, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 12:27:15,373 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-04-30 12:27:38,737 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 12:27:52,949 INFO [train.py:904] (0/8) Epoch 17, batch 3900, loss[loss=0.1782, simple_loss=0.2549, pruned_loss=0.05075, over 16433.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2557, pruned_loss=0.04877, over 3281653.37 frames. ], batch size: 146, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:28:00,973 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166307.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:28:37,864 INFO [optim.py:368] (0/8) 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 12:29:07,072 INFO [train.py:904] (0/8) Epoch 17, batch 3950, loss[loss=0.1732, simple_loss=0.245, pruned_loss=0.05074, over 16785.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2557, pruned_loss=0.04931, over 3278490.81 frames. ], batch size: 83, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:29:30,618 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:18,533 INFO [train.py:904] (0/8) Epoch 17, batch 4000, loss[loss=0.1625, simple_loss=0.2474, pruned_loss=0.03877, over 16586.00 frames. ], tot_loss[loss=0.178, simple_loss=0.256, pruned_loss=0.04999, over 3271360.27 frames. ], batch size: 62, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:30:25,145 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:30:31,140 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166410.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 12:31:03,721 INFO [optim.py:368] (0/8) 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 12:31:31,418 INFO [train.py:904] (0/8) Epoch 17, batch 4050, loss[loss=0.1698, simple_loss=0.2519, pruned_loss=0.04389, over 16756.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2565, pruned_loss=0.04919, over 3280836.15 frames. ], batch size: 124, lr: 4.01e-03, grad_scale: 8.0 2023-04-30 12:31:34,057 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166454.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:31:41,999 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166459.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:11,910 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 12:32:14,542 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:31,468 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.68 vs. limit=5.0 2023-04-30 12:32:44,116 INFO [train.py:904] (0/8) Epoch 17, batch 4100, loss[loss=0.1923, simple_loss=0.2772, pruned_loss=0.05375, over 16573.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2578, pruned_loss=0.04832, over 3280970.07 frames. ], batch size: 57, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:32:45,677 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:32:54,585 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166509.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:28,701 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166531.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:31,676 INFO [optim.py:368] (0/8) 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 12:33:45,784 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:33:59,693 INFO [train.py:904] (0/8) Epoch 17, batch 4150, loss[loss=0.2055, simple_loss=0.2985, pruned_loss=0.0562, over 16243.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.265, pruned_loss=0.05092, over 3225947.93 frames. ], batch size: 165, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:34:28,432 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:00,474 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166592.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:35:14,012 INFO [train.py:904] (0/8) Epoch 17, batch 4200, loss[loss=0.2295, simple_loss=0.3115, pruned_loss=0.07377, over 16879.00 frames. ], tot_loss[loss=0.189, simple_loss=0.2724, pruned_loss=0.05278, over 3201585.38 frames. ], batch size: 109, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:00,021 INFO [optim.py:368] (0/8) 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 12:36:27,728 INFO [train.py:904] (0/8) Epoch 17, batch 4250, loss[loss=0.1735, simple_loss=0.2668, pruned_loss=0.04009, over 17222.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2753, pruned_loss=0.05194, over 3198452.09 frames. ], batch size: 45, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:36:42,896 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166663.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:37:10,591 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6909, 1.8002, 2.2875, 2.6403, 2.6479, 2.9480, 1.9021, 2.8428], device='cuda:0'), covar=tensor([0.0181, 0.0454, 0.0277, 0.0263, 0.0257, 0.0168, 0.0478, 0.0134], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0187, 0.0173, 0.0178, 0.0187, 0.0144, 0.0188, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:37:19,958 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5472, 3.5994, 3.3862, 3.0445, 3.2048, 3.4900, 3.2955, 3.3167], device='cuda:0'), covar=tensor([0.0539, 0.0536, 0.0314, 0.0253, 0.0576, 0.0389, 0.1375, 0.0459], device='cuda:0'), in_proj_covar=tensor([0.0282, 0.0391, 0.0334, 0.0320, 0.0343, 0.0372, 0.0227, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:37:39,014 INFO [train.py:904] (0/8) Epoch 17, batch 4300, loss[loss=0.1945, simple_loss=0.2836, pruned_loss=0.05269, over 17009.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2768, pruned_loss=0.05128, over 3197364.55 frames. ], batch size: 50, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:37:51,453 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166710.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 12:38:02,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7206, 6.0375, 5.7328, 5.8564, 5.5518, 5.2088, 5.4318, 6.1907], device='cuda:0'), covar=tensor([0.1096, 0.0802, 0.1086, 0.0808, 0.0845, 0.0671, 0.1066, 0.0787], device='cuda:0'), in_proj_covar=tensor([0.0636, 0.0789, 0.0640, 0.0586, 0.0499, 0.0504, 0.0653, 0.0607], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:38:24,700 INFO [optim.py:368] (0/8) 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 12:38:28,344 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:38:38,510 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1840, 5.4569, 5.1598, 5.2119, 4.9921, 4.7664, 4.7921, 5.5216], device='cuda:0'), covar=tensor([0.1016, 0.0707, 0.0999, 0.0831, 0.0793, 0.0855, 0.1073, 0.0860], device='cuda:0'), in_proj_covar=tensor([0.0637, 0.0788, 0.0639, 0.0585, 0.0499, 0.0504, 0.0653, 0.0607], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:38:52,936 INFO [train.py:904] (0/8) Epoch 17, batch 4350, loss[loss=0.1995, simple_loss=0.2848, pruned_loss=0.05711, over 16511.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.28, pruned_loss=0.05248, over 3190949.90 frames. ], batch size: 62, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:39:01,859 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166758.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 12:39:10,382 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:39:57,558 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:05,344 INFO [train.py:904] (0/8) Epoch 17, batch 4400, loss[loss=0.1862, simple_loss=0.2755, pruned_loss=0.04842, over 16655.00 frames. ], tot_loss[loss=0.1938, simple_loss=0.2815, pruned_loss=0.05311, over 3200995.47 frames. ], batch size: 57, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:40:06,853 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166803.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:15,845 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.50 vs. limit=5.0 2023-04-30 12:40:38,007 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:40:49,146 INFO [optim.py:368] (0/8) 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 12:40:55,210 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:41:05,292 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-04-30 12:41:09,392 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.49 vs. limit=5.0 2023-04-30 12:41:15,125 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=166851.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:41:16,605 INFO [train.py:904] (0/8) Epoch 17, batch 4450, loss[loss=0.213, simple_loss=0.31, pruned_loss=0.05797, over 16975.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2847, pruned_loss=0.05449, over 3194157.55 frames. ], batch size: 41, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:41:36,448 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166865.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:07,450 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=166887.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:42:28,865 INFO [train.py:904] (0/8) Epoch 17, batch 4500, loss[loss=0.2035, simple_loss=0.2876, pruned_loss=0.05974, over 16525.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2854, pruned_loss=0.05501, over 3193029.37 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:42:42,480 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 12:42:58,218 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 12:43:07,242 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=166929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:43:12,783 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.487e+02 1.901e+02 2.204e+02 2.572e+02 5.344e+02, threshold=4.409e+02, percent-clipped=1.0 2023-04-30 12:43:40,963 INFO [train.py:904] (0/8) Epoch 17, batch 4550, loss[loss=0.1728, simple_loss=0.2688, pruned_loss=0.03834, over 16945.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2859, pruned_loss=0.05575, over 3209508.42 frames. ], batch size: 102, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:43:41,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2233, 4.3285, 4.1140, 3.8074, 3.8806, 4.2344, 3.8588, 4.0019], device='cuda:0'), covar=tensor([0.0490, 0.0311, 0.0224, 0.0225, 0.0632, 0.0285, 0.0789, 0.0534], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0380, 0.0325, 0.0313, 0.0335, 0.0362, 0.0222, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:43:57,147 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=166963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:35,673 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=166990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:44:50,659 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5225, 3.5782, 3.2991, 2.8954, 3.1893, 3.4262, 3.2875, 3.2685], device='cuda:0'), covar=tensor([0.0511, 0.0388, 0.0255, 0.0253, 0.0532, 0.0338, 0.1224, 0.0477], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0379, 0.0324, 0.0312, 0.0334, 0.0360, 0.0221, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:44:50,743 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8479, 1.2359, 1.6705, 1.6529, 1.8174, 1.9221, 1.4860, 1.7597], device='cuda:0'), covar=tensor([0.0221, 0.0367, 0.0201, 0.0247, 0.0221, 0.0185, 0.0405, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0177, 0.0186, 0.0143, 0.0187, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:44:53,176 INFO [train.py:904] (0/8) Epoch 17, batch 4600, loss[loss=0.1842, simple_loss=0.2769, pruned_loss=0.0458, over 16697.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2872, pruned_loss=0.05615, over 3214198.33 frames. ], batch size: 134, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:45:07,099 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167011.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:45:38,258 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.923e+02 2.253e+02 2.642e+02 5.543e+02, threshold=4.506e+02, percent-clipped=2.0 2023-04-30 12:45:47,491 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-04-30 12:46:05,387 INFO [train.py:904] (0/8) Epoch 17, batch 4650, loss[loss=0.1883, simple_loss=0.2652, pruned_loss=0.05575, over 16232.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2865, pruned_loss=0.05626, over 3229878.41 frames. ], batch size: 35, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:47:00,061 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:13,805 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167100.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:16,423 INFO [train.py:904] (0/8) Epoch 17, batch 4700, loss[loss=0.2074, simple_loss=0.2869, pruned_loss=0.06392, over 16196.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2835, pruned_loss=0.05539, over 3217958.24 frames. ], batch size: 165, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:47:18,981 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167104.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:24,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6170, 2.5710, 1.8207, 2.7194, 2.1279, 2.7677, 2.0428, 2.2934], device='cuda:0'), covar=tensor([0.0333, 0.0362, 0.1335, 0.0184, 0.0671, 0.0430, 0.1145, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0170, 0.0189, 0.0150, 0.0169, 0.0210, 0.0196, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 12:47:42,119 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167120.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:47:54,362 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.74 vs. limit=2.0 2023-04-30 12:48:00,932 INFO [optim.py:368] (0/8) 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 12:48:06,811 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:17,228 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7266, 3.7478, 2.2990, 4.4749, 2.9505, 4.3858, 2.4734, 2.9806], device='cuda:0'), covar=tensor([0.0260, 0.0351, 0.1616, 0.0114, 0.0781, 0.0386, 0.1422, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0170, 0.0190, 0.0150, 0.0169, 0.0210, 0.0197, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 12:48:28,397 INFO [train.py:904] (0/8) Epoch 17, batch 4750, loss[loss=0.1828, simple_loss=0.2773, pruned_loss=0.04412, over 16909.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2792, pruned_loss=0.05322, over 3207523.55 frames. ], batch size: 109, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:48:41,429 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:48,125 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:48:48,171 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:16,410 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:18,959 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167187.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:49:40,732 INFO [train.py:904] (0/8) Epoch 17, batch 4800, loss[loss=0.1831, simple_loss=0.2624, pruned_loss=0.05195, over 12107.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2753, pruned_loss=0.05123, over 3203661.43 frames. ], batch size: 246, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:49:58,979 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167213.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:08,948 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4621, 3.4775, 2.5906, 2.0033, 2.3622, 2.2180, 3.4489, 3.1581], device='cuda:0'), covar=tensor([0.3042, 0.0769, 0.2041, 0.3073, 0.2710, 0.2106, 0.0722, 0.1237], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0262, 0.0297, 0.0301, 0.0291, 0.0242, 0.0283, 0.0321], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:50:24,410 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9692, 4.2158, 4.0267, 4.0711, 3.7579, 3.8196, 3.8314, 4.1929], device='cuda:0'), covar=tensor([0.1054, 0.0852, 0.0971, 0.0738, 0.0796, 0.1746, 0.0982, 0.1064], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0762, 0.0623, 0.0565, 0.0484, 0.0489, 0.0635, 0.0590], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:50:27,976 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.365e+02 2.008e+02 2.170e+02 2.594e+02 4.937e+02, threshold=4.340e+02, percent-clipped=1.0 2023-04-30 12:50:31,913 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167235.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:50:33,754 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9658, 3.8365, 4.0346, 4.1489, 4.2487, 3.8338, 4.1895, 4.2930], device='cuda:0'), covar=tensor([0.1363, 0.1047, 0.1158, 0.0586, 0.0471, 0.1450, 0.0658, 0.0538], device='cuda:0'), in_proj_covar=tensor([0.0589, 0.0728, 0.0862, 0.0740, 0.0556, 0.0589, 0.0586, 0.0689], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:50:42,349 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 12:50:45,579 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4002, 4.3407, 4.3054, 3.4912, 4.3090, 1.5827, 4.0622, 3.9033], device='cuda:0'), covar=tensor([0.0099, 0.0107, 0.0156, 0.0454, 0.0098, 0.2815, 0.0167, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0140, 0.0187, 0.0171, 0.0161, 0.0197, 0.0176, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:50:58,082 INFO [train.py:904] (0/8) Epoch 17, batch 4850, loss[loss=0.1948, simple_loss=0.2797, pruned_loss=0.05499, over 12070.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2764, pruned_loss=0.05036, over 3185926.60 frames. ], batch size: 248, lr: 4.00e-03, grad_scale: 8.0 2023-04-30 12:51:06,951 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7629, 4.7181, 4.7269, 3.5789, 4.7177, 1.6720, 4.4109, 4.3787], device='cuda:0'), covar=tensor([0.0111, 0.0111, 0.0164, 0.0661, 0.0115, 0.2952, 0.0173, 0.0300], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0140, 0.0187, 0.0171, 0.0161, 0.0197, 0.0176, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:51:46,666 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:52:12,198 INFO [train.py:904] (0/8) Epoch 17, batch 4900, loss[loss=0.1718, simple_loss=0.2626, pruned_loss=0.04046, over 17043.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.276, pruned_loss=0.04954, over 3174481.57 frames. ], batch size: 50, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:52:52,003 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3185, 4.2039, 4.3821, 4.5468, 4.7033, 4.3354, 4.6364, 4.7268], device='cuda:0'), covar=tensor([0.1657, 0.1244, 0.1538, 0.0786, 0.0554, 0.0906, 0.0712, 0.0569], device='cuda:0'), in_proj_covar=tensor([0.0595, 0.0731, 0.0865, 0.0744, 0.0558, 0.0591, 0.0591, 0.0693], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:52:55,967 INFO [optim.py:368] (0/8) 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 12:53:24,991 INFO [train.py:904] (0/8) Epoch 17, batch 4950, loss[loss=0.1977, simple_loss=0.292, pruned_loss=0.05165, over 16708.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.2763, pruned_loss=0.04938, over 3169362.33 frames. ], batch size: 76, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:54:15,040 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4079, 2.2592, 2.2969, 4.1353, 2.1944, 2.5709, 2.3375, 2.4506], device='cuda:0'), covar=tensor([0.1194, 0.3584, 0.2628, 0.0439, 0.3812, 0.2503, 0.3503, 0.3025], device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0427, 0.0352, 0.0322, 0.0427, 0.0494, 0.0396, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:54:22,565 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167391.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:54:37,760 INFO [train.py:904] (0/8) Epoch 17, batch 5000, loss[loss=0.1998, simple_loss=0.2933, pruned_loss=0.05315, over 16471.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2782, pruned_loss=0.04975, over 3187496.61 frames. ], batch size: 146, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:02,764 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:19,931 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:20,680 INFO [optim.py:368] (0/8) 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 12:55:29,659 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167439.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:46,696 INFO [train.py:904] (0/8) Epoch 17, batch 5050, loss[loss=0.1766, simple_loss=0.2675, pruned_loss=0.04287, over 16657.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2782, pruned_loss=0.04921, over 3198721.97 frames. ], batch size: 62, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:55:51,626 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:55:57,169 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:43,501 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167493.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:56:56,785 INFO [train.py:904] (0/8) Epoch 17, batch 5100, loss[loss=0.1855, simple_loss=0.2766, pruned_loss=0.04719, over 15512.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2766, pruned_loss=0.04809, over 3207286.57 frames. ], batch size: 191, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:57:39,790 INFO [optim.py:368] (0/8) 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 12:57:50,563 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9753, 5.3166, 5.5321, 5.2736, 5.3210, 5.8877, 5.3501, 5.0799], device='cuda:0'), covar=tensor([0.0839, 0.1669, 0.1320, 0.1744, 0.2246, 0.0760, 0.1174, 0.2253], device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0550, 0.0596, 0.0460, 0.0617, 0.0640, 0.0471, 0.0619], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:58:08,926 INFO [train.py:904] (0/8) Epoch 17, batch 5150, loss[loss=0.1643, simple_loss=0.2674, pruned_loss=0.03062, over 16932.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2765, pruned_loss=0.04775, over 3168451.91 frames. ], batch size: 96, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:58:09,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8073, 3.7943, 3.9695, 3.7462, 3.8488, 4.2852, 3.9274, 3.6067], device='cuda:0'), covar=tensor([0.2032, 0.2210, 0.1807, 0.2204, 0.2437, 0.1550, 0.1470, 0.2716], device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0549, 0.0595, 0.0460, 0.0616, 0.0639, 0.0471, 0.0618], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 12:58:16,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6137, 3.6595, 2.0064, 4.2191, 2.7276, 4.1587, 2.4534, 2.9953], device='cuda:0'), covar=tensor([0.0272, 0.0361, 0.1788, 0.0190, 0.0859, 0.0474, 0.1450, 0.0763], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0169, 0.0188, 0.0148, 0.0168, 0.0207, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 12:58:31,513 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2075, 2.1539, 2.1557, 3.7977, 2.1084, 2.5701, 2.2673, 2.3551], device='cuda:0'), covar=tensor([0.1264, 0.3547, 0.2715, 0.0497, 0.3749, 0.2361, 0.3463, 0.2940], device='cuda:0'), in_proj_covar=tensor([0.0385, 0.0426, 0.0350, 0.0321, 0.0424, 0.0491, 0.0394, 0.0496], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:58:38,493 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6282, 1.5122, 2.1894, 2.5206, 2.4962, 2.8197, 1.5227, 2.8285], device='cuda:0'), covar=tensor([0.0171, 0.0633, 0.0278, 0.0287, 0.0283, 0.0165, 0.0690, 0.0131], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0188, 0.0174, 0.0178, 0.0187, 0.0143, 0.0188, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 12:58:56,786 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 12:59:21,208 INFO [train.py:904] (0/8) Epoch 17, batch 5200, loss[loss=0.1646, simple_loss=0.2456, pruned_loss=0.04179, over 17124.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2747, pruned_loss=0.04747, over 3174043.85 frames. ], batch size: 47, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 12:59:25,941 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8482, 3.7960, 4.4022, 2.0208, 4.5787, 4.5518, 3.1051, 3.2275], device='cuda:0'), covar=tensor([0.0762, 0.0231, 0.0103, 0.1244, 0.0043, 0.0080, 0.0396, 0.0481], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0106, 0.0093, 0.0137, 0.0075, 0.0119, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 13:00:07,267 INFO [optim.py:368] (0/8) 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 13:00:07,579 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167633.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:00:36,338 INFO [train.py:904] (0/8) Epoch 17, batch 5250, loss[loss=0.174, simple_loss=0.2627, pruned_loss=0.04264, over 16489.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2722, pruned_loss=0.04688, over 3180445.78 frames. ], batch size: 68, lr: 3.99e-03, grad_scale: 16.0 2023-04-30 13:01:48,355 INFO [train.py:904] (0/8) Epoch 17, batch 5300, loss[loss=0.2075, simple_loss=0.2807, pruned_loss=0.06712, over 12417.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2687, pruned_loss=0.04586, over 3186401.27 frames. ], batch size: 247, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:02:12,773 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:02:25,176 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4084, 2.6492, 2.1884, 2.4168, 2.9960, 2.6866, 3.0633, 3.2239], device='cuda:0'), covar=tensor([0.0095, 0.0355, 0.0478, 0.0400, 0.0232, 0.0348, 0.0188, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0223, 0.0215, 0.0216, 0.0225, 0.0224, 0.0225, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:02:35,323 INFO [optim.py:368] (0/8) 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 13:03:01,795 INFO [train.py:904] (0/8) Epoch 17, batch 5350, loss[loss=0.1795, simple_loss=0.2683, pruned_loss=0.04537, over 16789.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2672, pruned_loss=0.04525, over 3192107.98 frames. ], batch size: 39, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:03:08,257 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:13,832 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=167760.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:23,344 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=167767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:40,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7642, 3.6645, 4.2331, 1.8933, 4.4167, 4.4077, 3.0659, 3.2375], device='cuda:0'), covar=tensor([0.0729, 0.0228, 0.0126, 0.1201, 0.0048, 0.0084, 0.0376, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0106, 0.0093, 0.0138, 0.0075, 0.0118, 0.0125, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 13:03:41,738 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:03:44,332 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.97 vs. limit=5.0 2023-04-30 13:03:53,565 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=167788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:14,427 INFO [train.py:904] (0/8) Epoch 17, batch 5400, loss[loss=0.1867, simple_loss=0.2749, pruned_loss=0.04928, over 16952.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2701, pruned_loss=0.04608, over 3207157.30 frames. ], batch size: 109, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:04:17,678 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:24,147 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=167808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:04:54,068 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=167828.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:05:02,700 INFO [optim.py:368] (0/8) 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 13:05:31,692 INFO [train.py:904] (0/8) Epoch 17, batch 5450, loss[loss=0.1746, simple_loss=0.2655, pruned_loss=0.04182, over 17117.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2728, pruned_loss=0.04715, over 3214286.26 frames. ], batch size: 47, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:06:48,597 INFO [train.py:904] (0/8) Epoch 17, batch 5500, loss[loss=0.2507, simple_loss=0.325, pruned_loss=0.08819, over 11468.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2805, pruned_loss=0.05157, over 3191000.70 frames. ], batch size: 246, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:07:20,799 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2809, 4.3265, 4.1309, 3.8851, 3.8431, 4.2638, 3.9952, 3.9982], device='cuda:0'), covar=tensor([0.0689, 0.0529, 0.0307, 0.0284, 0.0866, 0.0465, 0.0680, 0.0637], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0388, 0.0328, 0.0316, 0.0337, 0.0370, 0.0223, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:07:39,671 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.878e+02 2.929e+02 3.465e+02 4.184e+02 8.083e+02, threshold=6.929e+02, percent-clipped=17.0 2023-04-30 13:08:05,267 INFO [train.py:904] (0/8) Epoch 17, batch 5550, loss[loss=0.2611, simple_loss=0.3225, pruned_loss=0.09982, over 11142.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.288, pruned_loss=0.05711, over 3157901.93 frames. ], batch size: 247, lr: 3.99e-03, grad_scale: 4.0 2023-04-30 13:08:58,454 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-04-30 13:09:25,545 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-168000.pt 2023-04-30 13:09:31,462 INFO [train.py:904] (0/8) Epoch 17, batch 5600, loss[loss=0.2294, simple_loss=0.3131, pruned_loss=0.07289, over 16860.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.294, pruned_loss=0.06277, over 3081406.30 frames. ], batch size: 116, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:09:37,398 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7384, 4.7453, 4.6011, 3.8998, 4.6591, 1.7780, 4.4499, 4.4271], device='cuda:0'), covar=tensor([0.0088, 0.0069, 0.0162, 0.0346, 0.0090, 0.2638, 0.0118, 0.0206], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0139, 0.0185, 0.0170, 0.0160, 0.0196, 0.0174, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:10:09,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 13:10:27,359 INFO [optim.py:368] (0/8) 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 13:10:27,869 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168035.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:10:54,422 INFO [train.py:904] (0/8) Epoch 17, batch 5650, loss[loss=0.2298, simple_loss=0.3114, pruned_loss=0.07405, over 16285.00 frames. ], tot_loss[loss=0.2146, simple_loss=0.2978, pruned_loss=0.0657, over 3077539.04 frames. ], batch size: 165, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:11:30,938 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168075.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:11:50,893 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:04,122 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168096.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:12:10,285 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7753, 4.7942, 5.1845, 5.1464, 5.1859, 4.8556, 4.7773, 4.5798], device='cuda:0'), covar=tensor([0.0331, 0.0571, 0.0345, 0.0388, 0.0462, 0.0359, 0.1002, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0412, 0.0401, 0.0378, 0.0448, 0.0421, 0.0521, 0.0336], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 13:12:12,316 INFO [train.py:904] (0/8) Epoch 17, batch 5700, loss[loss=0.2377, simple_loss=0.3182, pruned_loss=0.07861, over 16272.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.2992, pruned_loss=0.0671, over 3073541.93 frames. ], batch size: 165, lr: 3.99e-03, grad_scale: 8.0 2023-04-30 13:12:45,726 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168123.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:12:57,101 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 13:13:04,295 INFO [optim.py:368] (0/8) 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 13:13:06,080 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:13:31,693 INFO [train.py:904] (0/8) Epoch 17, batch 5750, loss[loss=0.222, simple_loss=0.2902, pruned_loss=0.07688, over 10872.00 frames. ], tot_loss[loss=0.2185, simple_loss=0.3013, pruned_loss=0.06789, over 3061244.03 frames. ], batch size: 247, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:14:32,895 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8091, 1.7830, 1.6286, 1.5757, 1.9323, 1.6341, 1.6240, 1.9192], device='cuda:0'), covar=tensor([0.0188, 0.0245, 0.0360, 0.0323, 0.0183, 0.0225, 0.0193, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0222, 0.0215, 0.0216, 0.0224, 0.0223, 0.0224, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:14:50,329 INFO [train.py:904] (0/8) Epoch 17, batch 5800, loss[loss=0.1837, simple_loss=0.2818, pruned_loss=0.04282, over 16903.00 frames. ], tot_loss[loss=0.2187, simple_loss=0.302, pruned_loss=0.06771, over 3037971.41 frames. ], batch size: 90, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:15:40,535 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168233.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:15:43,219 INFO [optim.py:368] (0/8) 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 13:16:04,385 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-04-30 13:16:09,629 INFO [train.py:904] (0/8) Epoch 17, batch 5850, loss[loss=0.2392, simple_loss=0.3062, pruned_loss=0.08606, over 11359.00 frames. ], tot_loss[loss=0.216, simple_loss=0.2996, pruned_loss=0.06614, over 3038937.13 frames. ], batch size: 246, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:17:19,309 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=168294.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:17:31,780 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5798, 3.5549, 3.9342, 1.8857, 4.0892, 4.0756, 3.1057, 3.0101], device='cuda:0'), covar=tensor([0.0792, 0.0219, 0.0148, 0.1225, 0.0056, 0.0133, 0.0351, 0.0466], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0104, 0.0092, 0.0136, 0.0074, 0.0118, 0.0124, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 13:17:32,954 INFO [train.py:904] (0/8) Epoch 17, batch 5900, loss[loss=0.1922, simple_loss=0.2947, pruned_loss=0.0449, over 16685.00 frames. ], tot_loss[loss=0.2147, simple_loss=0.2986, pruned_loss=0.06535, over 3055055.19 frames. ], batch size: 89, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:17:52,253 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7177, 1.7664, 1.6009, 1.5194, 1.8792, 1.5753, 1.6737, 1.8822], device='cuda:0'), covar=tensor([0.0153, 0.0230, 0.0332, 0.0272, 0.0168, 0.0222, 0.0139, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0222, 0.0215, 0.0215, 0.0224, 0.0222, 0.0224, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:18:28,194 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.786e+02 3.302e+02 4.334e+02 9.819e+02, threshold=6.604e+02, percent-clipped=5.0 2023-04-30 13:18:52,634 INFO [train.py:904] (0/8) Epoch 17, batch 5950, loss[loss=0.184, simple_loss=0.2814, pruned_loss=0.0433, over 16836.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2985, pruned_loss=0.06355, over 3078262.17 frames. ], batch size: 102, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:19:29,633 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168375.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:19:52,883 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168391.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:20:12,036 INFO [train.py:904] (0/8) Epoch 17, batch 6000, loss[loss=0.1866, simple_loss=0.2672, pruned_loss=0.05299, over 16461.00 frames. ], tot_loss[loss=0.2124, simple_loss=0.2979, pruned_loss=0.06346, over 3059276.93 frames. ], batch size: 68, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:20:12,037 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 13:20:21,979 INFO [train.py:938] (0/8) Epoch 17, validation: loss=0.1535, simple_loss=0.2667, pruned_loss=0.0202, over 944034.00 frames. 2023-04-30 13:20:21,980 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 13:20:53,936 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:20:53,974 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168423.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:21:15,441 INFO [optim.py:368] (0/8) 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 13:21:39,920 INFO [train.py:904] (0/8) Epoch 17, batch 6050, loss[loss=0.1906, simple_loss=0.2921, pruned_loss=0.04452, over 16377.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2959, pruned_loss=0.06212, over 3081998.00 frames. ], batch size: 35, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:22:09,419 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168471.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:22:44,898 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8204, 2.5914, 2.5486, 4.5518, 3.1354, 4.1964, 1.6152, 2.9949], device='cuda:0'), covar=tensor([0.1390, 0.0881, 0.1283, 0.0156, 0.0312, 0.0417, 0.1702, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0166, 0.0188, 0.0174, 0.0201, 0.0209, 0.0191, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 13:22:52,113 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.32 vs. limit=5.0 2023-04-30 13:22:59,848 INFO [train.py:904] (0/8) Epoch 17, batch 6100, loss[loss=0.1801, simple_loss=0.2702, pruned_loss=0.04501, over 17168.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2954, pruned_loss=0.06078, over 3107826.95 frames. ], batch size: 46, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:23:55,309 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.024e+02 2.750e+02 3.352e+02 4.017e+02 8.288e+02, threshold=6.704e+02, percent-clipped=2.0 2023-04-30 13:24:18,358 INFO [train.py:904] (0/8) Epoch 17, batch 6150, loss[loss=0.1962, simple_loss=0.2874, pruned_loss=0.05253, over 16464.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.294, pruned_loss=0.0606, over 3110597.90 frames. ], batch size: 146, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:25:17,302 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=168589.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:25:35,099 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-04-30 13:25:36,311 INFO [train.py:904] (0/8) Epoch 17, batch 6200, loss[loss=0.1962, simple_loss=0.2791, pruned_loss=0.05662, over 16660.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2912, pruned_loss=0.05962, over 3120159.14 frames. ], batch size: 134, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:26:31,029 INFO [optim.py:368] (0/8) 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 13:26:38,607 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1412, 5.1252, 4.9834, 4.5875, 4.6093, 5.0183, 4.9489, 4.7157], device='cuda:0'), covar=tensor([0.0804, 0.0762, 0.0333, 0.0365, 0.1076, 0.0675, 0.0465, 0.0907], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0384, 0.0324, 0.0312, 0.0334, 0.0365, 0.0220, 0.0388], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:26:53,714 INFO [train.py:904] (0/8) Epoch 17, batch 6250, loss[loss=0.199, simple_loss=0.2894, pruned_loss=0.05433, over 16365.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2904, pruned_loss=0.05882, over 3136587.20 frames. ], batch size: 146, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:27:53,825 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168691.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:27:58,379 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0111, 4.9995, 4.8360, 4.1548, 4.9080, 1.9159, 4.6681, 4.6475], device='cuda:0'), covar=tensor([0.0092, 0.0079, 0.0170, 0.0364, 0.0090, 0.2548, 0.0127, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0139, 0.0185, 0.0170, 0.0159, 0.0195, 0.0173, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:28:10,637 INFO [train.py:904] (0/8) Epoch 17, batch 6300, loss[loss=0.1998, simple_loss=0.2875, pruned_loss=0.0561, over 16946.00 frames. ], tot_loss[loss=0.203, simple_loss=0.29, pruned_loss=0.05802, over 3150002.55 frames. ], batch size: 109, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:29:06,432 INFO [optim.py:368] (0/8) 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 13:29:09,337 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168739.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:29:29,591 INFO [train.py:904] (0/8) Epoch 17, batch 6350, loss[loss=0.2355, simple_loss=0.2987, pruned_loss=0.08614, over 11175.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2919, pruned_loss=0.0604, over 3106036.67 frames. ], batch size: 247, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:30:46,703 INFO [train.py:904] (0/8) Epoch 17, batch 6400, loss[loss=0.1964, simple_loss=0.2866, pruned_loss=0.05306, over 16904.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2919, pruned_loss=0.06089, over 3113125.04 frames. ], batch size: 109, lr: 3.98e-03, grad_scale: 8.0 2023-04-30 13:31:24,048 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 13:31:42,104 INFO [optim.py:368] (0/8) 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 13:32:03,562 INFO [train.py:904] (0/8) Epoch 17, batch 6450, loss[loss=0.2045, simple_loss=0.2886, pruned_loss=0.06018, over 16294.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2914, pruned_loss=0.05958, over 3134233.28 frames. ], batch size: 165, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:33:02,235 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=168889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:33:20,848 INFO [train.py:904] (0/8) Epoch 17, batch 6500, loss[loss=0.1988, simple_loss=0.2797, pruned_loss=0.05896, over 17132.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2898, pruned_loss=0.05985, over 3110109.70 frames. ], batch size: 47, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:33:43,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3704, 3.3349, 3.4128, 3.4896, 3.5401, 3.3497, 3.5024, 3.5727], device='cuda:0'), covar=tensor([0.1149, 0.0919, 0.0981, 0.0616, 0.0622, 0.1961, 0.0964, 0.0702], device='cuda:0'), in_proj_covar=tensor([0.0587, 0.0722, 0.0860, 0.0740, 0.0554, 0.0588, 0.0592, 0.0689], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:33:48,449 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 13:34:15,421 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=168937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:34:16,271 INFO [optim.py:368] (0/8) 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 13:34:40,485 INFO [train.py:904] (0/8) Epoch 17, batch 6550, loss[loss=0.2198, simple_loss=0.3191, pruned_loss=0.06026, over 15310.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2927, pruned_loss=0.06073, over 3124084.55 frames. ], batch size: 190, lr: 3.98e-03, grad_scale: 4.0 2023-04-30 13:35:03,842 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168967.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:35:06,309 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=168969.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:35:58,169 INFO [train.py:904] (0/8) Epoch 17, batch 6600, loss[loss=0.1934, simple_loss=0.2806, pruned_loss=0.05315, over 16936.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2949, pruned_loss=0.06128, over 3115944.76 frames. ], batch size: 109, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:36:37,548 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:36:40,043 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169030.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 13:36:50,875 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.775e+02 2.872e+02 3.571e+02 4.497e+02 9.620e+02, threshold=7.142e+02, percent-clipped=5.0 2023-04-30 13:37:13,177 INFO [train.py:904] (0/8) Epoch 17, batch 6650, loss[loss=0.1989, simple_loss=0.2798, pruned_loss=0.05896, over 16837.00 frames. ], tot_loss[loss=0.2106, simple_loss=0.2957, pruned_loss=0.06279, over 3095705.89 frames. ], batch size: 116, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:37:59,846 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169082.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:38:01,992 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.42 vs. limit=5.0 2023-04-30 13:38:30,529 INFO [train.py:904] (0/8) Epoch 17, batch 6700, loss[loss=0.1879, simple_loss=0.2715, pruned_loss=0.05213, over 16753.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.294, pruned_loss=0.06212, over 3104403.07 frames. ], batch size: 39, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:38:53,628 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169116.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:00,611 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4318, 3.2712, 2.6199, 2.1413, 2.2257, 2.1930, 3.3086, 3.0634], device='cuda:0'), covar=tensor([0.2719, 0.0671, 0.1744, 0.2455, 0.2440, 0.2066, 0.0497, 0.1284], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0260, 0.0295, 0.0299, 0.0289, 0.0242, 0.0283, 0.0319], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 13:39:02,418 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9119, 3.8173, 4.0130, 4.1319, 4.2410, 3.8445, 4.2104, 4.2517], device='cuda:0'), covar=tensor([0.1672, 0.1287, 0.1399, 0.0710, 0.0613, 0.1627, 0.0749, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0722, 0.0858, 0.0740, 0.0557, 0.0588, 0.0593, 0.0691], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:39:23,217 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8554, 2.7101, 2.7487, 2.0611, 2.6136, 2.1187, 2.6322, 2.8897], device='cuda:0'), covar=tensor([0.0303, 0.0798, 0.0535, 0.1782, 0.0802, 0.0919, 0.0646, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0158, 0.0164, 0.0150, 0.0142, 0.0128, 0.0142, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 13:39:26,524 INFO [optim.py:368] (0/8) 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 13:39:34,929 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169143.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:39:47,631 INFO [train.py:904] (0/8) Epoch 17, batch 6750, loss[loss=0.1835, simple_loss=0.2688, pruned_loss=0.04915, over 16738.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2932, pruned_loss=0.06261, over 3096869.82 frames. ], batch size: 124, lr: 3.97e-03, grad_scale: 4.0 2023-04-30 13:39:58,661 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169159.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:40:26,226 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169177.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:41:04,389 INFO [train.py:904] (0/8) Epoch 17, batch 6800, loss[loss=0.2053, simple_loss=0.2807, pruned_loss=0.06488, over 16317.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2929, pruned_loss=0.06215, over 3105732.72 frames. ], batch size: 35, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:41:33,052 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169220.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:42:02,687 INFO [optim.py:368] (0/8) 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 13:42:17,651 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0565, 2.1134, 2.0716, 3.9039, 2.0546, 2.4983, 2.1501, 2.2583], device='cuda:0'), covar=tensor([0.1419, 0.3691, 0.2929, 0.0474, 0.4063, 0.2569, 0.3772, 0.3128], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0419, 0.0349, 0.0316, 0.0423, 0.0486, 0.0391, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:42:23,106 INFO [train.py:904] (0/8) Epoch 17, batch 6850, loss[loss=0.1929, simple_loss=0.2966, pruned_loss=0.04464, over 16714.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2939, pruned_loss=0.06208, over 3119641.68 frames. ], batch size: 76, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:42:59,310 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6222, 3.7019, 2.0449, 3.9869, 2.6808, 4.0178, 2.2923, 2.8804], device='cuda:0'), covar=tensor([0.0228, 0.0323, 0.1799, 0.0213, 0.0810, 0.0497, 0.1547, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0170, 0.0192, 0.0150, 0.0171, 0.0209, 0.0197, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 13:43:01,055 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.61 vs. limit=5.0 2023-04-30 13:43:37,895 INFO [train.py:904] (0/8) Epoch 17, batch 6900, loss[loss=0.2158, simple_loss=0.3004, pruned_loss=0.06559, over 16417.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2958, pruned_loss=0.06143, over 3133168.69 frames. ], batch size: 146, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:44:10,304 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:44:14,316 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169325.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 13:44:31,059 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.82 vs. limit=2.0 2023-04-30 13:44:33,269 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.634e+02 3.268e+02 4.133e+02 7.362e+02, threshold=6.535e+02, percent-clipped=1.0 2023-04-30 13:44:55,624 INFO [train.py:904] (0/8) Epoch 17, batch 6950, loss[loss=0.1899, simple_loss=0.2788, pruned_loss=0.05055, over 16603.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2973, pruned_loss=0.06307, over 3131225.05 frames. ], batch size: 76, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:45:57,681 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8647, 4.6683, 4.9189, 5.1357, 5.3289, 4.7931, 5.3288, 5.2754], device='cuda:0'), covar=tensor([0.1982, 0.1430, 0.1936, 0.0747, 0.0649, 0.0881, 0.0605, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0584, 0.0722, 0.0859, 0.0737, 0.0556, 0.0585, 0.0593, 0.0688], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:46:11,849 INFO [train.py:904] (0/8) Epoch 17, batch 7000, loss[loss=0.2191, simple_loss=0.3104, pruned_loss=0.06388, over 15453.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2976, pruned_loss=0.06254, over 3132389.13 frames. ], batch size: 191, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:47:07,100 INFO [optim.py:368] (0/8) 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 13:47:07,444 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:47:28,668 INFO [train.py:904] (0/8) Epoch 17, batch 7050, loss[loss=0.2166, simple_loss=0.2981, pruned_loss=0.06756, over 16214.00 frames. ], tot_loss[loss=0.2113, simple_loss=0.2981, pruned_loss=0.06227, over 3125232.42 frames. ], batch size: 165, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:48:00,532 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169472.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:48:47,199 INFO [train.py:904] (0/8) Epoch 17, batch 7100, loss[loss=0.2044, simple_loss=0.2821, pruned_loss=0.06333, over 16642.00 frames. ], tot_loss[loss=0.2109, simple_loss=0.2968, pruned_loss=0.06253, over 3098837.70 frames. ], batch size: 57, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:49:08,057 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169515.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:49:28,402 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4096, 2.9267, 2.7065, 2.2720, 2.2772, 2.2722, 2.9015, 2.8859], device='cuda:0'), covar=tensor([0.2250, 0.0625, 0.1400, 0.2191, 0.2146, 0.1933, 0.0475, 0.1069], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0261, 0.0297, 0.0300, 0.0289, 0.0242, 0.0284, 0.0320], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 13:49:42,146 INFO [optim.py:368] (0/8) 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 13:50:02,982 INFO [train.py:904] (0/8) Epoch 17, batch 7150, loss[loss=0.2678, simple_loss=0.3193, pruned_loss=0.1082, over 11471.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2952, pruned_loss=0.06257, over 3083674.35 frames. ], batch size: 246, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:50:48,946 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 13:51:19,568 INFO [train.py:904] (0/8) Epoch 17, batch 7200, loss[loss=0.1685, simple_loss=0.2599, pruned_loss=0.03854, over 16475.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2932, pruned_loss=0.06093, over 3073370.40 frames. ], batch size: 75, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:51:44,562 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5664, 3.5295, 2.7251, 2.2062, 2.4622, 2.3886, 3.7837, 3.3086], device='cuda:0'), covar=tensor([0.2818, 0.0756, 0.1861, 0.2567, 0.2510, 0.1959, 0.0474, 0.1195], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0263, 0.0299, 0.0301, 0.0290, 0.0243, 0.0285, 0.0322], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 13:51:48,209 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169621.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:51,392 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169623.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:51:55,202 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169625.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:52:16,449 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.648e+02 2.818e+02 3.396e+02 4.262e+02 6.944e+02, threshold=6.791e+02, percent-clipped=0.0 2023-04-30 13:52:38,964 INFO [train.py:904] (0/8) Epoch 17, batch 7250, loss[loss=0.1858, simple_loss=0.27, pruned_loss=0.05085, over 16911.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.291, pruned_loss=0.05965, over 3080838.01 frames. ], batch size: 109, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:53:08,594 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169671.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:12,522 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169673.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 13:53:25,490 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:53:28,123 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8383, 3.9672, 3.2290, 2.4428, 2.8366, 2.7616, 4.4685, 3.6529], device='cuda:0'), covar=tensor([0.2686, 0.0680, 0.1529, 0.2437, 0.2527, 0.1720, 0.0357, 0.1106], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0262, 0.0298, 0.0300, 0.0290, 0.0243, 0.0285, 0.0322], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 13:53:57,118 INFO [train.py:904] (0/8) Epoch 17, batch 7300, loss[loss=0.2131, simple_loss=0.301, pruned_loss=0.06259, over 16266.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2906, pruned_loss=0.05981, over 3081926.99 frames. ], batch size: 165, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:54:14,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7454, 2.7329, 2.7802, 2.1546, 2.6669, 2.1308, 2.5897, 2.8812], device='cuda:0'), covar=tensor([0.0291, 0.0777, 0.0478, 0.1747, 0.0823, 0.0882, 0.0548, 0.0675], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0159, 0.0165, 0.0151, 0.0143, 0.0128, 0.0142, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 13:54:36,543 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7202, 2.3562, 2.2039, 3.1474, 2.4032, 3.5134, 1.5124, 2.6391], device='cuda:0'), covar=tensor([0.1301, 0.0789, 0.1321, 0.0172, 0.0235, 0.0495, 0.1582, 0.0855], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0169, 0.0192, 0.0175, 0.0204, 0.0214, 0.0195, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 13:54:41,454 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:54:55,066 INFO [optim.py:368] (0/8) 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 13:54:55,477 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169738.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:10,622 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169747.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:55:17,637 INFO [train.py:904] (0/8) Epoch 17, batch 7350, loss[loss=0.2181, simple_loss=0.3132, pruned_loss=0.06147, over 16873.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.292, pruned_loss=0.06085, over 3064947.59 frames. ], batch size: 102, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:55:49,845 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,136 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:12,317 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=169786.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:19,145 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169790.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:20,948 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5708, 4.8263, 4.5619, 4.6352, 4.3506, 4.3753, 4.3224, 4.8897], device='cuda:0'), covar=tensor([0.1070, 0.0850, 0.1111, 0.0821, 0.0789, 0.1120, 0.1078, 0.0921], device='cuda:0'), in_proj_covar=tensor([0.0620, 0.0757, 0.0618, 0.0561, 0.0477, 0.0488, 0.0625, 0.0586], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:56:37,959 INFO [train.py:904] (0/8) Epoch 17, batch 7400, loss[loss=0.209, simple_loss=0.3002, pruned_loss=0.0589, over 17090.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2924, pruned_loss=0.06071, over 3090128.44 frames. ], batch size: 49, lr: 3.97e-03, grad_scale: 8.0 2023-04-30 13:56:48,646 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:56:59,905 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=169815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:06,370 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4250, 2.8774, 2.6782, 2.2692, 2.2961, 2.2452, 2.8557, 2.8788], device='cuda:0'), covar=tensor([0.2379, 0.0691, 0.1398, 0.2260, 0.2201, 0.2188, 0.0466, 0.1207], device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0264, 0.0301, 0.0303, 0.0293, 0.0246, 0.0288, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 13:57:07,987 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169820.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:36,688 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.712e+02 2.775e+02 3.366e+02 3.998e+02 8.087e+02, threshold=6.732e+02, percent-clipped=3.0 2023-04-30 13:57:51,854 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=169847.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:57:55,385 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3280, 2.1213, 1.6561, 1.9090, 2.4087, 2.1105, 2.1950, 2.5572], device='cuda:0'), covar=tensor([0.0178, 0.0378, 0.0501, 0.0437, 0.0225, 0.0333, 0.0204, 0.0228], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0221, 0.0214, 0.0215, 0.0220, 0.0220, 0.0221, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 13:57:59,465 INFO [train.py:904] (0/8) Epoch 17, batch 7450, loss[loss=0.1959, simple_loss=0.2836, pruned_loss=0.05414, over 16618.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2933, pruned_loss=0.06158, over 3104604.41 frames. ], batch size: 62, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 13:58:19,018 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=169863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 13:59:20,321 INFO [train.py:904] (0/8) Epoch 17, batch 7500, loss[loss=0.1987, simple_loss=0.2864, pruned_loss=0.05554, over 16451.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2936, pruned_loss=0.06135, over 3087763.11 frames. ], batch size: 146, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:00:17,627 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.828e+02 2.669e+02 3.210e+02 3.850e+02 6.683e+02, threshold=6.420e+02, percent-clipped=0.0 2023-04-30 14:00:39,258 INFO [train.py:904] (0/8) Epoch 17, batch 7550, loss[loss=0.193, simple_loss=0.2834, pruned_loss=0.05131, over 16157.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2925, pruned_loss=0.06136, over 3089974.24 frames. ], batch size: 165, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:00:42,991 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2172, 4.3116, 4.4555, 4.2847, 4.3435, 4.8436, 4.4233, 4.2349], device='cuda:0'), covar=tensor([0.1672, 0.2092, 0.2437, 0.2055, 0.2732, 0.1175, 0.1668, 0.2407], device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0560, 0.0616, 0.0468, 0.0633, 0.0647, 0.0488, 0.0630], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 14:00:54,295 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8388, 1.7626, 2.3692, 2.7860, 2.6382, 3.2021, 1.9715, 3.1043], device='cuda:0'), covar=tensor([0.0198, 0.0471, 0.0303, 0.0264, 0.0290, 0.0143, 0.0502, 0.0107], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0185, 0.0171, 0.0174, 0.0184, 0.0142, 0.0187, 0.0136], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:01:18,918 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=169977.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:01:27,386 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6490, 2.3361, 2.3177, 3.4865, 2.4557, 3.7860, 1.4616, 2.7509], device='cuda:0'), covar=tensor([0.1298, 0.0852, 0.1221, 0.0176, 0.0195, 0.0367, 0.1637, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0168, 0.0190, 0.0174, 0.0203, 0.0211, 0.0194, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 14:01:54,770 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-170000.pt 2023-04-30 14:02:00,833 INFO [train.py:904] (0/8) Epoch 17, batch 7600, loss[loss=0.1938, simple_loss=0.2862, pruned_loss=0.05073, over 16989.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2924, pruned_loss=0.06195, over 3076530.68 frames. ], batch size: 55, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:02:04,415 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9389, 3.3482, 3.0372, 1.6973, 2.6825, 1.8982, 3.4042, 3.5527], device='cuda:0'), covar=tensor([0.0238, 0.0657, 0.0752, 0.2406, 0.1082, 0.1280, 0.0649, 0.0843], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0159, 0.0164, 0.0151, 0.0143, 0.0128, 0.0142, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 14:02:06,188 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1968, 4.0986, 4.2828, 4.4198, 4.5304, 4.1034, 4.4419, 4.5397], device='cuda:0'), covar=tensor([0.1661, 0.1161, 0.1363, 0.0649, 0.0540, 0.1296, 0.0861, 0.0649], device='cuda:0'), in_proj_covar=tensor([0.0580, 0.0718, 0.0852, 0.0732, 0.0555, 0.0582, 0.0595, 0.0687], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:02:25,111 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7758, 1.7990, 1.5698, 1.4965, 1.8836, 1.5751, 1.6503, 1.9033], device='cuda:0'), covar=tensor([0.0150, 0.0238, 0.0369, 0.0303, 0.0186, 0.0252, 0.0168, 0.0180], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0220, 0.0213, 0.0215, 0.0220, 0.0219, 0.0221, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:02:58,193 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.022e+02 2.819e+02 3.511e+02 4.461e+02 7.625e+02, threshold=7.022e+02, percent-clipped=5.0 2023-04-30 14:03:07,370 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170045.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:03:11,023 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 14:03:17,373 INFO [train.py:904] (0/8) Epoch 17, batch 7650, loss[loss=0.1766, simple_loss=0.2634, pruned_loss=0.04486, over 17200.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2932, pruned_loss=0.06264, over 3077881.44 frames. ], batch size: 45, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:07,096 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170085.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:33,138 INFO [train.py:904] (0/8) Epoch 17, batch 7700, loss[loss=0.1971, simple_loss=0.2967, pruned_loss=0.04879, over 16565.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2931, pruned_loss=0.0629, over 3058160.85 frames. ], batch size: 75, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:04:34,740 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:04:39,597 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170106.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:29,558 INFO [optim.py:368] (0/8) 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 14:05:34,965 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170142.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:05:50,033 INFO [train.py:904] (0/8) Epoch 17, batch 7750, loss[loss=0.2038, simple_loss=0.2908, pruned_loss=0.05837, over 15334.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2933, pruned_loss=0.06276, over 3065707.98 frames. ], batch size: 190, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:06:07,394 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1901, 4.1008, 4.2383, 4.3923, 4.5184, 4.1649, 4.4525, 4.5327], device='cuda:0'), covar=tensor([0.1709, 0.1168, 0.1438, 0.0698, 0.0591, 0.1162, 0.0776, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0586, 0.0724, 0.0861, 0.0736, 0.0559, 0.0587, 0.0598, 0.0691], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:06:07,783 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 14:06:39,091 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8620, 2.7437, 2.5967, 1.9425, 2.5387, 2.6567, 2.5952, 1.9363], device='cuda:0'), covar=tensor([0.0424, 0.0072, 0.0081, 0.0338, 0.0131, 0.0134, 0.0120, 0.0372], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0076, 0.0078, 0.0132, 0.0091, 0.0102, 0.0090, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 14:06:50,684 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 14:07:07,220 INFO [train.py:904] (0/8) Epoch 17, batch 7800, loss[loss=0.1903, simple_loss=0.2906, pruned_loss=0.04505, over 16849.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2943, pruned_loss=0.0636, over 3076249.67 frames. ], batch size: 96, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:07:25,321 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 14:07:29,718 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9605, 3.1791, 3.1684, 2.1102, 2.9716, 3.1723, 3.0582, 1.8787], device='cuda:0'), covar=tensor([0.0517, 0.0063, 0.0069, 0.0390, 0.0102, 0.0119, 0.0101, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0076, 0.0077, 0.0132, 0.0091, 0.0102, 0.0089, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 14:07:48,628 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9534, 3.0610, 3.2446, 1.3707, 3.3035, 3.6668, 2.9398, 2.5305], device='cuda:0'), covar=tensor([0.1144, 0.0229, 0.0224, 0.1483, 0.0126, 0.0148, 0.0397, 0.0593], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0103, 0.0092, 0.0135, 0.0073, 0.0117, 0.0124, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 14:08:03,355 INFO [optim.py:368] (0/8) 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 14:08:16,402 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 14:08:23,350 INFO [train.py:904] (0/8) Epoch 17, batch 7850, loss[loss=0.1885, simple_loss=0.2821, pruned_loss=0.04748, over 16880.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2951, pruned_loss=0.06313, over 3070899.08 frames. ], batch size: 83, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:09:00,954 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170277.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:09:38,291 INFO [train.py:904] (0/8) Epoch 17, batch 7900, loss[loss=0.2246, simple_loss=0.31, pruned_loss=0.06961, over 15247.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2939, pruned_loss=0.06234, over 3097700.82 frames. ], batch size: 190, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:10:12,457 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170325.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:10:22,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9535, 5.0067, 4.8087, 4.4964, 4.4784, 4.8912, 4.8111, 4.5200], device='cuda:0'), covar=tensor([0.0628, 0.0453, 0.0313, 0.0304, 0.0941, 0.0464, 0.0330, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0266, 0.0378, 0.0316, 0.0302, 0.0324, 0.0352, 0.0218, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:10:36,116 INFO [optim.py:368] (0/8) 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 14:10:55,572 INFO [train.py:904] (0/8) Epoch 17, batch 7950, loss[loss=0.1971, simple_loss=0.276, pruned_loss=0.05904, over 16748.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2941, pruned_loss=0.06267, over 3092594.98 frames. ], batch size: 57, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:11:06,309 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.60 vs. limit=2.0 2023-04-30 14:11:31,458 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8700, 2.7545, 2.7548, 2.0930, 2.6371, 2.0498, 2.7417, 2.9456], device='cuda:0'), covar=tensor([0.0285, 0.0771, 0.0509, 0.1747, 0.0782, 0.0958, 0.0590, 0.0714], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0141, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 14:11:49,412 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170385.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:12,699 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170401.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:12:14,155 INFO [train.py:904] (0/8) Epoch 17, batch 8000, loss[loss=0.2162, simple_loss=0.3158, pruned_loss=0.05833, over 16783.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.2951, pruned_loss=0.0644, over 3057916.16 frames. ], batch size: 102, lr: 3.96e-03, grad_scale: 8.0 2023-04-30 14:12:15,944 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170403.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:00,980 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 14:13:15,188 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170442.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:28,430 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170451.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:13:29,855 INFO [train.py:904] (0/8) Epoch 17, batch 8050, loss[loss=0.2007, simple_loss=0.2852, pruned_loss=0.05812, over 16676.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2952, pruned_loss=0.06436, over 3047200.89 frames. ], batch size: 57, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:13:43,767 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170461.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:28,129 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170490.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:14:37,738 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.43 vs. limit=5.0 2023-04-30 14:14:46,142 INFO [train.py:904] (0/8) Epoch 17, batch 8100, loss[loss=0.1831, simple_loss=0.2669, pruned_loss=0.04967, over 17187.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2946, pruned_loss=0.06379, over 3046737.10 frames. ], batch size: 45, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:15:16,566 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170522.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 14:15:42,386 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.803e+02 2.570e+02 3.195e+02 3.934e+02 7.205e+02, threshold=6.390e+02, percent-clipped=3.0 2023-04-30 14:16:00,571 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 14:16:00,989 INFO [train.py:904] (0/8) Epoch 17, batch 8150, loss[loss=0.1788, simple_loss=0.2665, pruned_loss=0.04554, over 16917.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2926, pruned_loss=0.06299, over 3061470.62 frames. ], batch size: 96, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:16:27,590 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170569.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:16:36,795 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 14:17:14,966 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.69 vs. limit=5.0 2023-04-30 14:17:18,499 INFO [train.py:904] (0/8) Epoch 17, batch 8200, loss[loss=0.2183, simple_loss=0.3129, pruned_loss=0.06181, over 16858.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2905, pruned_loss=0.06219, over 3073702.37 frames. ], batch size: 116, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:18:05,335 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170630.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:18:21,307 INFO [optim.py:368] (0/8) 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 14:18:41,080 INFO [train.py:904] (0/8) Epoch 17, batch 8250, loss[loss=0.2038, simple_loss=0.2942, pruned_loss=0.05668, over 16872.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2896, pruned_loss=0.05934, over 3084078.87 frames. ], batch size: 116, lr: 3.96e-03, grad_scale: 4.0 2023-04-30 14:20:04,078 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=170701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:20:05,306 INFO [train.py:904] (0/8) Epoch 17, batch 8300, loss[loss=0.1744, simple_loss=0.2779, pruned_loss=0.0355, over 16726.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2868, pruned_loss=0.05633, over 3065458.37 frames. ], batch size: 76, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:20:45,023 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-04-30 14:21:09,572 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.327e+02 2.774e+02 3.132e+02 7.452e+02, threshold=5.547e+02, percent-clipped=1.0 2023-04-30 14:21:25,415 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=170749.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:21:29,853 INFO [train.py:904] (0/8) Epoch 17, batch 8350, loss[loss=0.1998, simple_loss=0.2763, pruned_loss=0.06167, over 12069.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2861, pruned_loss=0.05442, over 3055654.13 frames. ], batch size: 247, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:22:51,865 INFO [train.py:904] (0/8) Epoch 17, batch 8400, loss[loss=0.1815, simple_loss=0.2697, pruned_loss=0.04662, over 15307.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2834, pruned_loss=0.05231, over 3048676.11 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:23:13,252 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 14:23:16,445 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170817.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 14:23:16,861 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 14:23:52,184 INFO [optim.py:368] (0/8) 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 14:24:10,760 INFO [train.py:904] (0/8) Epoch 17, batch 8450, loss[loss=0.1758, simple_loss=0.2705, pruned_loss=0.04052, over 15391.00 frames. ], tot_loss[loss=0.191, simple_loss=0.281, pruned_loss=0.05055, over 3041873.19 frames. ], batch size: 190, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:24:15,730 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0863, 5.0388, 4.8124, 4.3601, 4.9011, 2.0111, 4.6752, 4.7276], device='cuda:0'), covar=tensor([0.0058, 0.0062, 0.0154, 0.0254, 0.0073, 0.2231, 0.0102, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0134, 0.0180, 0.0163, 0.0154, 0.0190, 0.0167, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:24:41,460 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 14:25:32,480 INFO [train.py:904] (0/8) Epoch 17, batch 8500, loss[loss=0.1529, simple_loss=0.2336, pruned_loss=0.03611, over 11811.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2773, pruned_loss=0.04876, over 3003280.79 frames. ], batch size: 247, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:25:50,872 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-30 14:26:10,259 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=170925.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:14,637 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=170928.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:26:34,260 INFO [optim.py:368] (0/8) 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 14:26:56,007 INFO [train.py:904] (0/8) Epoch 17, batch 8550, loss[loss=0.1618, simple_loss=0.2469, pruned_loss=0.03837, over 12033.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2744, pruned_loss=0.047, over 3005735.22 frames. ], batch size: 246, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:28:08,368 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=170989.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:28:33,701 INFO [train.py:904] (0/8) Epoch 17, batch 8600, loss[loss=0.167, simple_loss=0.2521, pruned_loss=0.04096, over 12526.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2745, pruned_loss=0.04592, over 2995325.27 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:29:36,379 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-04-30 14:29:52,153 INFO [optim.py:368] (0/8) 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 14:30:15,676 INFO [train.py:904] (0/8) Epoch 17, batch 8650, loss[loss=0.1632, simple_loss=0.2543, pruned_loss=0.03611, over 12168.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2722, pruned_loss=0.04436, over 2979576.36 frames. ], batch size: 247, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:31:43,991 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171092.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:32:02,429 INFO [train.py:904] (0/8) Epoch 17, batch 8700, loss[loss=0.1738, simple_loss=0.2603, pruned_loss=0.04364, over 12397.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2692, pruned_loss=0.04279, over 3001422.26 frames. ], batch size: 248, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:32:19,472 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7276, 2.4525, 2.3833, 3.5788, 2.0603, 3.6570, 1.5128, 2.8311], device='cuda:0'), covar=tensor([0.1446, 0.0781, 0.1238, 0.0201, 0.0101, 0.0384, 0.1818, 0.0754], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0166, 0.0188, 0.0173, 0.0201, 0.0211, 0.0194, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 14:32:31,825 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171117.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 14:33:13,277 INFO [optim.py:368] (0/8) 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 14:33:23,653 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7920, 5.1245, 5.2171, 5.0189, 5.0217, 5.5994, 5.0793, 4.7798], device='cuda:0'), covar=tensor([0.0956, 0.1810, 0.2028, 0.1897, 0.2556, 0.0992, 0.1587, 0.2360], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0538, 0.0590, 0.0451, 0.0602, 0.0626, 0.0469, 0.0604], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 14:33:38,789 INFO [train.py:904] (0/8) Epoch 17, batch 8750, loss[loss=0.2136, simple_loss=0.3076, pruned_loss=0.05986, over 16171.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.269, pruned_loss=0.04224, over 3015874.30 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 4.0 2023-04-30 14:33:43,350 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171165.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:34:22,951 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-04-30 14:34:39,108 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-04-30 14:35:21,373 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171196.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:35:33,196 INFO [train.py:904] (0/8) Epoch 17, batch 8800, loss[loss=0.1753, simple_loss=0.2696, pruned_loss=0.04044, over 16764.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2676, pruned_loss=0.04105, over 3032530.12 frames. ], batch size: 124, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:36:03,222 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6428, 1.7370, 2.2005, 2.5927, 2.5136, 2.9946, 1.8334, 2.8638], device='cuda:0'), covar=tensor([0.0224, 0.0495, 0.0332, 0.0290, 0.0293, 0.0185, 0.0521, 0.0167], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0183, 0.0170, 0.0172, 0.0183, 0.0139, 0.0185, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:36:22,100 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171225.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:36:54,983 INFO [optim.py:368] (0/8) 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 14:37:17,927 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 14:37:18,453 INFO [train.py:904] (0/8) Epoch 17, batch 8850, loss[loss=0.174, simple_loss=0.283, pruned_loss=0.03247, over 16167.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2711, pruned_loss=0.04049, over 3056853.62 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:37:22,654 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 14:37:28,396 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:03,153 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171273.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:38:12,796 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1209, 2.0676, 2.1208, 3.7859, 1.9825, 2.3836, 2.1998, 2.1893], device='cuda:0'), covar=tensor([0.1219, 0.3836, 0.3077, 0.0490, 0.4668, 0.2736, 0.3774, 0.3717], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0418, 0.0348, 0.0311, 0.0422, 0.0479, 0.0389, 0.0484], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:38:27,226 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171284.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:39:01,949 INFO [train.py:904] (0/8) Epoch 17, batch 8900, loss[loss=0.1742, simple_loss=0.2822, pruned_loss=0.03315, over 16904.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2719, pruned_loss=0.03996, over 3060718.80 frames. ], batch size: 96, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:40:38,367 INFO [optim.py:368] (0/8) 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 14:40:50,860 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171345.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:41:04,770 INFO [train.py:904] (0/8) Epoch 17, batch 8950, loss[loss=0.1567, simple_loss=0.2552, pruned_loss=0.0291, over 16879.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.271, pruned_loss=0.04002, over 3074935.33 frames. ], batch size: 96, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:41:06,007 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4902, 4.6041, 4.7439, 4.6272, 4.6871, 5.1682, 4.7125, 4.3934], device='cuda:0'), covar=tensor([0.1248, 0.1797, 0.1754, 0.1951, 0.2433, 0.0973, 0.1615, 0.2356], device='cuda:0'), in_proj_covar=tensor([0.0369, 0.0535, 0.0587, 0.0450, 0.0600, 0.0626, 0.0469, 0.0601], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 14:42:53,003 INFO [train.py:904] (0/8) Epoch 17, batch 9000, loss[loss=0.1626, simple_loss=0.2579, pruned_loss=0.03371, over 16097.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2679, pruned_loss=0.03846, over 3087917.57 frames. ], batch size: 165, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:42:53,004 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 14:43:02,952 INFO [train.py:938] (0/8) Epoch 17, validation: loss=0.1478, simple_loss=0.2519, pruned_loss=0.02187, over 944034.00 frames. 2023-04-30 14:43:02,953 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 14:43:11,001 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:44:23,321 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 2.012e+02 2.627e+02 3.134e+02 6.797e+02, threshold=5.255e+02, percent-clipped=3.0 2023-04-30 14:44:41,005 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171448.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:44:47,400 INFO [train.py:904] (0/8) Epoch 17, batch 9050, loss[loss=0.1851, simple_loss=0.2722, pruned_loss=0.049, over 12981.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2679, pruned_loss=0.03922, over 3058963.89 frames. ], batch size: 246, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:45:32,463 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1733, 2.6197, 2.7682, 1.9430, 2.8752, 2.9456, 2.5601, 2.5178], device='cuda:0'), covar=tensor([0.0634, 0.0200, 0.0172, 0.0891, 0.0083, 0.0187, 0.0386, 0.0401], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0101, 0.0088, 0.0133, 0.0071, 0.0112, 0.0119, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 14:46:22,876 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 14:46:24,585 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:46:30,495 INFO [train.py:904] (0/8) Epoch 17, batch 9100, loss[loss=0.187, simple_loss=0.2798, pruned_loss=0.04711, over 16921.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2677, pruned_loss=0.03964, over 3072173.97 frames. ], batch size: 109, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:47:02,090 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4294, 2.4362, 2.0255, 2.2929, 2.8734, 2.5045, 3.0243, 3.0796], device='cuda:0'), covar=tensor([0.0125, 0.0452, 0.0590, 0.0469, 0.0272, 0.0418, 0.0216, 0.0252], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0218, 0.0211, 0.0212, 0.0217, 0.0216, 0.0215, 0.0208], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:47:33,876 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1154, 3.2401, 3.1682, 2.1954, 2.9673, 3.2261, 3.1139, 1.9468], device='cuda:0'), covar=tensor([0.0478, 0.0043, 0.0057, 0.0386, 0.0112, 0.0086, 0.0086, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0075, 0.0076, 0.0130, 0.0089, 0.0100, 0.0087, 0.0123], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 14:48:02,857 INFO [optim.py:368] (0/8) 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 14:48:28,380 INFO [train.py:904] (0/8) Epoch 17, batch 9150, loss[loss=0.1771, simple_loss=0.26, pruned_loss=0.04716, over 11896.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2678, pruned_loss=0.03961, over 3044676.10 frames. ], batch size: 250, lr: 3.95e-03, grad_scale: 8.0 2023-04-30 14:48:28,878 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171552.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:48:44,336 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171559.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:49:36,175 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:50:13,338 INFO [train.py:904] (0/8) Epoch 17, batch 9200, loss[loss=0.1522, simple_loss=0.239, pruned_loss=0.03275, over 12184.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2639, pruned_loss=0.0389, over 3047998.61 frames. ], batch size: 249, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:51:10,344 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171632.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:51:27,843 INFO [optim.py:368] (0/8) 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 14:51:42,670 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-04-30 14:51:50,826 INFO [train.py:904] (0/8) Epoch 17, batch 9250, loss[loss=0.1688, simple_loss=0.259, pruned_loss=0.0393, over 16754.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2636, pruned_loss=0.03914, over 3030786.73 frames. ], batch size: 134, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:53:16,193 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2825, 3.3397, 1.9097, 3.6445, 2.4045, 3.5975, 2.1691, 2.7532], device='cuda:0'), covar=tensor([0.0273, 0.0364, 0.1670, 0.0209, 0.0907, 0.0482, 0.1489, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0165, 0.0188, 0.0145, 0.0168, 0.0203, 0.0196, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 14:53:42,130 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:53:42,883 INFO [train.py:904] (0/8) Epoch 17, batch 9300, loss[loss=0.1396, simple_loss=0.2349, pruned_loss=0.02217, over 16508.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2621, pruned_loss=0.03864, over 3034379.38 frames. ], batch size: 68, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 14:55:09,820 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.061e+02 2.530e+02 2.950e+02 5.223e+02, threshold=5.060e+02, percent-clipped=1.0 2023-04-30 14:55:20,621 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171748.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:55:27,782 INFO [train.py:904] (0/8) Epoch 17, batch 9350, loss[loss=0.1878, simple_loss=0.278, pruned_loss=0.04884, over 16615.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2617, pruned_loss=0.03823, over 3036730.79 frames. ], batch size: 62, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:55:40,782 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4336, 3.3809, 3.5002, 3.5672, 3.6062, 3.3277, 3.5575, 3.6539], device='cuda:0'), covar=tensor([0.1337, 0.0882, 0.1080, 0.0645, 0.0646, 0.2524, 0.0990, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0565, 0.0691, 0.0817, 0.0708, 0.0536, 0.0558, 0.0571, 0.0667], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 14:56:57,424 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:57:07,774 INFO [train.py:904] (0/8) Epoch 17, batch 9400, loss[loss=0.178, simple_loss=0.2769, pruned_loss=0.03954, over 15509.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.263, pruned_loss=0.03838, over 3050363.34 frames. ], batch size: 191, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:29,671 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.619e+02 2.242e+02 2.701e+02 3.369e+02 7.708e+02, threshold=5.403e+02, percent-clipped=2.0 2023-04-30 14:58:48,378 INFO [train.py:904] (0/8) Epoch 17, batch 9450, loss[loss=0.189, simple_loss=0.2698, pruned_loss=0.05409, over 12317.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2646, pruned_loss=0.039, over 3036768.75 frames. ], batch size: 249, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 14:58:49,006 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=171852.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 14:58:52,143 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=171854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:00:01,057 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3914, 4.3338, 4.1850, 3.6869, 4.2817, 1.6802, 4.0462, 3.9153], device='cuda:0'), covar=tensor([0.0065, 0.0085, 0.0155, 0.0214, 0.0076, 0.2535, 0.0111, 0.0214], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0133, 0.0175, 0.0158, 0.0152, 0.0190, 0.0164, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:00:26,652 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=171900.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:00:31,040 INFO [train.py:904] (0/8) Epoch 17, batch 9500, loss[loss=0.1815, simple_loss=0.2736, pruned_loss=0.04469, over 16345.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2645, pruned_loss=0.03882, over 3054994.25 frames. ], batch size: 146, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:01:41,178 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=171937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:01:51,141 INFO [optim.py:368] (0/8) 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 15:02:14,963 INFO [train.py:904] (0/8) Epoch 17, batch 9550, loss[loss=0.1981, simple_loss=0.296, pruned_loss=0.05011, over 15263.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2642, pruned_loss=0.03877, over 3063083.15 frames. ], batch size: 190, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:02:55,372 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5747, 3.6597, 2.8549, 2.0800, 2.3942, 2.3364, 3.8769, 3.2283], device='cuda:0'), covar=tensor([0.2833, 0.0625, 0.1635, 0.2825, 0.2742, 0.1972, 0.0445, 0.1315], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0251, 0.0289, 0.0291, 0.0273, 0.0236, 0.0274, 0.0309], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:03:49,247 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=171998.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:03:52,389 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-172000.pt 2023-04-30 15:03:57,293 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 17, batch 9600, loss[loss=0.2062, simple_loss=0.3001, pruned_loss=0.05608, over 16737.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2652, pruned_loss=0.03949, over 3047661.00 frames. ], batch size: 83, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:04:12,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3305, 1.6149, 1.9530, 2.3436, 2.3519, 2.5615, 1.7955, 2.5403], device='cuda:0'), covar=tensor([0.0233, 0.0504, 0.0351, 0.0332, 0.0323, 0.0207, 0.0514, 0.0140], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0181, 0.0168, 0.0169, 0.0180, 0.0138, 0.0183, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:04:13,985 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8761, 1.8817, 2.3287, 2.8889, 2.6697, 3.1966, 2.0650, 3.1708], device='cuda:0'), covar=tensor([0.0218, 0.0501, 0.0379, 0.0305, 0.0312, 0.0167, 0.0493, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0181, 0.0168, 0.0169, 0.0180, 0.0138, 0.0183, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:04:31,443 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4855, 1.9931, 1.6733, 1.7396, 2.3345, 1.8877, 1.9593, 2.3657], device='cuda:0'), covar=tensor([0.0130, 0.0385, 0.0496, 0.0424, 0.0247, 0.0368, 0.0161, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0220, 0.0211, 0.0212, 0.0218, 0.0217, 0.0213, 0.0207], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:05:19,861 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7412, 4.7141, 4.5445, 4.0415, 4.5576, 1.7252, 4.3740, 4.4705], device='cuda:0'), covar=tensor([0.0091, 0.0095, 0.0177, 0.0338, 0.0124, 0.2582, 0.0132, 0.0212], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0132, 0.0175, 0.0158, 0.0151, 0.0189, 0.0163, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:05:20,543 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.253e+02 2.956e+02 3.661e+02 6.195e+02, threshold=5.912e+02, percent-clipped=3.0 2023-04-30 15:05:38,180 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172049.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:05:44,198 INFO [train.py:904] (0/8) Epoch 17, batch 9650, loss[loss=0.1735, simple_loss=0.273, pruned_loss=0.03694, over 16689.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2674, pruned_loss=0.03977, over 3063668.42 frames. ], batch size: 134, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:05:59,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 15:06:15,916 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4163, 3.3525, 3.5074, 3.5617, 3.6134, 3.3056, 3.5925, 3.6362], device='cuda:0'), covar=tensor([0.1294, 0.0948, 0.1085, 0.0660, 0.0622, 0.2493, 0.0809, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0562, 0.0690, 0.0810, 0.0706, 0.0530, 0.0557, 0.0569, 0.0663], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:07:32,492 INFO [train.py:904] (0/8) Epoch 17, batch 9700, loss[loss=0.1542, simple_loss=0.247, pruned_loss=0.03071, over 12171.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.266, pruned_loss=0.03947, over 3050678.74 frames. ], batch size: 248, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:08:05,528 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172118.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:08:57,403 INFO [optim.py:368] (0/8) 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 15:09:14,189 INFO [train.py:904] (0/8) Epoch 17, batch 9750, loss[loss=0.1701, simple_loss=0.2696, pruned_loss=0.03533, over 16399.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2653, pruned_loss=0.03967, over 3047294.78 frames. ], batch size: 146, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:09:18,394 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172154.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:09:41,137 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7839, 4.7419, 4.5072, 4.0701, 4.6296, 1.7377, 4.4304, 4.3507], device='cuda:0'), covar=tensor([0.0063, 0.0078, 0.0156, 0.0252, 0.0089, 0.2501, 0.0107, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0134, 0.0178, 0.0160, 0.0154, 0.0192, 0.0166, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:09:49,151 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9435, 4.4078, 4.3760, 3.3287, 3.8313, 4.3821, 3.9654, 2.9290], device='cuda:0'), covar=tensor([0.0408, 0.0026, 0.0029, 0.0264, 0.0085, 0.0055, 0.0059, 0.0323], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0075, 0.0075, 0.0131, 0.0089, 0.0099, 0.0087, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 15:10:09,718 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:10:54,107 INFO [train.py:904] (0/8) Epoch 17, batch 9800, loss[loss=0.1892, simple_loss=0.2916, pruned_loss=0.0434, over 16568.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2647, pruned_loss=0.03801, over 3073555.07 frames. ], batch size: 62, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:10:54,785 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172202.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:11:10,897 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3145, 3.1549, 2.8860, 3.3845, 3.3461, 3.2316, 3.4115, 3.4143], device='cuda:0'), covar=tensor([0.1311, 0.1476, 0.2427, 0.1197, 0.1342, 0.3435, 0.1679, 0.1664], device='cuda:0'), in_proj_covar=tensor([0.0558, 0.0685, 0.0805, 0.0705, 0.0529, 0.0554, 0.0566, 0.0660], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:12:17,140 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 2.047e+02 2.504e+02 3.250e+02 5.993e+02, threshold=5.009e+02, percent-clipped=1.0 2023-04-30 15:12:39,001 INFO [train.py:904] (0/8) Epoch 17, batch 9850, loss[loss=0.1632, simple_loss=0.2563, pruned_loss=0.03507, over 15394.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2658, pruned_loss=0.03805, over 3060161.73 frames. ], batch size: 191, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:14:13,573 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172293.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:14:32,004 INFO [train.py:904] (0/8) Epoch 17, batch 9900, loss[loss=0.1666, simple_loss=0.2695, pruned_loss=0.03188, over 16725.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2655, pruned_loss=0.03763, over 3056813.45 frames. ], batch size: 83, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:16:10,691 INFO [optim.py:368] (0/8) 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 15:16:31,657 INFO [train.py:904] (0/8) Epoch 17, batch 9950, loss[loss=0.1729, simple_loss=0.2685, pruned_loss=0.03863, over 16684.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2672, pruned_loss=0.03785, over 3062835.73 frames. ], batch size: 134, lr: 3.94e-03, grad_scale: 4.0 2023-04-30 15:17:05,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4712, 4.6280, 4.7896, 4.5791, 4.6219, 5.1583, 4.7325, 4.4490], device='cuda:0'), covar=tensor([0.1227, 0.1960, 0.2087, 0.2066, 0.2434, 0.1100, 0.1510, 0.2463], device='cuda:0'), in_proj_covar=tensor([0.0359, 0.0527, 0.0578, 0.0439, 0.0591, 0.0613, 0.0457, 0.0584], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 15:18:33,060 INFO [train.py:904] (0/8) Epoch 17, batch 10000, loss[loss=0.1783, simple_loss=0.2707, pruned_loss=0.04297, over 16774.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2657, pruned_loss=0.03731, over 3069158.70 frames. ], batch size: 124, lr: 3.94e-03, grad_scale: 8.0 2023-04-30 15:19:56,015 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.275e+02 2.112e+02 2.435e+02 3.073e+02 5.359e+02, threshold=4.869e+02, percent-clipped=1.0 2023-04-30 15:20:13,520 INFO [train.py:904] (0/8) Epoch 17, batch 10050, loss[loss=0.1713, simple_loss=0.2654, pruned_loss=0.03856, over 11982.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2653, pruned_loss=0.03704, over 3068580.54 frames. ], batch size: 248, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:20:19,913 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 15:20:30,434 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-04-30 15:20:58,139 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:21:11,194 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9940, 2.7013, 2.9398, 2.1666, 2.6881, 2.0760, 2.7104, 2.9368], device='cuda:0'), covar=tensor([0.0291, 0.0860, 0.0476, 0.1657, 0.0751, 0.0982, 0.0573, 0.0907], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0147, 0.0158, 0.0145, 0.0136, 0.0123, 0.0136, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 15:21:13,585 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1002, 2.0909, 2.1997, 3.5221, 2.0910, 2.3547, 2.2124, 2.2307], device='cuda:0'), covar=tensor([0.1150, 0.3638, 0.2811, 0.0549, 0.4123, 0.2581, 0.3480, 0.3449], device='cuda:0'), in_proj_covar=tensor([0.0370, 0.0411, 0.0343, 0.0307, 0.0413, 0.0468, 0.0380, 0.0475], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:21:47,157 INFO [train.py:904] (0/8) Epoch 17, batch 10100, loss[loss=0.1668, simple_loss=0.2588, pruned_loss=0.03744, over 16332.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2659, pruned_loss=0.03757, over 3061638.65 frames. ], batch size: 146, lr: 3.93e-03, grad_scale: 8.0 2023-04-30 15:22:55,171 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9793, 4.0791, 3.9052, 3.6082, 3.6097, 4.0156, 3.7061, 3.7667], device='cuda:0'), covar=tensor([0.0637, 0.0583, 0.0336, 0.0290, 0.0821, 0.0561, 0.0945, 0.0563], device='cuda:0'), in_proj_covar=tensor([0.0255, 0.0357, 0.0300, 0.0287, 0.0308, 0.0336, 0.0206, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-30 15:22:57,877 INFO [optim.py:368] (0/8) 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 15:23:09,139 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-17.pt 2023-04-30 15:23:33,090 INFO [train.py:904] (0/8) Epoch 18, batch 0, loss[loss=0.2235, simple_loss=0.317, pruned_loss=0.06505, over 17045.00 frames. ], tot_loss[loss=0.2235, simple_loss=0.317, pruned_loss=0.06505, over 17045.00 frames. ], batch size: 55, lr: 3.82e-03, grad_scale: 8.0 2023-04-30 15:23:33,091 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 15:23:40,344 INFO [train.py:938] (0/8) Epoch 18, validation: loss=0.148, simple_loss=0.2515, pruned_loss=0.02223, over 944034.00 frames. 2023-04-30 15:23:40,345 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 15:24:36,550 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172593.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:44,211 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:24:50,174 INFO [train.py:904] (0/8) Epoch 18, batch 50, loss[loss=0.1594, simple_loss=0.2516, pruned_loss=0.0336, over 17011.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.273, pruned_loss=0.05028, over 750930.41 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:25:21,095 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:42,964 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172641.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:25:48,587 INFO [optim.py:368] (0/8) 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 15:25:56,046 INFO [train.py:904] (0/8) Epoch 18, batch 100, loss[loss=0.192, simple_loss=0.2811, pruned_loss=0.0514, over 17072.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2725, pruned_loss=0.05168, over 1299639.84 frames. ], batch size: 55, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:26:06,001 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172659.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:26:44,090 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:27:02,724 INFO [train.py:904] (0/8) Epoch 18, batch 150, loss[loss=0.1672, simple_loss=0.2704, pruned_loss=0.03198, over 17053.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2701, pruned_loss=0.05033, over 1745294.81 frames. ], batch size: 50, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:27:23,093 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9267, 4.6518, 4.9395, 5.1236, 5.3317, 4.6720, 5.3229, 5.2878], device='cuda:0'), covar=tensor([0.1727, 0.1328, 0.1552, 0.0719, 0.0576, 0.0886, 0.0493, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0569, 0.0704, 0.0826, 0.0718, 0.0537, 0.0566, 0.0580, 0.0675], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:27:46,387 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:28:01,449 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 18, batch 200, loss[loss=0.1388, simple_loss=0.2249, pruned_loss=0.02638, over 16859.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2703, pruned_loss=0.04978, over 2099872.03 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:28:41,305 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=172774.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:28:44,899 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8681, 3.1339, 3.1287, 2.1050, 2.8086, 2.2700, 3.4019, 3.4301], device='cuda:0'), covar=tensor([0.0300, 0.0859, 0.0702, 0.1890, 0.0887, 0.0979, 0.0591, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0152, 0.0161, 0.0148, 0.0139, 0.0126, 0.0139, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 15:29:10,008 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:29:17,955 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7657, 2.7278, 2.2758, 2.6897, 3.1857, 2.9442, 3.4656, 3.3808], device='cuda:0'), covar=tensor([0.0174, 0.0456, 0.0592, 0.0440, 0.0306, 0.0353, 0.0287, 0.0286], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0229, 0.0220, 0.0221, 0.0228, 0.0227, 0.0226, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:29:18,649 INFO [train.py:904] (0/8) Epoch 18, batch 250, loss[loss=0.1519, simple_loss=0.2393, pruned_loss=0.03229, over 16991.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2675, pruned_loss=0.04913, over 2373777.52 frames. ], batch size: 41, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:29:47,729 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=172822.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:18,734 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=172845.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:30:19,549 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 2.240e+02 2.726e+02 3.152e+02 7.045e+02, threshold=5.451e+02, percent-clipped=2.0 2023-04-30 15:30:28,993 INFO [train.py:904] (0/8) Epoch 18, batch 300, loss[loss=0.1945, simple_loss=0.2727, pruned_loss=0.05812, over 16763.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2636, pruned_loss=0.04721, over 2577389.98 frames. ], batch size: 124, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:02,739 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9541, 4.4055, 3.2569, 2.4044, 2.8403, 2.5362, 4.6868, 3.6429], device='cuda:0'), covar=tensor([0.2618, 0.0553, 0.1604, 0.2637, 0.2594, 0.1924, 0.0356, 0.1273], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0257, 0.0296, 0.0297, 0.0281, 0.0242, 0.0281, 0.0317], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 15:31:03,938 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0322, 4.3314, 3.1664, 2.3501, 2.8444, 2.4999, 4.5646, 3.6849], device='cuda:0'), covar=tensor([0.2541, 0.0593, 0.1666, 0.2809, 0.2630, 0.2016, 0.0417, 0.1260], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0257, 0.0296, 0.0297, 0.0281, 0.0243, 0.0281, 0.0317], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 15:31:39,808 INFO [train.py:904] (0/8) Epoch 18, batch 350, loss[loss=0.1471, simple_loss=0.2356, pruned_loss=0.02934, over 16850.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2612, pruned_loss=0.04609, over 2748584.22 frames. ], batch size: 39, lr: 3.82e-03, grad_scale: 1.0 2023-04-30 15:31:45,368 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=172906.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:31:50,002 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4806, 5.8275, 5.5908, 5.6143, 5.2437, 5.2156, 5.2419, 5.9357], device='cuda:0'), covar=tensor([0.1193, 0.0991, 0.1038, 0.0869, 0.0893, 0.0700, 0.1214, 0.1007], device='cuda:0'), in_proj_covar=tensor([0.0634, 0.0774, 0.0625, 0.0573, 0.0489, 0.0492, 0.0644, 0.0597], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:32:10,627 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-04-30 15:32:40,487 INFO [optim.py:368] (0/8) 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 15:32:47,898 INFO [train.py:904] (0/8) Epoch 18, batch 400, loss[loss=0.1832, simple_loss=0.2601, pruned_loss=0.05318, over 16760.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.259, pruned_loss=0.04571, over 2876729.38 frames. ], batch size: 102, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:32:51,109 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172954.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:30,367 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=172981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:33:59,531 INFO [train.py:904] (0/8) Epoch 18, batch 450, loss[loss=0.1542, simple_loss=0.2441, pruned_loss=0.03217, over 16507.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.258, pruned_loss=0.0453, over 2978304.94 frames. ], batch size: 68, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:34:24,390 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1597, 2.2263, 2.7249, 3.0890, 2.9762, 3.6230, 2.4595, 3.4186], device='cuda:0'), covar=tensor([0.0204, 0.0431, 0.0278, 0.0245, 0.0267, 0.0158, 0.0426, 0.0165], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0184, 0.0170, 0.0172, 0.0182, 0.0141, 0.0186, 0.0135], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:35:00,082 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.073e+02 2.463e+02 2.995e+02 6.633e+02, threshold=4.925e+02, percent-clipped=1.0 2023-04-30 15:35:08,316 INFO [train.py:904] (0/8) Epoch 18, batch 500, loss[loss=0.1654, simple_loss=0.2541, pruned_loss=0.03832, over 17175.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2563, pruned_loss=0.04427, over 3043197.70 frames. ], batch size: 46, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:35:59,658 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173090.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:36:14,713 INFO [train.py:904] (0/8) Epoch 18, batch 550, loss[loss=0.1913, simple_loss=0.2625, pruned_loss=0.06002, over 16867.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2561, pruned_loss=0.04437, over 3113951.88 frames. ], batch size: 116, lr: 3.82e-03, grad_scale: 2.0 2023-04-30 15:37:14,572 INFO [optim.py:368] (0/8) 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 15:37:20,901 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173150.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:37:22,688 INFO [train.py:904] (0/8) Epoch 18, batch 600, loss[loss=0.177, simple_loss=0.2469, pruned_loss=0.05354, over 16461.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2552, pruned_loss=0.044, over 3167904.34 frames. ], batch size: 75, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:37:37,223 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-04-30 15:38:30,585 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173201.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:38:31,380 INFO [train.py:904] (0/8) Epoch 18, batch 650, loss[loss=0.1765, simple_loss=0.2492, pruned_loss=0.05194, over 16872.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2543, pruned_loss=0.04362, over 3193352.24 frames. ], batch size: 116, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:38:44,886 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173211.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:39:32,449 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.711e+02 2.266e+02 2.757e+02 3.739e+02 8.562e+02, threshold=5.513e+02, percent-clipped=7.0 2023-04-30 15:39:40,843 INFO [train.py:904] (0/8) Epoch 18, batch 700, loss[loss=0.1569, simple_loss=0.2485, pruned_loss=0.03265, over 17129.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.255, pruned_loss=0.04326, over 3224609.64 frames. ], batch size: 48, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:39:43,556 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173254.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:22,726 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:40:46,206 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6818, 3.7803, 2.3279, 3.9732, 2.9216, 3.9538, 2.2884, 2.9424], device='cuda:0'), covar=tensor([0.0262, 0.0314, 0.1453, 0.0361, 0.0703, 0.0664, 0.1405, 0.0700], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0173, 0.0195, 0.0155, 0.0174, 0.0213, 0.0203, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 15:40:50,346 INFO [train.py:904] (0/8) Epoch 18, batch 750, loss[loss=0.162, simple_loss=0.259, pruned_loss=0.03248, over 17038.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2546, pruned_loss=0.04288, over 3247835.40 frames. ], batch size: 50, lr: 3.81e-03, grad_scale: 2.0 2023-04-30 15:40:51,631 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173302.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:00,047 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:26,656 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:41:50,586 INFO [optim.py:368] (0/8) 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 15:41:58,284 INFO [train.py:904] (0/8) Epoch 18, batch 800, loss[loss=0.1689, simple_loss=0.2481, pruned_loss=0.04488, over 16849.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2538, pruned_loss=0.0425, over 3266079.73 frames. ], batch size: 109, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:42:22,650 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:42:30,196 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8027, 2.8626, 2.4919, 2.7967, 3.1426, 2.9715, 3.4855, 3.4175], device='cuda:0'), covar=tensor([0.0106, 0.0369, 0.0457, 0.0369, 0.0260, 0.0346, 0.0239, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0231, 0.0221, 0.0223, 0.0231, 0.0230, 0.0231, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:42:43,051 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 15:42:49,290 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173390.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:43:05,173 INFO [train.py:904] (0/8) Epoch 18, batch 850, loss[loss=0.1645, simple_loss=0.26, pruned_loss=0.03452, over 16633.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2536, pruned_loss=0.04219, over 3279284.14 frames. ], batch size: 62, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:43:55,003 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173438.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:44:03,723 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4704, 2.3264, 2.2325, 4.2569, 2.2213, 2.6630, 2.3498, 2.4455], device='cuda:0'), covar=tensor([0.1198, 0.3615, 0.3024, 0.0501, 0.4229, 0.2578, 0.3482, 0.3441], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0426, 0.0355, 0.0322, 0.0427, 0.0490, 0.0396, 0.0498], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:44:07,470 INFO [optim.py:368] (0/8) 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 15:44:15,666 INFO [train.py:904] (0/8) Epoch 18, batch 900, loss[loss=0.1591, simple_loss=0.2385, pruned_loss=0.03986, over 16719.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2537, pruned_loss=0.04211, over 3286909.13 frames. ], batch size: 83, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:44:40,067 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7119, 2.4144, 2.5075, 3.8461, 3.1724, 3.9798, 1.5388, 2.8911], device='cuda:0'), covar=tensor([0.1386, 0.0755, 0.1066, 0.0184, 0.0139, 0.0378, 0.1588, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0167, 0.0189, 0.0175, 0.0198, 0.0213, 0.0194, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 15:45:11,391 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173491.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:45:24,058 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:45:25,653 INFO [train.py:904] (0/8) Epoch 18, batch 950, loss[loss=0.1628, simple_loss=0.2472, pruned_loss=0.0392, over 17200.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2537, pruned_loss=0.04236, over 3293732.73 frames. ], batch size: 46, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:45:31,772 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173506.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:45:57,724 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9301, 2.9862, 2.6693, 2.9186, 3.2876, 3.1038, 3.6191, 3.5434], device='cuda:0'), covar=tensor([0.0134, 0.0392, 0.0481, 0.0373, 0.0271, 0.0326, 0.0269, 0.0252], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0234, 0.0224, 0.0225, 0.0234, 0.0232, 0.0233, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:46:26,063 INFO [optim.py:368] (0/8) 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 15:46:31,274 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173549.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:46:35,238 INFO [train.py:904] (0/8) Epoch 18, batch 1000, loss[loss=0.1523, simple_loss=0.2482, pruned_loss=0.02821, over 17051.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2526, pruned_loss=0.04224, over 3295215.51 frames. ], batch size: 50, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:46:35,685 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173552.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:47:44,533 INFO [train.py:904] (0/8) Epoch 18, batch 1050, loss[loss=0.1403, simple_loss=0.2341, pruned_loss=0.0233, over 16807.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2522, pruned_loss=0.04198, over 3300148.48 frames. ], batch size: 39, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:48:02,446 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=173614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:48:32,957 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.26 vs. limit=5.0 2023-04-30 15:48:46,282 INFO [optim.py:368] (0/8) 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 15:48:54,963 INFO [train.py:904] (0/8) Epoch 18, batch 1100, loss[loss=0.1426, simple_loss=0.2304, pruned_loss=0.02737, over 17218.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2515, pruned_loss=0.04198, over 3303739.47 frames. ], batch size: 45, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:49:12,480 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:49:26,659 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=173675.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 15:49:33,270 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8997, 1.3769, 1.6966, 1.7078, 1.8347, 1.9597, 1.6484, 1.7692], device='cuda:0'), covar=tensor([0.0211, 0.0383, 0.0199, 0.0266, 0.0251, 0.0187, 0.0380, 0.0121], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0187, 0.0172, 0.0176, 0.0184, 0.0144, 0.0190, 0.0138], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:49:34,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5089, 4.5914, 4.7370, 4.5518, 4.5516, 5.1494, 4.6674, 4.3189], device='cuda:0'), covar=tensor([0.1535, 0.1986, 0.2175, 0.2204, 0.2820, 0.1144, 0.1617, 0.2787], device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0577, 0.0630, 0.0477, 0.0646, 0.0661, 0.0498, 0.0645], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 15:50:01,551 INFO [train.py:904] (0/8) Epoch 18, batch 1150, loss[loss=0.1648, simple_loss=0.2685, pruned_loss=0.03058, over 17049.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2516, pruned_loss=0.04162, over 3304847.39 frames. ], batch size: 50, lr: 3.81e-03, grad_scale: 4.0 2023-04-30 15:51:02,217 INFO [optim.py:368] (0/8) 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 15:51:11,443 INFO [train.py:904] (0/8) Epoch 18, batch 1200, loss[loss=0.1665, simple_loss=0.2638, pruned_loss=0.03466, over 16624.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2505, pruned_loss=0.04099, over 3303967.19 frames. ], batch size: 57, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:51:12,943 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3371, 4.0848, 4.2386, 4.5188, 4.6433, 4.2740, 4.5792, 4.6094], device='cuda:0'), covar=tensor([0.1612, 0.1617, 0.1928, 0.0948, 0.0843, 0.1343, 0.1969, 0.0927], device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0763, 0.0897, 0.0773, 0.0576, 0.0610, 0.0625, 0.0729], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:51:59,992 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9218, 3.0422, 2.8523, 5.0531, 4.2876, 4.5826, 1.8623, 3.4722], device='cuda:0'), covar=tensor([0.1254, 0.0684, 0.1089, 0.0186, 0.0223, 0.0360, 0.1440, 0.0657], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0167, 0.0188, 0.0176, 0.0198, 0.0213, 0.0193, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 15:52:12,498 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9519, 2.0783, 2.4428, 2.8211, 2.7777, 2.9422, 2.1137, 2.9910], device='cuda:0'), covar=tensor([0.0175, 0.0416, 0.0317, 0.0270, 0.0278, 0.0217, 0.0462, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0188, 0.0175, 0.0177, 0.0186, 0.0145, 0.0191, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:52:19,923 INFO [train.py:904] (0/8) Epoch 18, batch 1250, loss[loss=0.1689, simple_loss=0.2644, pruned_loss=0.03672, over 16675.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.251, pruned_loss=0.04117, over 3297716.73 frames. ], batch size: 57, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:52:24,708 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173806.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:52:59,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3861, 5.3352, 5.2433, 4.7414, 4.8029, 5.2857, 5.2452, 4.8921], device='cuda:0'), covar=tensor([0.0623, 0.0460, 0.0299, 0.0330, 0.1160, 0.0426, 0.0296, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0286, 0.0402, 0.0335, 0.0324, 0.0347, 0.0376, 0.0230, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:53:21,079 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173847.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:53:28,950 INFO [train.py:904] (0/8) Epoch 18, batch 1300, loss[loss=0.1433, simple_loss=0.2223, pruned_loss=0.03217, over 16168.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2507, pruned_loss=0.04144, over 3311392.95 frames. ], batch size: 36, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:53:32,737 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=173854.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:54:37,703 INFO [train.py:904] (0/8) Epoch 18, batch 1350, loss[loss=0.1639, simple_loss=0.2423, pruned_loss=0.04279, over 16425.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2516, pruned_loss=0.04136, over 3304353.61 frames. ], batch size: 146, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:54:44,048 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4957, 2.3709, 2.4401, 4.2387, 2.3017, 2.7014, 2.4073, 2.5380], device='cuda:0'), covar=tensor([0.1197, 0.3436, 0.2846, 0.0537, 0.4067, 0.2394, 0.3464, 0.3290], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0429, 0.0358, 0.0325, 0.0428, 0.0493, 0.0399, 0.0501], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:55:14,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2443, 3.2158, 3.3284, 2.3849, 3.0771, 3.3921, 3.1198, 1.9038], device='cuda:0'), covar=tensor([0.0464, 0.0109, 0.0062, 0.0363, 0.0109, 0.0106, 0.0108, 0.0468], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0079, 0.0079, 0.0133, 0.0093, 0.0103, 0.0091, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 15:55:37,173 INFO [optim.py:368] (0/8) 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 15:55:44,252 INFO [train.py:904] (0/8) Epoch 18, batch 1400, loss[loss=0.1837, simple_loss=0.258, pruned_loss=0.05467, over 16815.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2513, pruned_loss=0.0414, over 3307324.63 frames. ], batch size: 124, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:56:04,297 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=173965.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:56:09,772 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=173970.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 15:56:26,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7717, 2.9377, 2.8083, 4.9275, 3.9936, 4.3938, 1.7787, 3.2214], device='cuda:0'), covar=tensor([0.1357, 0.0718, 0.1122, 0.0189, 0.0234, 0.0403, 0.1500, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0167, 0.0187, 0.0175, 0.0197, 0.0212, 0.0192, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 15:56:38,037 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 15:56:51,147 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-174000.pt 2023-04-30 15:56:56,570 INFO [train.py:904] (0/8) Epoch 18, batch 1450, loss[loss=0.1607, simple_loss=0.2501, pruned_loss=0.03569, over 17291.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2506, pruned_loss=0.0412, over 3318181.71 frames. ], batch size: 52, lr: 3.81e-03, grad_scale: 8.0 2023-04-30 15:57:11,463 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174013.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:14,007 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:16,968 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:57:35,045 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8516, 2.5344, 1.9307, 2.3543, 2.8972, 2.6686, 2.9612, 2.9815], device='cuda:0'), covar=tensor([0.0225, 0.0363, 0.0555, 0.0401, 0.0223, 0.0314, 0.0206, 0.0273], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0234, 0.0223, 0.0225, 0.0233, 0.0232, 0.0233, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 15:57:54,755 INFO [optim.py:368] (0/8) 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 15:58:03,194 INFO [train.py:904] (0/8) Epoch 18, batch 1500, loss[loss=0.1645, simple_loss=0.2572, pruned_loss=0.03583, over 17063.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2502, pruned_loss=0.04093, over 3314656.84 frames. ], batch size: 50, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 15:58:36,167 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174076.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 15:58:38,864 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174078.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 15:59:10,941 INFO [train.py:904] (0/8) Epoch 18, batch 1550, loss[loss=0.1973, simple_loss=0.2784, pruned_loss=0.05809, over 15464.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2513, pruned_loss=0.04205, over 3315760.54 frames. ], batch size: 190, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 15:59:13,362 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2597, 3.6251, 3.8361, 2.1608, 3.1679, 2.4411, 3.6945, 3.7799], device='cuda:0'), covar=tensor([0.0318, 0.0869, 0.0473, 0.1902, 0.0757, 0.0903, 0.0650, 0.1015], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0158, 0.0164, 0.0151, 0.0141, 0.0126, 0.0141, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:00:00,472 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:00:09,837 INFO [optim.py:368] (0/8) 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 16:00:11,898 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174147.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:00:17,421 INFO [train.py:904] (0/8) Epoch 18, batch 1600, loss[loss=0.1532, simple_loss=0.2345, pruned_loss=0.03599, over 16970.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.253, pruned_loss=0.04301, over 3314282.99 frames. ], batch size: 41, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:16,144 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174195.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:01:23,757 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174200.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:01:25,621 INFO [train.py:904] (0/8) Epoch 18, batch 1650, loss[loss=0.156, simple_loss=0.2541, pruned_loss=0.02894, over 17115.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2549, pruned_loss=0.04367, over 3317369.39 frames. ], batch size: 47, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:01:38,297 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7592, 3.9190, 3.1141, 2.3185, 2.6110, 2.4708, 4.0528, 3.4507], device='cuda:0'), covar=tensor([0.2673, 0.0647, 0.1557, 0.2730, 0.2570, 0.1915, 0.0483, 0.1386], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0262, 0.0298, 0.0300, 0.0288, 0.0245, 0.0284, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:02:23,447 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 16:02:23,841 INFO [optim.py:368] (0/8) 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 16:02:32,769 INFO [train.py:904] (0/8) Epoch 18, batch 1700, loss[loss=0.173, simple_loss=0.2607, pruned_loss=0.04265, over 16885.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2569, pruned_loss=0.04418, over 3320988.30 frames. ], batch size: 90, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:02:57,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3356, 4.4467, 4.7614, 4.7150, 4.7627, 4.4407, 4.4781, 4.2947], device='cuda:0'), covar=tensor([0.0372, 0.0554, 0.0382, 0.0416, 0.0502, 0.0434, 0.0798, 0.0611], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0436, 0.0423, 0.0397, 0.0471, 0.0447, 0.0540, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 16:02:57,317 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174270.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 16:03:15,575 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:24,470 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174291.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:03:40,507 INFO [train.py:904] (0/8) Epoch 18, batch 1750, loss[loss=0.1891, simple_loss=0.2661, pruned_loss=0.05606, over 16436.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2576, pruned_loss=0.04367, over 3320751.35 frames. ], batch size: 146, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:01,307 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:04:07,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7981, 2.7887, 2.7122, 4.2916, 3.4707, 4.1877, 1.6580, 2.9816], device='cuda:0'), covar=tensor([0.1345, 0.0646, 0.1034, 0.0158, 0.0170, 0.0410, 0.1474, 0.0802], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0167, 0.0188, 0.0176, 0.0199, 0.0214, 0.0192, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:04:36,736 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 16:04:41,362 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6861, 2.8178, 2.8906, 4.8889, 3.7373, 4.4723, 1.7318, 3.2029], device='cuda:0'), covar=tensor([0.1544, 0.0895, 0.1226, 0.0237, 0.0269, 0.0408, 0.1711, 0.0831], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0168, 0.0189, 0.0177, 0.0199, 0.0214, 0.0193, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:04:46,178 INFO [train.py:904] (0/8) Epoch 18, batch 1800, loss[loss=0.1596, simple_loss=0.2601, pruned_loss=0.02961, over 17267.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2579, pruned_loss=0.04312, over 3330380.96 frames. ], batch size: 52, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:04:46,570 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174352.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 16:05:10,914 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174371.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 16:05:12,342 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8721, 1.8666, 2.4085, 2.8628, 2.6193, 3.1658, 2.1668, 3.1798], device='cuda:0'), covar=tensor([0.0214, 0.0460, 0.0326, 0.0250, 0.0309, 0.0179, 0.0464, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0190, 0.0176, 0.0179, 0.0187, 0.0147, 0.0192, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:05:13,914 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174373.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:33,376 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:05:51,533 INFO [train.py:904] (0/8) Epoch 18, batch 1850, loss[loss=0.1908, simple_loss=0.2614, pruned_loss=0.06013, over 16470.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2592, pruned_loss=0.0437, over 3321598.06 frames. ], batch size: 146, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:06:28,451 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6020, 4.7058, 4.8699, 4.6389, 4.6927, 5.3155, 4.8133, 4.4614], device='cuda:0'), covar=tensor([0.1572, 0.2032, 0.2018, 0.2079, 0.2696, 0.1048, 0.1509, 0.2515], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0581, 0.0637, 0.0480, 0.0652, 0.0665, 0.0501, 0.0651], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:06:51,037 INFO [optim.py:368] (0/8) 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 16:06:55,573 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174449.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:06:58,376 INFO [train.py:904] (0/8) Epoch 18, batch 1900, loss[loss=0.164, simple_loss=0.2454, pruned_loss=0.04124, over 16508.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2585, pruned_loss=0.04298, over 3317510.16 frames. ], batch size: 146, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:07:05,734 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1836, 1.5130, 1.9635, 2.1301, 2.1839, 2.2722, 1.6837, 2.2895], device='cuda:0'), covar=tensor([0.0198, 0.0492, 0.0266, 0.0294, 0.0287, 0.0247, 0.0463, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0190, 0.0176, 0.0180, 0.0188, 0.0147, 0.0193, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:07:05,803 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8194, 1.8791, 2.1645, 3.3845, 1.9424, 2.0541, 2.0464, 1.9971], device='cuda:0'), covar=tensor([0.1711, 0.4533, 0.3018, 0.0806, 0.4845, 0.3256, 0.4088, 0.4106], device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0431, 0.0359, 0.0326, 0.0431, 0.0497, 0.0401, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:07:40,427 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5411, 3.7246, 3.9870, 2.2883, 3.2122, 2.4947, 3.9607, 3.9411], device='cuda:0'), covar=tensor([0.0253, 0.0898, 0.0465, 0.1862, 0.0813, 0.0946, 0.0596, 0.0977], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0157, 0.0163, 0.0150, 0.0141, 0.0126, 0.0141, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:07:47,715 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5134, 4.8236, 4.6193, 4.6212, 4.3948, 4.3154, 4.3017, 4.9050], device='cuda:0'), covar=tensor([0.1205, 0.0857, 0.0939, 0.0817, 0.0743, 0.1400, 0.1157, 0.0804], device='cuda:0'), in_proj_covar=tensor([0.0656, 0.0807, 0.0648, 0.0598, 0.0507, 0.0508, 0.0670, 0.0621], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:07:57,458 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:08:06,702 INFO [train.py:904] (0/8) Epoch 18, batch 1950, loss[loss=0.1666, simple_loss=0.2579, pruned_loss=0.03768, over 17112.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2583, pruned_loss=0.04259, over 3322526.98 frames. ], batch size: 53, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:09:05,470 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.435e+02 2.251e+02 2.513e+02 3.165e+02 5.916e+02, threshold=5.027e+02, percent-clipped=3.0 2023-04-30 16:09:14,568 INFO [train.py:904] (0/8) Epoch 18, batch 2000, loss[loss=0.1796, simple_loss=0.2525, pruned_loss=0.05334, over 16848.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2582, pruned_loss=0.04256, over 3320830.77 frames. ], batch size: 116, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:10:11,423 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9500, 4.0358, 2.5146, 4.5797, 3.2097, 4.5488, 2.5558, 3.2424], device='cuda:0'), covar=tensor([0.0261, 0.0334, 0.1501, 0.0253, 0.0711, 0.0420, 0.1501, 0.0712], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0159, 0.0174, 0.0218, 0.0202, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:10:23,408 INFO [train.py:904] (0/8) Epoch 18, batch 2050, loss[loss=0.1719, simple_loss=0.2498, pruned_loss=0.047, over 16730.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2582, pruned_loss=0.04285, over 3310360.84 frames. ], batch size: 134, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:11:17,520 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:11:26,171 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 16:11:26,653 INFO [optim.py:368] (0/8) 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 16:11:28,697 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174647.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:11:35,064 INFO [train.py:904] (0/8) Epoch 18, batch 2100, loss[loss=0.1745, simple_loss=0.2576, pruned_loss=0.0457, over 16883.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2596, pruned_loss=0.04413, over 3315604.86 frames. ], batch size: 90, lr: 3.80e-03, grad_scale: 16.0 2023-04-30 16:12:02,084 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174671.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:12:05,612 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174673.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:12:45,805 INFO [train.py:904] (0/8) Epoch 18, batch 2150, loss[loss=0.1744, simple_loss=0.2725, pruned_loss=0.03819, over 17058.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2608, pruned_loss=0.04454, over 3308567.42 frames. ], batch size: 53, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:12:50,309 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=174705.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:10,640 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174719.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:12,891 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:45,383 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=174744.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:13:45,475 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8280, 3.7728, 3.9030, 4.0235, 4.0603, 3.6749, 3.9005, 4.0976], device='cuda:0'), covar=tensor([0.1562, 0.1057, 0.1157, 0.0617, 0.0647, 0.1897, 0.1872, 0.0687], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0772, 0.0913, 0.0788, 0.0583, 0.0622, 0.0637, 0.0743], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:13:49,521 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 18, batch 2200, loss[loss=0.1521, simple_loss=0.2397, pruned_loss=0.03222, over 17200.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2615, pruned_loss=0.04482, over 3294682.15 frames. ], batch size: 44, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:14:15,721 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=174766.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:14:51,004 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5825, 3.9596, 4.0941, 2.7877, 3.7352, 4.1643, 3.8129, 2.2862], device='cuda:0'), covar=tensor([0.0482, 0.0155, 0.0046, 0.0362, 0.0091, 0.0086, 0.0085, 0.0441], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0079, 0.0079, 0.0131, 0.0092, 0.0102, 0.0090, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:14:56,022 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:15:06,372 INFO [train.py:904] (0/8) Epoch 18, batch 2250, loss[loss=0.191, simple_loss=0.2765, pruned_loss=0.05275, over 16568.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2609, pruned_loss=0.04452, over 3294303.72 frames. ], batch size: 68, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:15:10,969 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1725, 5.1414, 4.8899, 4.4998, 5.0246, 1.7018, 4.7190, 4.8485], device='cuda:0'), covar=tensor([0.0073, 0.0082, 0.0206, 0.0346, 0.0093, 0.2874, 0.0131, 0.0190], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0146, 0.0192, 0.0173, 0.0167, 0.0202, 0.0181, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:15:25,038 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5165, 3.5943, 3.3130, 3.0059, 3.2128, 3.4821, 3.3087, 3.2999], device='cuda:0'), covar=tensor([0.0619, 0.0511, 0.0285, 0.0259, 0.0484, 0.0410, 0.1396, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0411, 0.0340, 0.0331, 0.0354, 0.0383, 0.0236, 0.0410], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:16:02,787 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:16:07,689 INFO [optim.py:368] (0/8) 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 16:16:14,126 INFO [train.py:904] (0/8) Epoch 18, batch 2300, loss[loss=0.1712, simple_loss=0.2538, pruned_loss=0.04432, over 16812.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2608, pruned_loss=0.04477, over 3307477.84 frames. ], batch size: 39, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:17:22,150 INFO [train.py:904] (0/8) Epoch 18, batch 2350, loss[loss=0.1555, simple_loss=0.2427, pruned_loss=0.03417, over 16038.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2619, pruned_loss=0.04568, over 3306818.11 frames. ], batch size: 35, lr: 3.80e-03, grad_scale: 8.0 2023-04-30 16:18:14,637 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:18:23,371 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 16:18:25,285 INFO [optim.py:368] (0/8) 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 16:18:25,683 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=174947.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 16:18:32,022 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2289, 4.1575, 4.1287, 3.8972, 3.9123, 4.2060, 3.8663, 3.9665], device='cuda:0'), covar=tensor([0.0623, 0.0726, 0.0285, 0.0260, 0.0637, 0.0451, 0.0785, 0.0600], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0410, 0.0339, 0.0331, 0.0352, 0.0382, 0.0235, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:18:32,780 INFO [train.py:904] (0/8) Epoch 18, batch 2400, loss[loss=0.1736, simple_loss=0.2588, pruned_loss=0.04422, over 16221.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2621, pruned_loss=0.04561, over 3301510.56 frames. ], batch size: 165, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:19:20,828 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:33,334 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=174995.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:19:42,209 INFO [train.py:904] (0/8) Epoch 18, batch 2450, loss[loss=0.1569, simple_loss=0.2546, pruned_loss=0.02957, over 17193.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.262, pruned_loss=0.04467, over 3301898.12 frames. ], batch size: 46, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:20:12,754 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0995, 2.1068, 2.7115, 3.1101, 2.8620, 3.5564, 2.4121, 3.5480], device='cuda:0'), covar=tensor([0.0222, 0.0487, 0.0292, 0.0294, 0.0304, 0.0166, 0.0458, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:20:24,398 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7883, 4.6653, 4.6918, 4.3679, 4.3759, 4.7304, 4.5280, 4.5004], device='cuda:0'), covar=tensor([0.0716, 0.0846, 0.0332, 0.0320, 0.0946, 0.0515, 0.0453, 0.0697], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0413, 0.0341, 0.0333, 0.0355, 0.0385, 0.0236, 0.0411], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:20:26,746 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9746, 3.9244, 4.3931, 1.9703, 4.6466, 4.6397, 3.3692, 3.5199], device='cuda:0'), covar=tensor([0.0718, 0.0216, 0.0175, 0.1204, 0.0057, 0.0182, 0.0370, 0.0381], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0107, 0.0095, 0.0140, 0.0077, 0.0123, 0.0126, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:20:31,933 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3142, 3.2703, 3.6004, 1.8436, 3.7333, 3.7091, 2.9739, 2.6678], device='cuda:0'), covar=tensor([0.0847, 0.0211, 0.0146, 0.1109, 0.0081, 0.0177, 0.0377, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0107, 0.0095, 0.0140, 0.0077, 0.0123, 0.0127, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:20:40,414 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175044.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:20:43,747 INFO [optim.py:368] (0/8) 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 16:20:51,502 INFO [train.py:904] (0/8) Epoch 18, batch 2500, loss[loss=0.1823, simple_loss=0.2655, pruned_loss=0.04953, over 16550.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2619, pruned_loss=0.0439, over 3311078.28 frames. ], batch size: 68, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:21:04,692 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175061.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:21:38,840 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9266, 4.0102, 3.0735, 2.4292, 2.6256, 2.4910, 4.1955, 3.5393], device='cuda:0'), covar=tensor([0.2336, 0.0634, 0.1633, 0.2606, 0.2790, 0.1954, 0.0461, 0.1339], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0266, 0.0302, 0.0303, 0.0293, 0.0248, 0.0288, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:21:46,603 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175092.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:22:00,372 INFO [train.py:904] (0/8) Epoch 18, batch 2550, loss[loss=0.141, simple_loss=0.2295, pruned_loss=0.02626, over 17014.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2619, pruned_loss=0.04394, over 3320839.56 frames. ], batch size: 41, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:22:13,110 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 16:22:25,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6077, 2.5152, 1.8557, 2.6279, 2.1328, 2.7622, 2.0262, 2.3656], device='cuda:0'), covar=tensor([0.0314, 0.0379, 0.1325, 0.0269, 0.0660, 0.0496, 0.1220, 0.0649], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0178, 0.0197, 0.0162, 0.0177, 0.0221, 0.0205, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:22:28,344 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5580, 3.5726, 3.9006, 2.7001, 3.5548, 3.9234, 3.6152, 2.2126], device='cuda:0'), covar=tensor([0.0429, 0.0228, 0.0054, 0.0359, 0.0101, 0.0091, 0.0093, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0080, 0.0080, 0.0132, 0.0094, 0.0104, 0.0091, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:22:30,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0122, 4.7157, 4.8519, 5.2251, 5.3608, 4.7229, 5.4051, 5.3605], device='cuda:0'), covar=tensor([0.1912, 0.1402, 0.2035, 0.0863, 0.0753, 0.1062, 0.0670, 0.0788], device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0772, 0.0908, 0.0790, 0.0583, 0.0623, 0.0637, 0.0740], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:22:51,970 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175139.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:23:03,051 INFO [optim.py:368] (0/8) 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 16:23:09,305 INFO [train.py:904] (0/8) Epoch 18, batch 2600, loss[loss=0.171, simple_loss=0.2544, pruned_loss=0.0438, over 16889.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2617, pruned_loss=0.04399, over 3310305.93 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:23:42,058 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3300, 2.2265, 1.7494, 1.8910, 2.5097, 2.2315, 2.3862, 2.5824], device='cuda:0'), covar=tensor([0.0221, 0.0353, 0.0513, 0.0454, 0.0218, 0.0334, 0.0207, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0234, 0.0223, 0.0223, 0.0234, 0.0232, 0.0237, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:23:59,685 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3736, 2.3316, 2.3338, 4.2392, 2.2381, 2.7068, 2.3667, 2.5231], device='cuda:0'), covar=tensor([0.1324, 0.3661, 0.2943, 0.0544, 0.4082, 0.2545, 0.3500, 0.3203], device='cuda:0'), in_proj_covar=tensor([0.0392, 0.0432, 0.0360, 0.0327, 0.0431, 0.0499, 0.0401, 0.0505], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:24:17,416 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 18, batch 2650, loss[loss=0.175, simple_loss=0.2655, pruned_loss=0.04229, over 16864.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2621, pruned_loss=0.04394, over 3310785.12 frames. ], batch size: 96, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:24:27,530 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175207.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:25:15,512 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7108, 4.7295, 5.0906, 5.0448, 5.1047, 4.7594, 4.7144, 4.5067], device='cuda:0'), covar=tensor([0.0291, 0.0596, 0.0393, 0.0415, 0.0524, 0.0411, 0.0984, 0.0545], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0432, 0.0420, 0.0395, 0.0467, 0.0444, 0.0539, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 16:25:22,166 INFO [optim.py:368] (0/8) 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 16:25:27,543 INFO [train.py:904] (0/8) Epoch 18, batch 2700, loss[loss=0.1762, simple_loss=0.2611, pruned_loss=0.04564, over 16846.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.262, pruned_loss=0.04357, over 3321569.75 frames. ], batch size: 102, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:25:49,514 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175268.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:26:36,868 INFO [train.py:904] (0/8) Epoch 18, batch 2750, loss[loss=0.1755, simple_loss=0.2704, pruned_loss=0.04027, over 17036.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2621, pruned_loss=0.04294, over 3330292.67 frames. ], batch size: 55, lr: 3.79e-03, grad_scale: 4.0 2023-04-30 16:26:37,340 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6990, 2.7767, 2.9088, 4.9101, 3.7472, 4.3886, 1.9158, 3.0838], device='cuda:0'), covar=tensor([0.1457, 0.0845, 0.1094, 0.0204, 0.0300, 0.0437, 0.1520, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0168, 0.0188, 0.0178, 0.0201, 0.0214, 0.0192, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:27:21,793 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9166, 2.0493, 2.5276, 2.8486, 2.7214, 3.3904, 2.2254, 3.3079], device='cuda:0'), covar=tensor([0.0228, 0.0435, 0.0279, 0.0322, 0.0298, 0.0175, 0.0447, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0191, 0.0177, 0.0182, 0.0190, 0.0149, 0.0194, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:27:40,032 INFO [optim.py:368] (0/8) 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 16:27:45,211 INFO [train.py:904] (0/8) Epoch 18, batch 2800, loss[loss=0.1465, simple_loss=0.2385, pruned_loss=0.0273, over 17112.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2616, pruned_loss=0.04297, over 3338421.86 frames. ], batch size: 47, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:27:58,706 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:11,751 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175370.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:45,672 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:28:47,862 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9655, 5.3368, 5.4978, 5.2442, 5.3598, 5.9334, 5.4025, 5.1227], device='cuda:0'), covar=tensor([0.0995, 0.1891, 0.2021, 0.2185, 0.2547, 0.1023, 0.1443, 0.2512], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0585, 0.0643, 0.0488, 0.0658, 0.0669, 0.0506, 0.0654], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:28:56,426 INFO [train.py:904] (0/8) Epoch 18, batch 2850, loss[loss=0.1949, simple_loss=0.2621, pruned_loss=0.06382, over 16900.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2613, pruned_loss=0.04327, over 3323334.06 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:29:05,424 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175409.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:15,093 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7803, 2.7094, 2.2736, 2.5908, 3.0871, 2.8223, 3.5011, 3.3167], device='cuda:0'), covar=tensor([0.0133, 0.0402, 0.0502, 0.0389, 0.0264, 0.0352, 0.0218, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0235, 0.0224, 0.0224, 0.0235, 0.0233, 0.0238, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:29:35,301 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175431.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:29:46,861 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6269, 2.5468, 2.0196, 2.4239, 2.8858, 2.6540, 3.2684, 3.1795], device='cuda:0'), covar=tensor([0.0164, 0.0416, 0.0606, 0.0454, 0.0298, 0.0392, 0.0274, 0.0276], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0235, 0.0225, 0.0225, 0.0236, 0.0234, 0.0240, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:29:58,777 INFO [optim.py:368] (0/8) 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 16:30:04,929 INFO [train.py:904] (0/8) Epoch 18, batch 2900, loss[loss=0.1707, simple_loss=0.2645, pruned_loss=0.03852, over 16498.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2602, pruned_loss=0.04366, over 3318822.40 frames. ], batch size: 68, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:30:08,913 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175455.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:45,660 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:30:50,675 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3034, 4.1953, 4.2024, 3.9611, 3.9945, 4.2918, 4.0311, 4.0701], device='cuda:0'), covar=tensor([0.0647, 0.0747, 0.0303, 0.0285, 0.0764, 0.0504, 0.0677, 0.0600], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0420, 0.0347, 0.0339, 0.0361, 0.0392, 0.0240, 0.0418], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:31:03,886 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175495.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:31:08,789 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1513, 2.1149, 2.3114, 3.7849, 2.1624, 2.4835, 2.2182, 2.2682], device='cuda:0'), covar=tensor([0.1347, 0.3729, 0.2625, 0.0600, 0.3606, 0.2450, 0.3533, 0.3159], device='cuda:0'), in_proj_covar=tensor([0.0392, 0.0432, 0.0359, 0.0327, 0.0430, 0.0499, 0.0401, 0.0505], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:31:13,719 INFO [train.py:904] (0/8) Epoch 18, batch 2950, loss[loss=0.1923, simple_loss=0.2647, pruned_loss=0.06002, over 16919.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.26, pruned_loss=0.04429, over 3315207.37 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:32:09,660 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:32:17,357 INFO [optim.py:368] (0/8) 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 16:32:23,321 INFO [train.py:904] (0/8) Epoch 18, batch 3000, loss[loss=0.2029, simple_loss=0.2768, pruned_loss=0.06449, over 16852.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2606, pruned_loss=0.04509, over 3324339.08 frames. ], batch size: 109, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:32:23,322 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 16:32:32,127 INFO [train.py:938] (0/8) Epoch 18, validation: loss=0.1359, simple_loss=0.2415, pruned_loss=0.01516, over 944034.00 frames. 2023-04-30 16:32:32,128 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 16:32:48,747 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175563.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:32:56,462 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 16:33:41,417 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5631, 6.0091, 5.7152, 5.7720, 5.3880, 5.3526, 5.3347, 6.1013], device='cuda:0'), covar=tensor([0.1245, 0.0824, 0.0910, 0.0776, 0.0873, 0.0606, 0.1140, 0.0865], device='cuda:0'), in_proj_covar=tensor([0.0664, 0.0817, 0.0658, 0.0603, 0.0511, 0.0515, 0.0678, 0.0631], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:33:42,216 INFO [train.py:904] (0/8) Epoch 18, batch 3050, loss[loss=0.1482, simple_loss=0.2401, pruned_loss=0.02818, over 17220.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2595, pruned_loss=0.04473, over 3330740.33 frames. ], batch size: 45, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:34:46,188 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 18, batch 3100, loss[loss=0.1797, simple_loss=0.2547, pruned_loss=0.05232, over 16929.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2589, pruned_loss=0.04447, over 3331737.71 frames. ], batch size: 116, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:01,379 INFO [train.py:904] (0/8) Epoch 18, batch 3150, loss[loss=0.1834, simple_loss=0.2574, pruned_loss=0.05469, over 16862.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2589, pruned_loss=0.0452, over 3324759.91 frames. ], batch size: 109, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:36:13,059 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175710.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:36:36,357 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:36:39,823 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0764, 5.0824, 4.8241, 4.2227, 4.9673, 1.9053, 4.6372, 4.7653], device='cuda:0'), covar=tensor([0.0091, 0.0085, 0.0214, 0.0417, 0.0101, 0.2604, 0.0151, 0.0190], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0147, 0.0195, 0.0176, 0.0169, 0.0203, 0.0185, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:37:05,955 INFO [optim.py:368] (0/8) 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 16:37:08,589 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175750.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:37:11,235 INFO [train.py:904] (0/8) Epoch 18, batch 3200, loss[loss=0.1806, simple_loss=0.2607, pruned_loss=0.05025, over 16816.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2586, pruned_loss=0.04467, over 3322777.18 frames. ], batch size: 102, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:37:37,212 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175771.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:11,036 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175795.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:38:20,682 INFO [train.py:904] (0/8) Epoch 18, batch 3250, loss[loss=0.1685, simple_loss=0.259, pruned_loss=0.03902, over 17117.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2582, pruned_loss=0.0443, over 3328527.82 frames. ], batch size: 47, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:38:45,637 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-04-30 16:39:09,524 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=175837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:17,731 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175843.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:39:23,830 INFO [optim.py:368] (0/8) 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 16:39:29,881 INFO [train.py:904] (0/8) Epoch 18, batch 3300, loss[loss=0.1576, simple_loss=0.244, pruned_loss=0.03558, over 17206.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2595, pruned_loss=0.04443, over 3322671.32 frames. ], batch size: 44, lr: 3.79e-03, grad_scale: 8.0 2023-04-30 16:39:45,199 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0620, 4.8450, 5.1038, 5.3044, 5.4724, 4.7838, 5.4200, 5.4620], device='cuda:0'), covar=tensor([0.1800, 0.1198, 0.1541, 0.0661, 0.0503, 0.1018, 0.0535, 0.0549], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0793, 0.0936, 0.0808, 0.0599, 0.0641, 0.0653, 0.0757], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:39:45,213 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=175863.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:40:11,009 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175881.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:40:38,821 INFO [train.py:904] (0/8) Epoch 18, batch 3350, loss[loss=0.152, simple_loss=0.2512, pruned_loss=0.02644, over 17236.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2593, pruned_loss=0.04396, over 3322460.09 frames. ], batch size: 45, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:40:50,865 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=175911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:08,409 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.58 vs. limit=5.0 2023-04-30 16:41:34,565 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=175942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:38,133 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=175945.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:41:41,871 INFO [optim.py:368] (0/8) 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 16:41:47,207 INFO [train.py:904] (0/8) Epoch 18, batch 3400, loss[loss=0.1633, simple_loss=0.2441, pruned_loss=0.0412, over 16323.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2582, pruned_loss=0.0435, over 3332616.56 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:41:52,755 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-04-30 16:41:56,178 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 16:42:08,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4241, 5.3491, 5.2536, 4.7713, 4.8948, 5.3245, 5.2628, 4.9249], device='cuda:0'), covar=tensor([0.0553, 0.0410, 0.0302, 0.0333, 0.1035, 0.0379, 0.0263, 0.0767], device='cuda:0'), in_proj_covar=tensor([0.0296, 0.0420, 0.0347, 0.0338, 0.0362, 0.0393, 0.0239, 0.0418], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:42:50,679 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 16:42:56,466 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-176000.pt 2023-04-30 16:43:01,869 INFO [train.py:904] (0/8) Epoch 18, batch 3450, loss[loss=0.1757, simple_loss=0.2551, pruned_loss=0.04818, over 16347.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2569, pruned_loss=0.04304, over 3329378.20 frames. ], batch size: 165, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:43:04,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7390, 2.5465, 2.4424, 3.6037, 2.8841, 3.8340, 1.6259, 2.7649], device='cuda:0'), covar=tensor([0.1396, 0.0748, 0.1159, 0.0228, 0.0164, 0.0388, 0.1629, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0169, 0.0188, 0.0181, 0.0203, 0.0215, 0.0193, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:43:07,967 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176006.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:43:35,378 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:06,116 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.158e+02 2.512e+02 2.983e+02 6.813e+02, threshold=5.023e+02, percent-clipped=2.0 2023-04-30 16:44:09,526 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:11,309 INFO [train.py:904] (0/8) Epoch 18, batch 3500, loss[loss=0.1898, simple_loss=0.2631, pruned_loss=0.05822, over 16451.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2568, pruned_loss=0.04299, over 3315684.94 frames. ], batch size: 146, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:44:31,126 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176066.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:44:41,814 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176074.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:03,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0538, 2.2763, 2.7059, 3.0949, 2.8492, 3.5740, 2.5524, 3.6124], device='cuda:0'), covar=tensor([0.0234, 0.0426, 0.0290, 0.0284, 0.0288, 0.0141, 0.0415, 0.0123], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0189, 0.0177, 0.0181, 0.0188, 0.0149, 0.0192, 0.0144], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:45:14,940 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176098.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:45:20,521 INFO [train.py:904] (0/8) Epoch 18, batch 3550, loss[loss=0.1632, simple_loss=0.2581, pruned_loss=0.03418, over 17109.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2554, pruned_loss=0.04243, over 3319929.34 frames. ], batch size: 49, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:46:10,656 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:46:26,487 INFO [optim.py:368] (0/8) 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 16:46:32,333 INFO [train.py:904] (0/8) Epoch 18, batch 3600, loss[loss=0.1918, simple_loss=0.2702, pruned_loss=0.05672, over 16524.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2553, pruned_loss=0.04247, over 3310528.28 frames. ], batch size: 146, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:47:18,735 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:47:43,325 INFO [train.py:904] (0/8) Epoch 18, batch 3650, loss[loss=0.1513, simple_loss=0.227, pruned_loss=0.03783, over 16360.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2547, pruned_loss=0.04306, over 3293190.58 frames. ], batch size: 68, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:47:43,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5424, 3.5539, 2.1929, 3.7905, 2.7911, 3.7387, 2.2465, 2.8932], device='cuda:0'), covar=tensor([0.0262, 0.0452, 0.1498, 0.0323, 0.0820, 0.0980, 0.1472, 0.0714], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0163, 0.0177, 0.0222, 0.0205, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:48:30,063 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7526, 2.5207, 2.4842, 3.8056, 3.2447, 3.9451, 1.5365, 2.7697], device='cuda:0'), covar=tensor([0.1347, 0.0742, 0.1147, 0.0195, 0.0149, 0.0342, 0.1560, 0.0824], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0169, 0.0189, 0.0182, 0.0203, 0.0215, 0.0194, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:48:35,862 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176237.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:43,992 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:48:48,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3608, 3.4617, 3.5935, 2.3073, 3.0385, 2.4870, 3.7932, 3.7286], device='cuda:0'), covar=tensor([0.0246, 0.0908, 0.0561, 0.1849, 0.0827, 0.0956, 0.0551, 0.0999], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0162, 0.0166, 0.0152, 0.0143, 0.0128, 0.0144, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:48:53,038 INFO [optim.py:368] (0/8) 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 16:48:58,536 INFO [train.py:904] (0/8) Epoch 18, batch 3700, loss[loss=0.1717, simple_loss=0.2468, pruned_loss=0.04835, over 16820.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2538, pruned_loss=0.04496, over 3280629.34 frames. ], batch size: 102, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:49:06,434 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:10,843 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176301.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:11,698 INFO [train.py:904] (0/8) Epoch 18, batch 3750, loss[loss=0.1956, simple_loss=0.2927, pruned_loss=0.04927, over 16750.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2552, pruned_loss=0.04681, over 3269774.85 frames. ], batch size: 57, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:50:13,739 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176303.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:50:35,190 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:51:17,343 INFO [optim.py:368] (0/8) 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 16:51:23,248 INFO [train.py:904] (0/8) Epoch 18, batch 3800, loss[loss=0.1645, simple_loss=0.2482, pruned_loss=0.04042, over 15515.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2555, pruned_loss=0.04766, over 3279254.26 frames. ], batch size: 190, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:51:44,217 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176366.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:52:31,399 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6748, 3.6881, 2.3146, 3.9756, 2.9694, 3.9804, 2.4105, 2.9439], device='cuda:0'), covar=tensor([0.0242, 0.0371, 0.1415, 0.0251, 0.0625, 0.0684, 0.1387, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0160, 0.0174, 0.0219, 0.0202, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:52:35,508 INFO [train.py:904] (0/8) Epoch 18, batch 3850, loss[loss=0.1725, simple_loss=0.2529, pruned_loss=0.04608, over 16666.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2554, pruned_loss=0.04835, over 3280216.61 frames. ], batch size: 57, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:52:37,411 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-04-30 16:52:53,512 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176414.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:52:54,700 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8189, 3.6630, 3.8650, 3.6429, 3.8071, 4.2452, 3.9401, 3.5965], device='cuda:0'), covar=tensor([0.2249, 0.2507, 0.2100, 0.2506, 0.2758, 0.2081, 0.1478, 0.2490], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0582, 0.0637, 0.0490, 0.0657, 0.0666, 0.0505, 0.0655], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 16:53:42,934 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.371e+02 2.342e+02 2.782e+02 3.184e+02 4.336e+02, threshold=5.564e+02, percent-clipped=0.0 2023-04-30 16:53:49,450 INFO [train.py:904] (0/8) Epoch 18, batch 3900, loss[loss=0.202, simple_loss=0.2724, pruned_loss=0.06578, over 16680.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2545, pruned_loss=0.04875, over 3282222.10 frames. ], batch size: 76, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:54:10,469 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176466.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:54:37,541 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1645, 3.9602, 4.5377, 2.4659, 4.8789, 4.8218, 3.5070, 3.5656], device='cuda:0'), covar=tensor([0.0639, 0.0243, 0.0148, 0.1047, 0.0039, 0.0076, 0.0334, 0.0385], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0106, 0.0094, 0.0138, 0.0076, 0.0122, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 16:55:00,910 INFO [train.py:904] (0/8) Epoch 18, batch 3950, loss[loss=0.179, simple_loss=0.2664, pruned_loss=0.04583, over 17117.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2542, pruned_loss=0.04876, over 3285834.28 frames. ], batch size: 48, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:55:10,799 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3010, 3.3309, 3.5378, 2.1647, 3.0033, 2.3623, 3.6977, 3.7099], device='cuda:0'), covar=tensor([0.0224, 0.0771, 0.0590, 0.1822, 0.0820, 0.0946, 0.0474, 0.0741], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0161, 0.0165, 0.0151, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:55:35,305 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176527.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:55:50,488 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176537.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:56:05,777 INFO [optim.py:368] (0/8) 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 16:56:12,336 INFO [train.py:904] (0/8) Epoch 18, batch 4000, loss[loss=0.1642, simple_loss=0.2495, pruned_loss=0.0395, over 16711.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2543, pruned_loss=0.04884, over 3284854.72 frames. ], batch size: 62, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:56:49,976 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7459, 5.0324, 4.8294, 4.8228, 4.6072, 4.5264, 4.4862, 5.1288], device='cuda:0'), covar=tensor([0.1102, 0.0819, 0.0958, 0.0800, 0.0717, 0.1080, 0.1067, 0.0861], device='cuda:0'), in_proj_covar=tensor([0.0656, 0.0808, 0.0655, 0.0603, 0.0507, 0.0515, 0.0673, 0.0627], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:57:00,439 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:18,438 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176598.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:22,932 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176601.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:24,379 INFO [train.py:904] (0/8) Epoch 18, batch 4050, loss[loss=0.1762, simple_loss=0.2642, pruned_loss=0.04415, over 17109.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2544, pruned_loss=0.04762, over 3289970.83 frames. ], batch size: 47, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:57:39,765 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176613.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:57:40,023 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5020, 3.4318, 3.8223, 2.2495, 3.2175, 2.4069, 3.8288, 3.8353], device='cuda:0'), covar=tensor([0.0176, 0.0763, 0.0480, 0.1800, 0.0746, 0.0889, 0.0469, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0161, 0.0165, 0.0151, 0.0142, 0.0127, 0.0143, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 16:57:57,935 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176625.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:30,485 INFO [optim.py:368] (0/8) 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 16:58:33,054 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176649.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:58:36,781 INFO [train.py:904] (0/8) Epoch 18, batch 4100, loss[loss=0.162, simple_loss=0.2486, pruned_loss=0.03771, over 16417.00 frames. ], tot_loss[loss=0.1754, simple_loss=0.2565, pruned_loss=0.0472, over 3277761.64 frames. ], batch size: 35, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 16:58:57,220 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176665.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:59:26,254 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9470, 1.9245, 2.1372, 3.3504, 1.9405, 2.2460, 2.0948, 2.0728], device='cuda:0'), covar=tensor([0.1493, 0.4067, 0.2862, 0.0752, 0.4894, 0.2884, 0.3602, 0.4194], device='cuda:0'), in_proj_covar=tensor([0.0392, 0.0437, 0.0359, 0.0328, 0.0430, 0.0503, 0.0405, 0.0510], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:59:28,803 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 16:59:44,288 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6325, 1.8777, 2.3340, 2.6500, 2.6232, 3.0266, 1.9305, 2.9232], device='cuda:0'), covar=tensor([0.0200, 0.0458, 0.0284, 0.0255, 0.0281, 0.0169, 0.0469, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0188, 0.0177, 0.0180, 0.0188, 0.0148, 0.0191, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 16:59:51,173 INFO [train.py:904] (0/8) Epoch 18, batch 4150, loss[loss=0.1906, simple_loss=0.2843, pruned_loss=0.04845, over 16359.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2627, pruned_loss=0.04951, over 3236318.57 frames. ], batch size: 146, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:00:27,799 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176726.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:00:51,099 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5208, 4.6146, 4.7677, 4.5487, 4.5293, 5.1217, 4.6713, 4.3165], device='cuda:0'), covar=tensor([0.1109, 0.1790, 0.1656, 0.2054, 0.2649, 0.1051, 0.1489, 0.2577], device='cuda:0'), in_proj_covar=tensor([0.0396, 0.0579, 0.0631, 0.0485, 0.0649, 0.0661, 0.0502, 0.0649], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:01:00,610 INFO [optim.py:368] (0/8) 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 17:01:06,249 INFO [train.py:904] (0/8) Epoch 18, batch 4200, loss[loss=0.2402, simple_loss=0.3233, pruned_loss=0.07861, over 16705.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2702, pruned_loss=0.05148, over 3201379.41 frames. ], batch size: 134, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:02:13,954 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 17:02:21,472 INFO [train.py:904] (0/8) Epoch 18, batch 4250, loss[loss=0.194, simple_loss=0.2703, pruned_loss=0.05884, over 11894.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2736, pruned_loss=0.05106, over 3188476.68 frames. ], batch size: 246, lr: 3.78e-03, grad_scale: 8.0 2023-04-30 17:02:24,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2941, 5.1838, 5.3091, 5.5007, 5.6773, 4.9341, 5.6189, 5.6859], device='cuda:0'), covar=tensor([0.1668, 0.1067, 0.1420, 0.0613, 0.0478, 0.0766, 0.0588, 0.0428], device='cuda:0'), in_proj_covar=tensor([0.0614, 0.0760, 0.0892, 0.0772, 0.0568, 0.0612, 0.0624, 0.0725], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:02:51,002 INFO [zipformer.py:625] (0/8) 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] (0/8) attn_weights_entropy = tensor([3.4120, 3.4244, 1.8958, 3.9532, 2.5650, 3.9476, 2.2242, 2.7208], device='cuda:0'), covar=tensor([0.0305, 0.0397, 0.1938, 0.0177, 0.0927, 0.0555, 0.1542, 0.0862], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0175, 0.0192, 0.0156, 0.0174, 0.0216, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 17:03:30,839 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.576e+02 2.197e+02 2.694e+02 3.150e+02 7.622e+02, threshold=5.388e+02, percent-clipped=1.0 2023-04-30 17:03:35,717 INFO [train.py:904] (0/8) Epoch 18, batch 4300, loss[loss=0.1902, simple_loss=0.2811, pruned_loss=0.04965, over 16792.00 frames. ], tot_loss[loss=0.1881, simple_loss=0.2755, pruned_loss=0.05038, over 3189911.74 frames. ], batch size: 39, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:03:37,843 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=176853.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:44,715 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176898.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:04:49,707 INFO [train.py:904] (0/8) Epoch 18, batch 4350, loss[loss=0.2097, simple_loss=0.2834, pruned_loss=0.06799, over 11732.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.279, pruned_loss=0.05121, over 3201756.25 frames. ], batch size: 246, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:04:55,948 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4027, 4.3816, 4.6099, 4.3586, 4.4945, 5.0092, 4.5328, 4.1521], device='cuda:0'), covar=tensor([0.1376, 0.1840, 0.1822, 0.2001, 0.2475, 0.1019, 0.1544, 0.2372], device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0572, 0.0624, 0.0478, 0.0642, 0.0654, 0.0497, 0.0640], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:05:06,485 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=176913.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:07,719 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=176914.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 17:05:55,205 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176946.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:05:57,201 INFO [optim.py:368] (0/8) 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 17:06:00,559 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8658, 2.7609, 2.6662, 1.9173, 2.5949, 2.6786, 2.5670, 1.9454], device='cuda:0'), covar=tensor([0.0396, 0.0066, 0.0064, 0.0337, 0.0109, 0.0102, 0.0109, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0078, 0.0079, 0.0132, 0.0094, 0.0104, 0.0091, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:06:03,309 INFO [train.py:904] (0/8) Epoch 18, batch 4400, loss[loss=0.1909, simple_loss=0.2819, pruned_loss=0.04996, over 16729.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2816, pruned_loss=0.05285, over 3177207.62 frames. ], batch size: 76, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:06:17,359 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=176961.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:06:45,575 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=176981.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:07:05,509 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1166, 2.0922, 2.6210, 3.0832, 2.8731, 3.4724, 2.0422, 3.3703], device='cuda:0'), covar=tensor([0.0154, 0.0402, 0.0246, 0.0206, 0.0234, 0.0128, 0.0486, 0.0106], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0188, 0.0177, 0.0180, 0.0188, 0.0148, 0.0192, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:07:16,206 INFO [train.py:904] (0/8) Epoch 18, batch 4450, loss[loss=0.2213, simple_loss=0.3059, pruned_loss=0.06839, over 16507.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2848, pruned_loss=0.0539, over 3187733.68 frames. ], batch size: 35, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:07:29,080 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5642, 3.7750, 2.9390, 2.2929, 2.5588, 2.3675, 4.1048, 3.3945], device='cuda:0'), covar=tensor([0.2850, 0.0633, 0.1628, 0.2452, 0.2494, 0.1941, 0.0400, 0.1139], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0266, 0.0301, 0.0304, 0.0294, 0.0248, 0.0288, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:07:43,893 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:08:24,608 INFO [optim.py:368] (0/8) 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 17:08:30,161 INFO [train.py:904] (0/8) Epoch 18, batch 4500, loss[loss=0.2097, simple_loss=0.2882, pruned_loss=0.06554, over 16646.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2854, pruned_loss=0.05472, over 3189034.51 frames. ], batch size: 62, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:43,623 INFO [train.py:904] (0/8) Epoch 18, batch 4550, loss[loss=0.1899, simple_loss=0.2783, pruned_loss=0.05076, over 16771.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2857, pruned_loss=0.05525, over 3202763.67 frames. ], batch size: 124, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:09:53,036 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177109.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:09:55,579 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:11,723 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177122.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:10:32,095 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-04-30 17:10:45,881 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9037, 2.7333, 1.8977, 2.2018, 3.1949, 2.7309, 3.4776, 3.5095], device='cuda:0'), covar=tensor([0.0076, 0.0396, 0.0659, 0.0512, 0.0241, 0.0373, 0.0221, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0230, 0.0221, 0.0220, 0.0230, 0.0229, 0.0232, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:10:48,536 INFO [optim.py:368] (0/8) 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 17:10:53,019 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0647, 4.1968, 4.4265, 4.3710, 4.3981, 4.0941, 4.1588, 4.0677], device='cuda:0'), covar=tensor([0.0303, 0.0441, 0.0370, 0.0405, 0.0422, 0.0407, 0.0779, 0.0480], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0413, 0.0405, 0.0380, 0.0450, 0.0426, 0.0519, 0.0337], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 17:10:53,842 INFO [train.py:904] (0/8) Epoch 18, batch 4600, loss[loss=0.189, simple_loss=0.281, pruned_loss=0.04846, over 16362.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2872, pruned_loss=0.05597, over 3210301.45 frames. ], batch size: 146, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:11:16,529 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6429, 2.5733, 1.8851, 2.7371, 2.2272, 2.7511, 2.0951, 2.3933], device='cuda:0'), covar=tensor([0.0278, 0.0346, 0.1236, 0.0187, 0.0610, 0.0370, 0.1154, 0.0562], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0174, 0.0192, 0.0153, 0.0173, 0.0214, 0.0200, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 17:11:19,987 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:20,156 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:11:22,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7363, 2.5811, 2.4362, 3.1257, 2.4039, 3.5388, 1.5358, 2.6730], device='cuda:0'), covar=tensor([0.1341, 0.0713, 0.1126, 0.0161, 0.0187, 0.0362, 0.1669, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0171, 0.0192, 0.0182, 0.0206, 0.0216, 0.0197, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 17:11:24,069 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:12:05,783 INFO [train.py:904] (0/8) Epoch 18, batch 4650, loss[loss=0.1954, simple_loss=0.2741, pruned_loss=0.05834, over 16773.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.287, pruned_loss=0.05664, over 3198800.16 frames. ], batch size: 39, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:12:15,472 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177209.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:13:10,935 INFO [optim.py:368] (0/8) 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 17:13:16,225 INFO [train.py:904] (0/8) Epoch 18, batch 4700, loss[loss=0.2427, simple_loss=0.3048, pruned_loss=0.09025, over 11599.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2843, pruned_loss=0.05555, over 3197759.34 frames. ], batch size: 248, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:13:59,455 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:14:17,243 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2568, 5.1244, 5.2523, 5.4625, 5.6747, 4.9836, 5.6101, 5.6397], device='cuda:0'), covar=tensor([0.1635, 0.1116, 0.1620, 0.0666, 0.0463, 0.0757, 0.0477, 0.0480], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0744, 0.0875, 0.0763, 0.0560, 0.0601, 0.0613, 0.0709], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:14:28,065 INFO [train.py:904] (0/8) Epoch 18, batch 4750, loss[loss=0.1556, simple_loss=0.2478, pruned_loss=0.03168, over 16757.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2801, pruned_loss=0.05356, over 3183956.83 frames. ], batch size: 89, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:14:56,076 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177321.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:15:07,320 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177329.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:15:27,311 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2564, 3.2258, 3.5012, 1.7186, 3.7500, 3.7029, 2.8982, 2.7106], device='cuda:0'), covar=tensor([0.0852, 0.0244, 0.0161, 0.1245, 0.0059, 0.0137, 0.0384, 0.0482], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0107, 0.0095, 0.0139, 0.0076, 0.0123, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 17:15:34,637 INFO [optim.py:368] (0/8) 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 17:15:40,567 INFO [train.py:904] (0/8) Epoch 18, batch 4800, loss[loss=0.1989, simple_loss=0.2906, pruned_loss=0.05358, over 16208.00 frames. ], tot_loss[loss=0.1901, simple_loss=0.2767, pruned_loss=0.0517, over 3185021.08 frames. ], batch size: 165, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:15:46,548 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 17:16:06,837 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177369.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:16:57,032 INFO [train.py:904] (0/8) Epoch 18, batch 4850, loss[loss=0.179, simple_loss=0.2759, pruned_loss=0.04104, over 16400.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2771, pruned_loss=0.05055, over 3184540.90 frames. ], batch size: 146, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:17:19,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0640, 4.1515, 4.4297, 4.3986, 4.4105, 4.0994, 4.1128, 4.0162], device='cuda:0'), covar=tensor([0.0285, 0.0538, 0.0361, 0.0372, 0.0383, 0.0395, 0.0820, 0.0484], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0412, 0.0406, 0.0380, 0.0451, 0.0425, 0.0520, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 17:17:28,974 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7384, 3.6759, 2.2270, 4.3930, 2.8184, 4.2645, 2.3348, 3.0569], device='cuda:0'), covar=tensor([0.0253, 0.0351, 0.1605, 0.0096, 0.0884, 0.0409, 0.1524, 0.0708], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0188, 0.0150, 0.0171, 0.0210, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 17:17:36,066 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7975, 4.7714, 4.6581, 3.9624, 4.7316, 1.7359, 4.4142, 4.4653], device='cuda:0'), covar=tensor([0.0091, 0.0086, 0.0156, 0.0451, 0.0095, 0.2664, 0.0134, 0.0207], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0144, 0.0188, 0.0173, 0.0163, 0.0199, 0.0179, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:18:05,881 INFO [optim.py:368] (0/8) 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 17:18:11,216 INFO [train.py:904] (0/8) Epoch 18, batch 4900, loss[loss=0.1843, simple_loss=0.2689, pruned_loss=0.04984, over 12278.00 frames. ], tot_loss[loss=0.1871, simple_loss=0.2759, pruned_loss=0.04918, over 3179455.55 frames. ], batch size: 248, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:18:29,114 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5300, 3.5211, 2.1131, 4.0340, 2.6853, 3.9894, 2.2504, 2.8594], device='cuda:0'), covar=tensor([0.0278, 0.0358, 0.1717, 0.0122, 0.0905, 0.0502, 0.1583, 0.0777], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0170, 0.0188, 0.0149, 0.0170, 0.0209, 0.0196, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 17:18:30,789 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177465.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:18:33,003 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177467.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:18:44,984 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-04-30 17:19:07,810 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-04-30 17:19:25,483 INFO [train.py:904] (0/8) Epoch 18, batch 4950, loss[loss=0.1788, simple_loss=0.2717, pruned_loss=0.04298, over 17239.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2753, pruned_loss=0.04853, over 3177518.91 frames. ], batch size: 45, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:19:34,753 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177509.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 17:20:30,525 INFO [optim.py:368] (0/8) 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 17:20:33,534 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0820, 3.8461, 3.7951, 2.3959, 3.4306, 3.7544, 3.4708, 2.0476], device='cuda:0'), covar=tensor([0.0546, 0.0033, 0.0039, 0.0394, 0.0093, 0.0079, 0.0079, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0078, 0.0078, 0.0130, 0.0093, 0.0103, 0.0090, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:20:35,828 INFO [train.py:904] (0/8) Epoch 18, batch 5000, loss[loss=0.2005, simple_loss=0.2894, pruned_loss=0.05579, over 17092.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2771, pruned_loss=0.04874, over 3198040.55 frames. ], batch size: 49, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:20:37,348 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177553.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:20:41,863 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 18, batch 5050, loss[loss=0.1675, simple_loss=0.2639, pruned_loss=0.03551, over 16902.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2779, pruned_loss=0.04883, over 3192878.18 frames. ], batch size: 116, lr: 3.77e-03, grad_scale: 16.0 2023-04-30 17:21:48,860 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5061, 1.7728, 2.1969, 2.5337, 2.5580, 2.8538, 1.8351, 2.8127], device='cuda:0'), covar=tensor([0.0215, 0.0511, 0.0337, 0.0327, 0.0293, 0.0175, 0.0566, 0.0134], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0190, 0.0178, 0.0181, 0.0189, 0.0147, 0.0194, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:22:04,053 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:22:18,264 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 17:22:54,212 INFO [optim.py:368] (0/8) 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 17:22:58,431 INFO [train.py:904] (0/8) Epoch 18, batch 5100, loss[loss=0.1596, simple_loss=0.2519, pruned_loss=0.0336, over 16756.00 frames. ], tot_loss[loss=0.186, simple_loss=0.276, pruned_loss=0.04803, over 3190631.61 frames. ], batch size: 83, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:23:16,168 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177664.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:23:20,139 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0078, 3.9464, 3.9799, 3.2074, 3.9225, 1.7085, 3.6719, 3.4926], device='cuda:0'), covar=tensor([0.0118, 0.0125, 0.0144, 0.0359, 0.0097, 0.2789, 0.0145, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0145, 0.0190, 0.0175, 0.0165, 0.0200, 0.0181, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:24:10,484 INFO [train.py:904] (0/8) Epoch 18, batch 5150, loss[loss=0.1784, simple_loss=0.2778, pruned_loss=0.0395, over 16165.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2757, pruned_loss=0.04698, over 3202593.57 frames. ], batch size: 165, lr: 3.77e-03, grad_scale: 8.0 2023-04-30 17:24:46,068 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177725.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:20,512 INFO [optim.py:368] (0/8) 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 17:25:24,675 INFO [train.py:904] (0/8) Epoch 18, batch 5200, loss[loss=0.1912, simple_loss=0.2766, pruned_loss=0.0529, over 15280.00 frames. ], tot_loss[loss=0.1839, simple_loss=0.2745, pruned_loss=0.04665, over 3196500.11 frames. ], batch size: 190, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:25:44,147 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:25:47,177 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=177767.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:06,695 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177780.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:31,550 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=177797.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:39,167 INFO [train.py:904] (0/8) Epoch 18, batch 5250, loss[loss=0.1716, simple_loss=0.2591, pruned_loss=0.04205, over 15270.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2722, pruned_loss=0.04643, over 3198041.19 frames. ], batch size: 190, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:26:55,360 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177813.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:26:58,851 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=177815.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:05,422 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 17:27:06,288 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8439, 2.7839, 2.8033, 4.7230, 3.6519, 4.1624, 1.8338, 3.1941], device='cuda:0'), covar=tensor([0.1248, 0.0761, 0.1124, 0.0121, 0.0244, 0.0406, 0.1478, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0169, 0.0191, 0.0180, 0.0204, 0.0214, 0.0196, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 17:27:37,238 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:27:46,432 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3471, 5.6723, 5.3797, 5.4673, 5.1795, 5.0943, 5.0044, 5.7536], device='cuda:0'), covar=tensor([0.1175, 0.0815, 0.0939, 0.0656, 0.0739, 0.0684, 0.1082, 0.0807], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0772, 0.0633, 0.0576, 0.0487, 0.0495, 0.0644, 0.0601], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:27:48,910 INFO [optim.py:368] (0/8) 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 17:27:52,860 INFO [train.py:904] (0/8) Epoch 18, batch 5300, loss[loss=0.1662, simple_loss=0.2526, pruned_loss=0.03989, over 16708.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2684, pruned_loss=0.04533, over 3215109.75 frames. ], batch size: 134, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:28:02,069 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=177858.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:28:04,135 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6723, 3.7383, 2.8420, 2.2941, 2.4356, 2.4615, 4.0044, 3.3651], device='cuda:0'), covar=tensor([0.2699, 0.0614, 0.1777, 0.2514, 0.2525, 0.1819, 0.0402, 0.1158], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0264, 0.0298, 0.0302, 0.0291, 0.0246, 0.0288, 0.0326], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:28:14,288 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4113, 4.1677, 4.0908, 2.6275, 3.6741, 4.0793, 3.6035, 2.0628], device='cuda:0'), covar=tensor([0.0487, 0.0029, 0.0034, 0.0362, 0.0077, 0.0077, 0.0088, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0078, 0.0079, 0.0131, 0.0094, 0.0103, 0.0091, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:28:28,967 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9041, 1.9511, 2.4454, 2.8538, 2.7110, 3.2670, 2.0676, 3.1904], device='cuda:0'), covar=tensor([0.0193, 0.0471, 0.0318, 0.0284, 0.0310, 0.0149, 0.0495, 0.0130], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0188, 0.0176, 0.0179, 0.0187, 0.0145, 0.0192, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:28:43,040 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3783, 3.2816, 3.6053, 1.6756, 3.8372, 3.8151, 2.9371, 2.7121], device='cuda:0'), covar=tensor([0.0773, 0.0245, 0.0158, 0.1281, 0.0053, 0.0127, 0.0398, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0138, 0.0076, 0.0121, 0.0125, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 17:29:05,782 INFO [train.py:904] (0/8) Epoch 18, batch 5350, loss[loss=0.1748, simple_loss=0.2745, pruned_loss=0.03761, over 16901.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2668, pruned_loss=0.0445, over 3227999.10 frames. ], batch size: 96, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:29:06,191 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8628, 4.9388, 5.2739, 5.2538, 5.2681, 4.9285, 4.8556, 4.6206], device='cuda:0'), covar=tensor([0.0316, 0.0505, 0.0331, 0.0344, 0.0459, 0.0352, 0.0925, 0.0467], device='cuda:0'), in_proj_covar=tensor([0.0382, 0.0417, 0.0410, 0.0384, 0.0456, 0.0430, 0.0527, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 17:29:17,496 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=177909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:29:49,854 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 17:29:50,731 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2224, 2.4826, 2.0220, 2.2094, 2.8543, 2.4928, 2.8592, 2.9818], device='cuda:0'), covar=tensor([0.0105, 0.0378, 0.0513, 0.0424, 0.0249, 0.0345, 0.0208, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0226, 0.0218, 0.0216, 0.0227, 0.0226, 0.0226, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:30:08,780 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-30 17:30:10,644 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-04-30 17:30:15,240 INFO [optim.py:368] (0/8) 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 17:30:19,783 INFO [train.py:904] (0/8) Epoch 18, batch 5400, loss[loss=0.2017, simple_loss=0.296, pruned_loss=0.05375, over 16837.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2693, pruned_loss=0.04506, over 3217012.33 frames. ], batch size: 116, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:31:28,991 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 17:31:32,305 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-178000.pt 2023-04-30 17:31:39,773 INFO [train.py:904] (0/8) Epoch 18, batch 5450, loss[loss=0.1918, simple_loss=0.271, pruned_loss=0.05631, over 11693.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2718, pruned_loss=0.0462, over 3209435.68 frames. ], batch size: 247, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:32:08,127 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178020.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:32:41,217 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2693, 5.3202, 5.7303, 5.7324, 5.7387, 5.3262, 5.2376, 4.9562], device='cuda:0'), covar=tensor([0.0245, 0.0409, 0.0247, 0.0273, 0.0373, 0.0295, 0.0884, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0419, 0.0411, 0.0386, 0.0458, 0.0433, 0.0531, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 17:32:52,432 INFO [optim.py:368] (0/8) 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 17:32:56,920 INFO [train.py:904] (0/8) Epoch 18, batch 5500, loss[loss=0.2334, simple_loss=0.3108, pruned_loss=0.07804, over 16587.00 frames. ], tot_loss[loss=0.1892, simple_loss=0.2783, pruned_loss=0.05008, over 3193217.99 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:33:58,219 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 17:34:14,394 INFO [train.py:904] (0/8) Epoch 18, batch 5550, loss[loss=0.2866, simple_loss=0.3466, pruned_loss=0.1133, over 11485.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2853, pruned_loss=0.05529, over 3158187.15 frames. ], batch size: 248, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:09,019 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178136.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:35:31,155 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.108e+02 3.250e+02 3.857e+02 5.113e+02 1.112e+03, threshold=7.715e+02, percent-clipped=10.0 2023-04-30 17:35:34,172 INFO [train.py:904] (0/8) Epoch 18, batch 5600, loss[loss=0.2838, simple_loss=0.3277, pruned_loss=0.12, over 10893.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2903, pruned_loss=0.05941, over 3122998.16 frames. ], batch size: 248, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:35:36,609 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178153.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:36:08,139 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3837, 4.2644, 4.4499, 4.5766, 4.7057, 4.3049, 4.6675, 4.7038], device='cuda:0'), covar=tensor([0.1727, 0.1198, 0.1345, 0.0652, 0.0566, 0.1200, 0.0730, 0.0633], device='cuda:0'), in_proj_covar=tensor([0.0597, 0.0739, 0.0869, 0.0759, 0.0560, 0.0598, 0.0607, 0.0703], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:36:58,602 INFO [train.py:904] (0/8) Epoch 18, batch 5650, loss[loss=0.2697, simple_loss=0.3325, pruned_loss=0.1034, over 11267.00 frames. ], tot_loss[loss=0.2118, simple_loss=0.2957, pruned_loss=0.0639, over 3086866.36 frames. ], batch size: 248, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:37:09,715 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178209.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:38:05,082 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7410, 2.2486, 1.8843, 2.0417, 2.6225, 2.2511, 2.5642, 2.7759], device='cuda:0'), covar=tensor([0.0166, 0.0396, 0.0492, 0.0451, 0.0240, 0.0364, 0.0214, 0.0244], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0227, 0.0220, 0.0218, 0.0228, 0.0227, 0.0228, 0.0221], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:38:16,449 INFO [optim.py:368] (0/8) 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 17:38:17,801 INFO [train.py:904] (0/8) Epoch 18, batch 5700, loss[loss=0.231, simple_loss=0.3227, pruned_loss=0.06967, over 16867.00 frames. ], tot_loss[loss=0.2142, simple_loss=0.2976, pruned_loss=0.06545, over 3085698.93 frames. ], batch size: 116, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:38:25,292 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178257.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:39:39,092 INFO [train.py:904] (0/8) Epoch 18, batch 5750, loss[loss=0.2, simple_loss=0.2851, pruned_loss=0.05747, over 16449.00 frames. ], tot_loss[loss=0.2171, simple_loss=0.3004, pruned_loss=0.06692, over 3077823.34 frames. ], batch size: 68, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:40:08,185 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:40:59,322 INFO [optim.py:368] (0/8) 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 17:41:00,712 INFO [train.py:904] (0/8) Epoch 18, batch 5800, loss[loss=0.2074, simple_loss=0.298, pruned_loss=0.05847, over 16673.00 frames. ], tot_loss[loss=0.2167, simple_loss=0.3006, pruned_loss=0.06641, over 3074459.76 frames. ], batch size: 134, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:41:27,558 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178368.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:42:19,753 INFO [train.py:904] (0/8) Epoch 18, batch 5850, loss[loss=0.209, simple_loss=0.2967, pruned_loss=0.06066, over 16656.00 frames. ], tot_loss[loss=0.2136, simple_loss=0.2982, pruned_loss=0.06447, over 3087569.06 frames. ], batch size: 76, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:16,127 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:39,874 INFO [optim.py:368] (0/8) 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 17:43:41,475 INFO [train.py:904] (0/8) Epoch 18, batch 5900, loss[loss=0.2013, simple_loss=0.2833, pruned_loss=0.05968, over 16356.00 frames. ], tot_loss[loss=0.2129, simple_loss=0.2974, pruned_loss=0.06422, over 3100162.89 frames. ], batch size: 146, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:43:42,266 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:43:44,135 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=178453.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:32,741 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178484.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:51,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5412, 3.9973, 4.0964, 2.7511, 3.7393, 4.1272, 3.7890, 2.1657], device='cuda:0'), covar=tensor([0.0490, 0.0058, 0.0048, 0.0371, 0.0084, 0.0099, 0.0075, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0079, 0.0080, 0.0132, 0.0094, 0.0105, 0.0091, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:44:58,647 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=178501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:44:59,559 INFO [train.py:904] (0/8) Epoch 18, batch 5950, loss[loss=0.1905, simple_loss=0.2827, pruned_loss=0.04916, over 16529.00 frames. ], tot_loss[loss=0.2114, simple_loss=0.2973, pruned_loss=0.06277, over 3081697.36 frames. ], batch size: 75, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:45:18,618 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178513.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:45:21,292 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8431, 3.8568, 2.4674, 4.6764, 3.0293, 4.5825, 2.6051, 3.0883], device='cuda:0'), covar=tensor([0.0257, 0.0376, 0.1562, 0.0211, 0.0790, 0.0628, 0.1423, 0.0744], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0171, 0.0190, 0.0151, 0.0173, 0.0211, 0.0198, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 17:45:29,041 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7293, 4.9728, 5.1155, 4.9818, 4.9730, 5.5293, 5.0477, 4.8534], device='cuda:0'), covar=tensor([0.1083, 0.1799, 0.2261, 0.1929, 0.2574, 0.0904, 0.1586, 0.2486], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0561, 0.0616, 0.0471, 0.0631, 0.0642, 0.0488, 0.0636], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 17:46:17,850 INFO [optim.py:368] (0/8) 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 17:46:19,088 INFO [train.py:904] (0/8) Epoch 18, batch 6000, loss[loss=0.1858, simple_loss=0.2788, pruned_loss=0.04636, over 16848.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.297, pruned_loss=0.06232, over 3098511.66 frames. ], batch size: 96, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:46:19,089 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 17:46:29,950 INFO [train.py:938] (0/8) Epoch 18, validation: loss=0.1531, simple_loss=0.2661, pruned_loss=0.02007, over 944034.00 frames. 2023-04-30 17:46:29,951 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 17:47:39,581 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-04-30 17:47:48,048 INFO [train.py:904] (0/8) Epoch 18, batch 6050, loss[loss=0.2012, simple_loss=0.2895, pruned_loss=0.05642, over 16376.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2952, pruned_loss=0.06106, over 3131365.01 frames. ], batch size: 146, lr: 3.76e-03, grad_scale: 8.0 2023-04-30 17:48:14,312 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178619.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:48:21,642 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3975, 2.1943, 2.9080, 3.2196, 3.0654, 3.7432, 2.4204, 3.7516], device='cuda:0'), covar=tensor([0.0171, 0.0471, 0.0316, 0.0267, 0.0268, 0.0124, 0.0496, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0187, 0.0174, 0.0178, 0.0186, 0.0145, 0.0191, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:49:00,526 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-04-30 17:49:06,484 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.761e+02 2.787e+02 3.191e+02 3.956e+02 7.476e+02, threshold=6.381e+02, percent-clipped=1.0 2023-04-30 17:49:06,500 INFO [train.py:904] (0/8) Epoch 18, batch 6100, loss[loss=0.1829, simple_loss=0.2791, pruned_loss=0.04338, over 16857.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2943, pruned_loss=0.05992, over 3137299.25 frames. ], batch size: 96, lr: 3.76e-03, grad_scale: 4.0 2023-04-30 17:49:50,799 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178680.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:12,347 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178694.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:50:16,881 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178697.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:50:23,994 INFO [train.py:904] (0/8) Epoch 18, batch 6150, loss[loss=0.2252, simple_loss=0.2952, pruned_loss=0.07758, over 11417.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2922, pruned_loss=0.05916, over 3132955.91 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:39,662 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.052e+02 2.816e+02 3.285e+02 4.160e+02 8.412e+02, threshold=6.570e+02, percent-clipped=4.0 2023-04-30 17:51:39,682 INFO [train.py:904] (0/8) Epoch 18, batch 6200, loss[loss=0.1963, simple_loss=0.2876, pruned_loss=0.05256, over 16776.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2907, pruned_loss=0.05873, over 3137734.09 frames. ], batch size: 124, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:51:45,194 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178755.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:51:49,748 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178758.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 17:52:09,017 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:52:14,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3719, 2.1217, 1.7321, 1.8605, 2.4446, 2.0801, 2.1785, 2.5300], device='cuda:0'), covar=tensor([0.0167, 0.0370, 0.0495, 0.0456, 0.0229, 0.0344, 0.0200, 0.0254], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0224, 0.0218, 0.0217, 0.0226, 0.0224, 0.0227, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:52:14,302 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5714, 3.0125, 3.1492, 1.9338, 2.7465, 2.1054, 3.1455, 3.2457], device='cuda:0'), covar=tensor([0.0264, 0.0742, 0.0575, 0.1955, 0.0866, 0.0995, 0.0661, 0.0844], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0161, 0.0166, 0.0150, 0.0143, 0.0128, 0.0143, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 17:52:52,920 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0830, 5.0622, 4.9357, 4.1996, 4.9991, 1.8234, 4.7428, 4.6748], device='cuda:0'), covar=tensor([0.0117, 0.0110, 0.0184, 0.0413, 0.0109, 0.2723, 0.0195, 0.0210], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0142, 0.0187, 0.0172, 0.0162, 0.0198, 0.0177, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:52:56,578 INFO [train.py:904] (0/8) Epoch 18, batch 6250, loss[loss=0.1862, simple_loss=0.2836, pruned_loss=0.04444, over 16790.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2901, pruned_loss=0.0583, over 3139029.81 frames. ], batch size: 83, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:53:07,456 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178808.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:53:41,407 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:54:15,697 INFO [train.py:904] (0/8) Epoch 18, batch 6300, loss[loss=0.2317, simple_loss=0.3247, pruned_loss=0.06929, over 16876.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2903, pruned_loss=0.05833, over 3133700.66 frames. ], batch size: 116, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:54:17,524 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.018e+02 2.791e+02 3.273e+02 3.863e+02 9.337e+02, threshold=6.545e+02, percent-clipped=2.0 2023-04-30 17:55:22,168 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=178894.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 17:55:34,231 INFO [train.py:904] (0/8) Epoch 18, batch 6350, loss[loss=0.2001, simple_loss=0.2891, pruned_loss=0.05558, over 16270.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2906, pruned_loss=0.05949, over 3126196.77 frames. ], batch size: 165, lr: 3.75e-03, grad_scale: 2.0 2023-04-30 17:56:52,040 INFO [train.py:904] (0/8) Epoch 18, batch 6400, loss[loss=0.1886, simple_loss=0.2835, pruned_loss=0.04684, over 16685.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2913, pruned_loss=0.06129, over 3097711.38 frames. ], batch size: 76, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:56:53,840 INFO [optim.py:368] (0/8) 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 17:56:57,533 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=178955.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 17:57:27,448 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=178975.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 17:58:07,514 INFO [train.py:904] (0/8) Epoch 18, batch 6450, loss[loss=0.1962, simple_loss=0.2835, pruned_loss=0.05441, over 16757.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2917, pruned_loss=0.06137, over 3078362.34 frames. ], batch size: 124, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:58:23,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8517, 5.1504, 4.8763, 4.8839, 4.6397, 4.5959, 4.5295, 5.2325], device='cuda:0'), covar=tensor([0.1222, 0.0886, 0.1013, 0.0940, 0.0856, 0.1016, 0.1280, 0.0957], device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0770, 0.0631, 0.0575, 0.0481, 0.0493, 0.0639, 0.0598], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 17:59:24,079 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179050.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 17:59:26,168 INFO [train.py:904] (0/8) Epoch 18, batch 6500, loss[loss=0.2078, simple_loss=0.2847, pruned_loss=0.06548, over 16642.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.29, pruned_loss=0.06073, over 3084790.04 frames. ], batch size: 57, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 17:59:27,322 INFO [optim.py:368] (0/8) 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 17:59:27,823 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179053.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:00:28,356 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6492, 3.9610, 3.1155, 2.2357, 2.6217, 2.3594, 4.0390, 3.4324], device='cuda:0'), covar=tensor([0.2926, 0.0623, 0.1585, 0.2998, 0.2674, 0.2151, 0.0554, 0.1386], device='cuda:0'), in_proj_covar=tensor([0.0318, 0.0261, 0.0296, 0.0301, 0.0288, 0.0246, 0.0285, 0.0324], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:00:39,666 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7066, 2.6973, 2.5769, 4.1764, 2.8850, 4.0593, 1.4651, 2.9773], device='cuda:0'), covar=tensor([0.1414, 0.0769, 0.1184, 0.0164, 0.0298, 0.0425, 0.1794, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0168, 0.0190, 0.0178, 0.0204, 0.0213, 0.0195, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:00:44,336 INFO [train.py:904] (0/8) Epoch 18, batch 6550, loss[loss=0.228, simple_loss=0.3021, pruned_loss=0.07698, over 11143.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2931, pruned_loss=0.06143, over 3081672.79 frames. ], batch size: 246, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:00:53,823 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179108.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:20,188 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179126.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:01:58,905 INFO [train.py:904] (0/8) Epoch 18, batch 6600, loss[loss=0.2258, simple_loss=0.3257, pruned_loss=0.06302, over 16795.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2952, pruned_loss=0.06143, over 3090754.03 frames. ], batch size: 83, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:02:00,671 INFO [optim.py:368] (0/8) 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 18:02:05,104 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179156.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:02:56,065 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 18:03:17,510 INFO [train.py:904] (0/8) Epoch 18, batch 6650, loss[loss=0.2384, simple_loss=0.3055, pruned_loss=0.08566, over 11593.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2949, pruned_loss=0.06185, over 3092979.82 frames. ], batch size: 248, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:30,599 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179250.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 18:04:32,473 INFO [train.py:904] (0/8) Epoch 18, batch 6700, loss[loss=0.2201, simple_loss=0.3026, pruned_loss=0.06878, over 16925.00 frames. ], tot_loss[loss=0.21, simple_loss=0.2948, pruned_loss=0.06257, over 3094854.25 frames. ], batch size: 109, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:04:34,184 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.091e+02 2.690e+02 3.432e+02 4.163e+02 9.246e+02, threshold=6.864e+02, percent-clipped=3.0 2023-04-30 18:05:08,737 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179275.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:05:48,870 INFO [train.py:904] (0/8) Epoch 18, batch 6750, loss[loss=0.1881, simple_loss=0.2753, pruned_loss=0.05044, over 16682.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2926, pruned_loss=0.06182, over 3111275.42 frames. ], batch size: 76, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:06:14,524 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1903, 4.2920, 4.4487, 4.1632, 4.3089, 4.8130, 4.3361, 4.0438], device='cuda:0'), covar=tensor([0.1807, 0.1962, 0.2123, 0.2267, 0.2545, 0.1111, 0.1691, 0.2575], device='cuda:0'), in_proj_covar=tensor([0.0394, 0.0567, 0.0626, 0.0477, 0.0635, 0.0652, 0.0491, 0.0641], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:06:19,479 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179323.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:06:25,025 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 18:06:45,742 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7971, 5.1256, 5.2975, 5.1115, 5.1567, 5.6732, 5.1319, 4.9329], device='cuda:0'), covar=tensor([0.1019, 0.1629, 0.1800, 0.1690, 0.2165, 0.0845, 0.1373, 0.2064], device='cuda:0'), in_proj_covar=tensor([0.0395, 0.0568, 0.0627, 0.0478, 0.0636, 0.0653, 0.0492, 0.0641], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:07:00,962 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179350.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:03,097 INFO [train.py:904] (0/8) Epoch 18, batch 6800, loss[loss=0.1858, simple_loss=0.2744, pruned_loss=0.04854, over 17142.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.293, pruned_loss=0.06181, over 3121585.95 frames. ], batch size: 47, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:07:04,930 INFO [optim.py:368] (0/8) 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 18:07:05,345 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179353.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:07:18,364 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179361.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:07:48,624 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7979, 2.3396, 1.9373, 2.2112, 2.7091, 2.3529, 2.6049, 2.8806], device='cuda:0'), covar=tensor([0.0175, 0.0433, 0.0509, 0.0416, 0.0245, 0.0367, 0.0212, 0.0246], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0225, 0.0221, 0.0220, 0.0228, 0.0226, 0.0228, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:08:15,643 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179398.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:20,459 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179401.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:08:21,228 INFO [train.py:904] (0/8) Epoch 18, batch 6850, loss[loss=0.2372, simple_loss=0.3328, pruned_loss=0.07082, over 17286.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2942, pruned_loss=0.06249, over 3122788.41 frames. ], batch size: 52, lr: 3.75e-03, grad_scale: 8.0 2023-04-30 18:08:50,212 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179422.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:08:56,266 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:09:02,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0666, 5.3923, 5.1026, 5.1277, 4.8846, 4.7654, 4.6932, 5.4459], device='cuda:0'), covar=tensor([0.1226, 0.0798, 0.0969, 0.0762, 0.0835, 0.0926, 0.1086, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0623, 0.0765, 0.0625, 0.0571, 0.0478, 0.0491, 0.0634, 0.0591], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:09:23,559 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179444.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:09:34,745 INFO [train.py:904] (0/8) Epoch 18, batch 6900, loss[loss=0.2808, simple_loss=0.3382, pruned_loss=0.1117, over 11384.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2964, pruned_loss=0.06225, over 3118781.46 frames. ], batch size: 247, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:09:38,469 INFO [optim.py:368] (0/8) 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 18:10:10,147 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179474.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:10:48,210 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2762, 3.4269, 3.5725, 3.5456, 3.5505, 3.3604, 3.4008, 3.4496], device='cuda:0'), covar=tensor([0.0416, 0.0673, 0.0479, 0.0452, 0.0541, 0.0572, 0.0837, 0.0545], device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0429, 0.0417, 0.0394, 0.0468, 0.0440, 0.0539, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 18:10:53,470 INFO [train.py:904] (0/8) Epoch 18, batch 6950, loss[loss=0.1896, simple_loss=0.2785, pruned_loss=0.05035, over 17197.00 frames. ], tot_loss[loss=0.2128, simple_loss=0.2979, pruned_loss=0.06384, over 3102866.22 frames. ], batch size: 44, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:10:58,246 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179505.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:11:34,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6607, 2.4303, 2.1513, 3.1921, 2.0595, 3.5462, 1.4769, 2.5993], device='cuda:0'), covar=tensor([0.1462, 0.0810, 0.1423, 0.0206, 0.0154, 0.0415, 0.1854, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0167, 0.0189, 0.0176, 0.0203, 0.0211, 0.0194, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:12:07,081 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=179550.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:12:09,856 INFO [train.py:904] (0/8) Epoch 18, batch 7000, loss[loss=0.2037, simple_loss=0.3091, pruned_loss=0.04918, over 16789.00 frames. ], tot_loss[loss=0.2125, simple_loss=0.2984, pruned_loss=0.06332, over 3097181.78 frames. ], batch size: 83, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:12:12,181 INFO [optim.py:368] (0/8) 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 18:12:55,832 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-04-30 18:13:16,942 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=179598.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:13:21,689 INFO [train.py:904] (0/8) Epoch 18, batch 7050, loss[loss=0.2028, simple_loss=0.2882, pruned_loss=0.05865, over 16686.00 frames. ], tot_loss[loss=0.2134, simple_loss=0.2996, pruned_loss=0.06363, over 3073910.57 frames. ], batch size: 57, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:13:35,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3996, 3.2424, 2.6950, 2.1396, 2.2509, 2.2333, 3.3349, 3.0318], device='cuda:0'), covar=tensor([0.2878, 0.0779, 0.1673, 0.2577, 0.2477, 0.2116, 0.0505, 0.1260], device='cuda:0'), in_proj_covar=tensor([0.0322, 0.0265, 0.0300, 0.0306, 0.0292, 0.0249, 0.0287, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:14:12,913 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179636.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:14:37,719 INFO [train.py:904] (0/8) Epoch 18, batch 7100, loss[loss=0.2089, simple_loss=0.2903, pruned_loss=0.06377, over 16566.00 frames. ], tot_loss[loss=0.2122, simple_loss=0.2979, pruned_loss=0.0633, over 3070952.61 frames. ], batch size: 35, lr: 3.75e-03, grad_scale: 4.0 2023-04-30 18:14:40,317 INFO [optim.py:368] (0/8) 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 18:15:33,590 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8027, 1.2874, 1.6856, 1.6541, 1.7705, 1.9561, 1.5729, 1.7585], device='cuda:0'), covar=tensor([0.0263, 0.0380, 0.0222, 0.0279, 0.0248, 0.0163, 0.0404, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0189, 0.0175, 0.0178, 0.0187, 0.0145, 0.0191, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:15:46,313 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179697.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:15:53,688 INFO [train.py:904] (0/8) Epoch 18, batch 7150, loss[loss=0.1993, simple_loss=0.2856, pruned_loss=0.05645, over 16430.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.296, pruned_loss=0.06313, over 3065705.10 frames. ], batch size: 146, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:15:55,575 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-30 18:16:08,467 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4867, 3.5782, 2.6676, 2.1928, 2.4045, 2.2647, 3.8527, 3.2431], device='cuda:0'), covar=tensor([0.3111, 0.0693, 0.1863, 0.2519, 0.2759, 0.2080, 0.0485, 0.1197], device='cuda:0'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0292, 0.0249, 0.0287, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:16:16,383 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179717.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:16:26,714 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-04-30 18:16:59,856 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3304, 2.9398, 2.6531, 2.2207, 2.1957, 2.2018, 2.8996, 2.8466], device='cuda:0'), covar=tensor([0.2536, 0.0817, 0.1626, 0.2543, 0.2445, 0.2158, 0.0517, 0.1245], device='cuda:0'), in_proj_covar=tensor([0.0322, 0.0265, 0.0299, 0.0305, 0.0291, 0.0249, 0.0286, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:17:08,206 INFO [train.py:904] (0/8) Epoch 18, batch 7200, loss[loss=0.1798, simple_loss=0.2715, pruned_loss=0.04401, over 16545.00 frames. ], tot_loss[loss=0.2086, simple_loss=0.2938, pruned_loss=0.06167, over 3048420.96 frames. ], batch size: 75, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:17:10,638 INFO [optim.py:368] (0/8) 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 18:17:29,404 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 18:17:30,277 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6701, 2.2254, 1.8821, 1.9747, 2.5753, 2.2088, 2.5131, 2.7378], device='cuda:0'), covar=tensor([0.0189, 0.0441, 0.0552, 0.0502, 0.0290, 0.0403, 0.0236, 0.0269], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:17:54,708 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-04-30 18:18:17,316 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179796.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:18:24,747 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179800.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 18:18:27,058 INFO [train.py:904] (0/8) Epoch 18, batch 7250, loss[loss=0.1866, simple_loss=0.2675, pruned_loss=0.05285, over 16878.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2908, pruned_loss=0.05997, over 3070408.37 frames. ], batch size: 116, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:18:29,316 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7168, 4.7734, 5.1508, 5.0943, 5.1115, 4.8096, 4.7917, 4.5226], device='cuda:0'), covar=tensor([0.0268, 0.0472, 0.0334, 0.0403, 0.0417, 0.0335, 0.0807, 0.0465], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0426, 0.0416, 0.0393, 0.0467, 0.0438, 0.0534, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 18:19:41,459 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1644, 1.4511, 1.9561, 2.0583, 2.1453, 2.3951, 1.6892, 2.2826], device='cuda:0'), covar=tensor([0.0210, 0.0454, 0.0243, 0.0315, 0.0277, 0.0174, 0.0457, 0.0122], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0186, 0.0173, 0.0176, 0.0185, 0.0143, 0.0189, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:19:42,149 INFO [train.py:904] (0/8) Epoch 18, batch 7300, loss[loss=0.2102, simple_loss=0.2951, pruned_loss=0.06264, over 15330.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2904, pruned_loss=0.05967, over 3083189.90 frames. ], batch size: 190, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:19:45,258 INFO [optim.py:368] (0/8) 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 18:19:50,398 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179857.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:20:33,699 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3707, 4.5261, 4.6735, 4.4737, 4.5544, 5.0254, 4.5756, 4.3082], device='cuda:0'), covar=tensor([0.1456, 0.1769, 0.2036, 0.1986, 0.2171, 0.1006, 0.1655, 0.2459], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0559, 0.0618, 0.0470, 0.0628, 0.0644, 0.0486, 0.0631], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:20:58,400 INFO [train.py:904] (0/8) Epoch 18, batch 7350, loss[loss=0.2011, simple_loss=0.2866, pruned_loss=0.0578, over 15428.00 frames. ], tot_loss[loss=0.207, simple_loss=0.292, pruned_loss=0.06101, over 3053298.69 frames. ], batch size: 191, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:21:04,882 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179906.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:21:32,202 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=179924.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:22:16,607 INFO [train.py:904] (0/8) Epoch 18, batch 7400, loss[loss=0.2423, simple_loss=0.3209, pruned_loss=0.08187, over 15375.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2933, pruned_loss=0.06155, over 3060904.20 frames. ], batch size: 190, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:22:19,966 INFO [optim.py:368] (0/8) 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 18:22:30,146 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9362, 2.3926, 1.9301, 2.1075, 2.7544, 2.3503, 2.7031, 2.9036], device='cuda:0'), covar=tensor([0.0148, 0.0381, 0.0546, 0.0521, 0.0231, 0.0425, 0.0213, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0222, 0.0217, 0.0217, 0.0225, 0.0222, 0.0225, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:22:41,122 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179967.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:23:09,669 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=179985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:21,177 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=179992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:23:33,676 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-180000.pt 2023-04-30 18:23:39,779 INFO [train.py:904] (0/8) Epoch 18, batch 7450, loss[loss=0.1821, simple_loss=0.2832, pruned_loss=0.04053, over 16842.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2938, pruned_loss=0.06216, over 3058268.48 frames. ], batch size: 96, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:23:51,154 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3008, 2.1795, 2.7534, 3.2895, 3.1084, 3.7160, 2.3611, 3.6946], device='cuda:0'), covar=tensor([0.0162, 0.0457, 0.0281, 0.0213, 0.0224, 0.0115, 0.0463, 0.0104], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0186, 0.0172, 0.0175, 0.0184, 0.0143, 0.0188, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:24:02,602 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8919, 2.5919, 2.5912, 1.8857, 2.5867, 2.6694, 2.5186, 1.8322], device='cuda:0'), covar=tensor([0.0427, 0.0106, 0.0077, 0.0372, 0.0126, 0.0130, 0.0111, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0079, 0.0079, 0.0133, 0.0093, 0.0105, 0.0091, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:24:05,911 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180017.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:24:10,816 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9375, 2.4953, 2.6050, 1.9477, 2.5667, 2.6330, 2.5497, 1.8158], device='cuda:0'), covar=tensor([0.0410, 0.0135, 0.0080, 0.0326, 0.0121, 0.0130, 0.0112, 0.0434], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0079, 0.0079, 0.0133, 0.0093, 0.0105, 0.0091, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:24:19,180 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180026.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:25:01,490 INFO [train.py:904] (0/8) Epoch 18, batch 7500, loss[loss=0.2112, simple_loss=0.2971, pruned_loss=0.06265, over 16226.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2945, pruned_loss=0.06171, over 3054463.81 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:25:04,523 INFO [optim.py:368] (0/8) 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 18:25:22,817 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:25:56,706 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180087.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:26:16,490 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180100.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:26:19,023 INFO [train.py:904] (0/8) Epoch 18, batch 7550, loss[loss=0.2021, simple_loss=0.2869, pruned_loss=0.05863, over 16166.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2939, pruned_loss=0.06224, over 3041664.42 frames. ], batch size: 165, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:26:21,114 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-04-30 18:27:29,786 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180148.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:27:36,554 INFO [train.py:904] (0/8) Epoch 18, batch 7600, loss[loss=0.191, simple_loss=0.2854, pruned_loss=0.04831, over 16842.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2927, pruned_loss=0.06204, over 3050498.37 frames. ], batch size: 102, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:27:36,947 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2477, 5.5531, 5.3206, 5.3178, 5.0439, 4.9393, 4.9662, 5.6589], device='cuda:0'), covar=tensor([0.1108, 0.0738, 0.0876, 0.0820, 0.0740, 0.0759, 0.1045, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0764, 0.0624, 0.0570, 0.0477, 0.0491, 0.0638, 0.0595], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:27:36,950 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180152.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:27:39,411 INFO [optim.py:368] (0/8) 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 18:28:34,518 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3527, 2.9905, 2.9502, 1.8613, 2.6789, 2.0566, 2.9167, 3.1960], device='cuda:0'), covar=tensor([0.0330, 0.0754, 0.0633, 0.2139, 0.0930, 0.1093, 0.0700, 0.0908], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0159, 0.0165, 0.0150, 0.0142, 0.0127, 0.0141, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:28:55,657 INFO [train.py:904] (0/8) Epoch 18, batch 7650, loss[loss=0.2027, simple_loss=0.2903, pruned_loss=0.05755, over 16655.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2927, pruned_loss=0.06183, over 3078841.28 frames. ], batch size: 134, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:29:00,775 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:13,296 INFO [train.py:904] (0/8) Epoch 18, batch 7700, loss[loss=0.1951, simple_loss=0.2773, pruned_loss=0.05644, over 16434.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2929, pruned_loss=0.06223, over 3082838.01 frames. ], batch size: 75, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:30:18,209 INFO [optim.py:368] (0/8) 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 18:30:20,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2314, 4.3280, 4.6497, 4.5961, 4.6184, 4.3514, 4.3187, 4.1932], device='cuda:0'), covar=tensor([0.0344, 0.0521, 0.0411, 0.0448, 0.0476, 0.0416, 0.0946, 0.0531], device='cuda:0'), in_proj_covar=tensor([0.0381, 0.0416, 0.0406, 0.0383, 0.0456, 0.0427, 0.0522, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 18:30:29,177 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180262.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:30:35,088 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180266.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:30:44,605 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2457, 3.2841, 1.9121, 3.5922, 2.4201, 3.6119, 2.0271, 2.6430], device='cuda:0'), covar=tensor([0.0296, 0.0425, 0.1777, 0.0230, 0.0941, 0.0662, 0.1658, 0.0892], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0174, 0.0192, 0.0151, 0.0176, 0.0214, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:30:56,915 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:15,918 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180292.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:31:30,826 INFO [train.py:904] (0/8) Epoch 18, batch 7750, loss[loss=0.2109, simple_loss=0.3048, pruned_loss=0.05851, over 16829.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2929, pruned_loss=0.06172, over 3090616.51 frames. ], batch size: 96, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:31:50,433 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7869, 1.3587, 1.7076, 1.6769, 1.7902, 1.9545, 1.5486, 1.7492], device='cuda:0'), covar=tensor([0.0228, 0.0353, 0.0196, 0.0255, 0.0236, 0.0168, 0.0384, 0.0111], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0186, 0.0173, 0.0175, 0.0185, 0.0143, 0.0189, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:32:23,183 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180336.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:23,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7244, 1.7991, 1.6177, 1.5161, 1.9027, 1.6149, 1.5969, 1.9297], device='cuda:0'), covar=tensor([0.0212, 0.0294, 0.0415, 0.0370, 0.0243, 0.0286, 0.0191, 0.0211], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0224, 0.0218, 0.0219, 0.0227, 0.0223, 0.0226, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:32:28,055 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 18:32:29,343 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180340.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:32:32,601 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-04-30 18:32:46,614 INFO [train.py:904] (0/8) Epoch 18, batch 7800, loss[loss=0.2076, simple_loss=0.2939, pruned_loss=0.06072, over 15352.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2946, pruned_loss=0.06318, over 3063226.16 frames. ], batch size: 191, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:32:51,028 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.798e+02 2.996e+02 3.455e+02 4.260e+02 9.369e+02, threshold=6.911e+02, percent-clipped=1.0 2023-04-30 18:33:19,878 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-04-30 18:33:33,637 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180382.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:33:55,626 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180397.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:34:00,879 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1437, 2.3968, 1.9633, 2.1640, 2.7503, 2.4085, 2.8542, 2.9463], device='cuda:0'), covar=tensor([0.0135, 0.0388, 0.0502, 0.0433, 0.0257, 0.0369, 0.0207, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0224, 0.0218, 0.0219, 0.0227, 0.0223, 0.0226, 0.0219], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:34:02,028 INFO [train.py:904] (0/8) Epoch 18, batch 7850, loss[loss=0.2069, simple_loss=0.2985, pruned_loss=0.05766, over 16763.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2948, pruned_loss=0.06295, over 3047949.00 frames. ], batch size: 89, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:34:18,830 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 18:35:16,229 INFO [train.py:904] (0/8) Epoch 18, batch 7900, loss[loss=0.198, simple_loss=0.2867, pruned_loss=0.0546, over 16715.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2934, pruned_loss=0.06176, over 3077611.64 frames. ], batch size: 76, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:35:17,197 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180452.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:35:20,366 INFO [optim.py:368] (0/8) 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 18:35:44,198 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7571, 4.7723, 4.5481, 3.8339, 4.6458, 1.7357, 4.4103, 4.3264], device='cuda:0'), covar=tensor([0.0097, 0.0085, 0.0193, 0.0420, 0.0104, 0.2698, 0.0143, 0.0237], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0138, 0.0184, 0.0168, 0.0159, 0.0195, 0.0173, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:36:31,925 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180500.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:36:34,742 INFO [train.py:904] (0/8) Epoch 18, batch 7950, loss[loss=0.246, simple_loss=0.3071, pruned_loss=0.09245, over 11634.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2939, pruned_loss=0.06245, over 3081626.39 frames. ], batch size: 246, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:37:52,816 INFO [train.py:904] (0/8) Epoch 18, batch 8000, loss[loss=0.2138, simple_loss=0.2954, pruned_loss=0.06604, over 15386.00 frames. ], tot_loss[loss=0.2102, simple_loss=0.2947, pruned_loss=0.06287, over 3088932.28 frames. ], batch size: 190, lr: 3.74e-03, grad_scale: 8.0 2023-04-30 18:37:57,089 INFO [optim.py:368] (0/8) 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 18:38:07,578 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180561.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:08,864 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180562.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 18:38:21,345 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:38:36,635 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:06,228 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-04-30 18:39:10,323 INFO [train.py:904] (0/8) Epoch 18, batch 8050, loss[loss=0.1834, simple_loss=0.2751, pruned_loss=0.04582, over 16274.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2941, pruned_loss=0.06223, over 3090223.95 frames. ], batch size: 35, lr: 3.74e-03, grad_scale: 4.0 2023-04-30 18:39:15,764 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8804, 2.6741, 2.8426, 1.9981, 2.6487, 2.0861, 2.5847, 2.9330], device='cuda:0'), covar=tensor([0.0268, 0.0757, 0.0466, 0.1746, 0.0772, 0.0929, 0.0555, 0.0660], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0158, 0.0163, 0.0149, 0.0141, 0.0127, 0.0141, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:39:23,084 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180610.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 18:39:45,022 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 18:39:50,354 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:39:54,560 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180631.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:40:26,596 INFO [train.py:904] (0/8) Epoch 18, batch 8100, loss[loss=0.2282, simple_loss=0.2954, pruned_loss=0.08051, over 11453.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2935, pruned_loss=0.06163, over 3084151.33 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:40:32,031 INFO [optim.py:368] (0/8) 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 18:40:41,251 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7919, 5.1610, 5.3464, 5.1491, 5.2507, 5.7311, 5.1623, 4.9556], device='cuda:0'), covar=tensor([0.0974, 0.1718, 0.1961, 0.1872, 0.2160, 0.0920, 0.1678, 0.2296], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0559, 0.0620, 0.0471, 0.0626, 0.0645, 0.0488, 0.0632], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 18:41:11,952 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:26,976 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180692.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:41:41,233 INFO [train.py:904] (0/8) Epoch 18, batch 8150, loss[loss=0.1986, simple_loss=0.2756, pruned_loss=0.06085, over 15459.00 frames. ], tot_loss[loss=0.206, simple_loss=0.2912, pruned_loss=0.0604, over 3093066.96 frames. ], batch size: 190, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:42:12,980 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0100, 2.4425, 1.9469, 2.2046, 2.7843, 2.4380, 2.8402, 2.9533], device='cuda:0'), covar=tensor([0.0150, 0.0347, 0.0498, 0.0396, 0.0239, 0.0342, 0.0187, 0.0236], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0222, 0.0218, 0.0217, 0.0225, 0.0222, 0.0226, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:42:14,543 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9497, 4.2069, 4.0312, 4.0400, 3.7652, 3.7827, 3.8358, 4.2061], device='cuda:0'), covar=tensor([0.1065, 0.0856, 0.1001, 0.0880, 0.0741, 0.1756, 0.0972, 0.0970], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0763, 0.0629, 0.0570, 0.0476, 0.0491, 0.0637, 0.0591], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:42:15,789 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6774, 4.4991, 4.6665, 4.8394, 4.9900, 4.4989, 4.9699, 4.9862], device='cuda:0'), covar=tensor([0.1736, 0.1168, 0.1570, 0.0725, 0.0576, 0.0999, 0.0584, 0.0629], device='cuda:0'), in_proj_covar=tensor([0.0595, 0.0731, 0.0859, 0.0746, 0.0559, 0.0588, 0.0609, 0.0704], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:42:24,164 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180730.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:26,666 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180732.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:42:56,048 INFO [train.py:904] (0/8) Epoch 18, batch 8200, loss[loss=0.1982, simple_loss=0.2902, pruned_loss=0.05308, over 16794.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2885, pruned_loss=0.05969, over 3103849.52 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:43:01,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-04-30 18:43:02,073 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.834e+02 3.334e+02 4.232e+02 1.685e+03, threshold=6.669e+02, percent-clipped=1.0 2023-04-30 18:44:01,857 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180793.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:44:14,895 INFO [train.py:904] (0/8) Epoch 18, batch 8250, loss[loss=0.174, simple_loss=0.2625, pruned_loss=0.04273, over 12147.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2872, pruned_loss=0.05663, over 3117022.51 frames. ], batch size: 247, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:44:18,557 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0559, 1.8073, 1.5881, 1.4494, 1.9306, 1.6146, 1.6443, 1.9323], device='cuda:0'), covar=tensor([0.0215, 0.0268, 0.0398, 0.0342, 0.0226, 0.0260, 0.0203, 0.0230], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0221, 0.0217, 0.0216, 0.0224, 0.0221, 0.0225, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:44:39,138 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5946, 2.9858, 3.3105, 2.0259, 2.8902, 2.1288, 3.2941, 3.1608], device='cuda:0'), covar=tensor([0.0269, 0.0820, 0.0487, 0.2004, 0.0756, 0.0997, 0.0612, 0.0896], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0158, 0.0163, 0.0149, 0.0141, 0.0127, 0.0140, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:44:50,310 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6593, 2.6412, 1.8884, 2.8044, 2.1556, 2.8305, 2.1320, 2.4360], device='cuda:0'), covar=tensor([0.0274, 0.0372, 0.1207, 0.0320, 0.0661, 0.0524, 0.1154, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0190, 0.0150, 0.0173, 0.0211, 0.0199, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:44:53,348 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=180825.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:45:37,342 INFO [train.py:904] (0/8) Epoch 18, batch 8300, loss[loss=0.1788, simple_loss=0.2639, pruned_loss=0.04687, over 12386.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2847, pruned_loss=0.05423, over 3090236.93 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:45:38,269 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 18:45:43,844 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.692e+02 2.270e+02 2.749e+02 3.299e+02 7.574e+02, threshold=5.499e+02, percent-clipped=1.0 2023-04-30 18:45:52,368 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180861.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:21,649 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9199, 2.7060, 2.9225, 2.0563, 2.6810, 2.1337, 2.7768, 2.8910], device='cuda:0'), covar=tensor([0.0307, 0.0911, 0.0478, 0.1872, 0.0801, 0.0995, 0.0621, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0156, 0.0161, 0.0147, 0.0140, 0.0126, 0.0139, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:46:27,929 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7618, 3.1810, 3.4447, 2.0940, 2.8687, 2.1934, 3.4625, 3.4015], device='cuda:0'), covar=tensor([0.0288, 0.0815, 0.0479, 0.1929, 0.0820, 0.1036, 0.0576, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0156, 0.0161, 0.0147, 0.0140, 0.0126, 0.0139, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 18:46:32,564 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=180886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:46:58,364 INFO [train.py:904] (0/8) Epoch 18, batch 8350, loss[loss=0.2139, simple_loss=0.2914, pruned_loss=0.0682, over 12299.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2847, pruned_loss=0.05325, over 3071533.04 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:47:00,741 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 18:47:06,746 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9322, 4.9135, 4.6823, 4.0746, 4.7720, 1.7441, 4.5121, 4.5813], device='cuda:0'), covar=tensor([0.0087, 0.0099, 0.0198, 0.0381, 0.0107, 0.2569, 0.0140, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0138, 0.0183, 0.0167, 0.0158, 0.0194, 0.0172, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:47:09,896 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=180909.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:47:35,969 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=180926.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:47:48,794 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1620, 4.1172, 4.4465, 4.4273, 4.4428, 4.2315, 4.1881, 4.2144], device='cuda:0'), covar=tensor([0.0310, 0.0735, 0.0451, 0.0396, 0.0451, 0.0406, 0.0934, 0.0463], device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0419, 0.0408, 0.0385, 0.0453, 0.0430, 0.0525, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 18:48:13,275 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8096, 4.6211, 4.8699, 5.0083, 5.1766, 4.6087, 5.1820, 5.1702], device='cuda:0'), covar=tensor([0.1968, 0.1329, 0.1603, 0.0743, 0.0567, 0.0964, 0.0546, 0.0711], device='cuda:0'), in_proj_covar=tensor([0.0589, 0.0729, 0.0853, 0.0744, 0.0556, 0.0588, 0.0606, 0.0702], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:48:16,308 INFO [train.py:904] (0/8) Epoch 18, batch 8400, loss[loss=0.1705, simple_loss=0.2489, pruned_loss=0.04605, over 12539.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2821, pruned_loss=0.05121, over 3072034.79 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:48:19,832 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 18:48:22,153 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.789e+02 2.220e+02 2.634e+02 3.245e+02 6.969e+02, threshold=5.268e+02, percent-clipped=3.0 2023-04-30 18:48:51,837 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6604, 2.6004, 1.8884, 2.7619, 2.1207, 2.7745, 2.0859, 2.3929], device='cuda:0'), covar=tensor([0.0273, 0.0316, 0.1179, 0.0277, 0.0602, 0.0422, 0.1236, 0.0564], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0188, 0.0148, 0.0171, 0.0207, 0.0197, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 18:49:16,492 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=180992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:49:27,502 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-04-30 18:49:31,287 INFO [train.py:904] (0/8) Epoch 18, batch 8450, loss[loss=0.1753, simple_loss=0.2617, pruned_loss=0.04445, over 12615.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.28, pruned_loss=0.04937, over 3061378.45 frames. ], batch size: 248, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 18:50:31,763 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181040.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:50:50,507 INFO [train.py:904] (0/8) Epoch 18, batch 8500, loss[loss=0.1752, simple_loss=0.2658, pruned_loss=0.04237, over 16940.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.2763, pruned_loss=0.04722, over 3057256.98 frames. ], batch size: 109, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:50:58,705 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.304e+02 3.083e+02 3.822e+02 7.945e+02, threshold=6.166e+02, percent-clipped=7.0 2023-04-30 18:51:48,658 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181088.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:52:12,280 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0361, 1.8357, 1.6789, 1.4868, 1.9915, 1.6358, 1.6091, 1.8972], device='cuda:0'), covar=tensor([0.0163, 0.0251, 0.0335, 0.0355, 0.0202, 0.0252, 0.0171, 0.0209], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0218, 0.0213, 0.0213, 0.0221, 0.0218, 0.0221, 0.0213], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:52:13,303 INFO [train.py:904] (0/8) Epoch 18, batch 8550, loss[loss=0.1775, simple_loss=0.2566, pruned_loss=0.04921, over 12122.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2734, pruned_loss=0.04597, over 3038086.85 frames. ], batch size: 247, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:53:50,715 INFO [train.py:904] (0/8) Epoch 18, batch 8600, loss[loss=0.1763, simple_loss=0.2626, pruned_loss=0.04498, over 12401.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2738, pruned_loss=0.0453, over 3021216.89 frames. ], batch size: 247, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:53:54,242 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7012, 4.7088, 4.4987, 3.8539, 4.5553, 1.7497, 4.3554, 4.3690], device='cuda:0'), covar=tensor([0.0069, 0.0071, 0.0171, 0.0283, 0.0091, 0.2535, 0.0116, 0.0188], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0137, 0.0181, 0.0165, 0.0157, 0.0194, 0.0171, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:54:01,222 INFO [optim.py:368] (0/8) 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 18:54:09,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4328, 4.4432, 4.3022, 3.5887, 4.3203, 1.6226, 4.1122, 4.1412], device='cuda:0'), covar=tensor([0.0089, 0.0077, 0.0158, 0.0268, 0.0089, 0.2588, 0.0117, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0137, 0.0181, 0.0166, 0.0158, 0.0194, 0.0172, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 18:54:49,999 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181181.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:55:06,737 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9695, 3.6646, 4.0781, 2.0344, 4.2793, 4.3201, 3.2693, 3.2400], device='cuda:0'), covar=tensor([0.0589, 0.0238, 0.0169, 0.1099, 0.0053, 0.0107, 0.0334, 0.0380], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0102, 0.0090, 0.0134, 0.0072, 0.0115, 0.0120, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 18:55:29,858 INFO [train.py:904] (0/8) Epoch 18, batch 8650, loss[loss=0.1588, simple_loss=0.2634, pruned_loss=0.02715, over 16801.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2721, pruned_loss=0.04351, over 3032165.92 frames. ], batch size: 76, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:56:25,732 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181226.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:57:15,678 INFO [train.py:904] (0/8) Epoch 18, batch 8700, loss[loss=0.1853, simple_loss=0.2858, pruned_loss=0.04236, over 15292.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2696, pruned_loss=0.04247, over 3034021.66 frames. ], batch size: 191, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:57:25,082 INFO [optim.py:368] (0/8) 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 18:57:56,332 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181274.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:39,127 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181297.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 18:58:48,992 INFO [train.py:904] (0/8) Epoch 18, batch 8750, loss[loss=0.185, simple_loss=0.2861, pruned_loss=0.04196, over 16833.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2691, pruned_loss=0.04193, over 3043958.35 frames. ], batch size: 124, lr: 3.73e-03, grad_scale: 4.0 2023-04-30 18:59:02,471 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5068, 2.2630, 2.3924, 4.2315, 2.1236, 2.6490, 2.3891, 2.4383], device='cuda:0'), covar=tensor([0.1115, 0.3732, 0.2836, 0.0443, 0.4530, 0.2614, 0.3532, 0.3586], device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0420, 0.0347, 0.0311, 0.0420, 0.0481, 0.0389, 0.0489], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:00:41,592 INFO [train.py:904] (0/8) Epoch 18, batch 8800, loss[loss=0.1763, simple_loss=0.2698, pruned_loss=0.04139, over 16286.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2678, pruned_loss=0.04109, over 3041605.03 frames. ], batch size: 146, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:00:51,215 INFO [optim.py:368] (0/8) 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 19:00:54,119 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181358.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:01:41,589 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-04-30 19:01:57,932 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181388.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:02:27,193 INFO [train.py:904] (0/8) Epoch 18, batch 8850, loss[loss=0.1649, simple_loss=0.2644, pruned_loss=0.0327, over 15303.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2704, pruned_loss=0.04068, over 3048535.67 frames. ], batch size: 191, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:03:44,281 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181436.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:04:15,180 INFO [train.py:904] (0/8) Epoch 18, batch 8900, loss[loss=0.1721, simple_loss=0.2702, pruned_loss=0.03695, over 16999.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2704, pruned_loss=0.03962, over 3064906.72 frames. ], batch size: 55, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:04:25,753 INFO [optim.py:368] (0/8) 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 19:04:57,336 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-04-30 19:05:23,278 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:05:55,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3635, 3.8141, 3.8050, 2.4615, 3.4397, 3.7889, 3.5370, 2.1838], device='cuda:0'), covar=tensor([0.0490, 0.0035, 0.0038, 0.0366, 0.0088, 0.0075, 0.0066, 0.0420], device='cuda:0'), in_proj_covar=tensor([0.0129, 0.0075, 0.0075, 0.0126, 0.0089, 0.0098, 0.0087, 0.0120], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 19:06:18,290 INFO [train.py:904] (0/8) Epoch 18, batch 8950, loss[loss=0.1587, simple_loss=0.2508, pruned_loss=0.0333, over 15356.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2702, pruned_loss=0.04037, over 3043190.78 frames. ], batch size: 191, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:06:22,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4445, 3.3875, 2.7096, 2.1898, 2.1879, 2.2683, 3.4485, 3.0028], device='cuda:0'), covar=tensor([0.2783, 0.0657, 0.1709, 0.2753, 0.2426, 0.2064, 0.0430, 0.1301], device='cuda:0'), in_proj_covar=tensor([0.0312, 0.0255, 0.0290, 0.0295, 0.0279, 0.0241, 0.0278, 0.0315], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:07:17,053 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=181529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:07:19,841 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181530.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:07:37,583 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-04-30 19:08:08,297 INFO [train.py:904] (0/8) Epoch 18, batch 9000, loss[loss=0.1543, simple_loss=0.2457, pruned_loss=0.03147, over 16962.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2667, pruned_loss=0.03878, over 3055436.44 frames. ], batch size: 116, lr: 3.73e-03, grad_scale: 8.0 2023-04-30 19:08:08,298 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 19:08:14,382 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5605, 4.0436, 4.3991, 2.4970, 4.6924, 4.6967, 3.7390, 3.8263], device='cuda:0'), covar=tensor([0.0450, 0.0206, 0.0187, 0.1034, 0.0041, 0.0091, 0.0274, 0.0293], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0101, 0.0089, 0.0134, 0.0072, 0.0115, 0.0120, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 19:08:17,833 INFO [train.py:938] (0/8) Epoch 18, validation: loss=0.1475, simple_loss=0.2516, pruned_loss=0.02169, over 944034.00 frames. 2023-04-30 19:08:17,834 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 19:08:27,944 INFO [optim.py:368] (0/8) 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 19:08:37,694 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-04-30 19:09:39,036 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=181591.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:10:01,204 INFO [train.py:904] (0/8) Epoch 18, batch 9050, loss[loss=0.1635, simple_loss=0.253, pruned_loss=0.03698, over 13017.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2677, pruned_loss=0.03914, over 3069510.46 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:11,846 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-04-30 19:11:46,782 INFO [train.py:904] (0/8) Epoch 18, batch 9100, loss[loss=0.1896, simple_loss=0.2894, pruned_loss=0.04489, over 16915.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2675, pruned_loss=0.03986, over 3072220.80 frames. ], batch size: 109, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:11:49,675 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181653.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:11:55,856 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.522e+02 2.401e+02 2.871e+02 3.524e+02 6.480e+02, threshold=5.743e+02, percent-clipped=6.0 2023-04-30 19:12:49,817 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 19:13:25,706 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-04-30 19:13:28,076 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-30 19:13:43,592 INFO [train.py:904] (0/8) Epoch 18, batch 9150, loss[loss=0.1698, simple_loss=0.255, pruned_loss=0.04234, over 11800.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2683, pruned_loss=0.03972, over 3064484.47 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:15:27,792 INFO [train.py:904] (0/8) Epoch 18, batch 9200, loss[loss=0.1829, simple_loss=0.2748, pruned_loss=0.04554, over 15257.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2636, pruned_loss=0.0386, over 3059070.61 frames. ], batch size: 190, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:15:36,946 INFO [optim.py:368] (0/8) 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 19:15:41,785 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5142, 4.8584, 4.6308, 4.5949, 4.4138, 4.3919, 4.2442, 4.9050], device='cuda:0'), covar=tensor([0.1079, 0.0840, 0.0936, 0.0808, 0.0777, 0.1163, 0.1221, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0754, 0.0621, 0.0563, 0.0474, 0.0488, 0.0632, 0.0587], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:16:56,936 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-04-30 19:17:05,947 INFO [train.py:904] (0/8) Epoch 18, batch 9250, loss[loss=0.1675, simple_loss=0.2574, pruned_loss=0.03878, over 16530.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2633, pruned_loss=0.03857, over 3056932.21 frames. ], batch size: 147, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:18:57,439 INFO [train.py:904] (0/8) Epoch 18, batch 9300, loss[loss=0.1522, simple_loss=0.2536, pruned_loss=0.02537, over 16924.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2619, pruned_loss=0.03798, over 3053995.40 frames. ], batch size: 102, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:19:06,082 INFO [optim.py:368] (0/8) 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 19:20:12,461 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=181886.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:20:17,204 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5076, 3.7346, 2.8786, 2.1600, 2.3740, 2.4365, 3.9475, 3.2800], device='cuda:0'), covar=tensor([0.3135, 0.0663, 0.1889, 0.3070, 0.2870, 0.2040, 0.0472, 0.1425], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0255, 0.0290, 0.0294, 0.0278, 0.0241, 0.0278, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:20:41,430 INFO [train.py:904] (0/8) Epoch 18, batch 9350, loss[loss=0.1789, simple_loss=0.2688, pruned_loss=0.0445, over 16200.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2614, pruned_loss=0.03779, over 3048380.20 frames. ], batch size: 165, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:21:37,000 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0456, 3.0524, 1.8699, 3.3094, 2.3355, 3.2976, 2.0662, 2.5591], device='cuda:0'), covar=tensor([0.0326, 0.0447, 0.1769, 0.0287, 0.0921, 0.0553, 0.1707, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0166, 0.0185, 0.0144, 0.0169, 0.0201, 0.0193, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 19:22:11,657 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4019, 3.2937, 2.6021, 2.1352, 2.1264, 2.2029, 3.3394, 2.9780], device='cuda:0'), covar=tensor([0.2796, 0.0693, 0.1879, 0.2900, 0.2708, 0.2108, 0.0422, 0.1313], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0255, 0.0290, 0.0294, 0.0278, 0.0241, 0.0278, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:22:24,040 INFO [train.py:904] (0/8) Epoch 18, batch 9400, loss[loss=0.1892, simple_loss=0.2898, pruned_loss=0.04431, over 16659.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2616, pruned_loss=0.03762, over 3041325.96 frames. ], batch size: 134, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:22:25,090 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6197, 2.9341, 3.2566, 1.8870, 2.7918, 2.0907, 3.1801, 3.1529], device='cuda:0'), covar=tensor([0.0254, 0.0838, 0.0515, 0.2040, 0.0820, 0.0986, 0.0688, 0.0928], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0150, 0.0158, 0.0144, 0.0137, 0.0123, 0.0137, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 19:22:25,483 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 19:22:27,729 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=181953.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:22:33,109 INFO [optim.py:368] (0/8) 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 19:23:16,351 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=181978.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:24:00,350 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-182000.pt 2023-04-30 19:24:05,164 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182001.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:24:06,048 INFO [train.py:904] (0/8) Epoch 18, batch 9450, loss[loss=0.1549, simple_loss=0.2502, pruned_loss=0.02981, over 16482.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2633, pruned_loss=0.03779, over 3037655.50 frames. ], batch size: 68, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:19,115 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182039.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:25:41,967 INFO [train.py:904] (0/8) Epoch 18, batch 9500, loss[loss=0.1727, simple_loss=0.2658, pruned_loss=0.03979, over 16605.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2614, pruned_loss=0.03665, over 3041590.35 frames. ], batch size: 57, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:25:55,089 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.398e+02 2.182e+02 2.603e+02 3.134e+02 5.554e+02, threshold=5.207e+02, percent-clipped=1.0 2023-04-30 19:26:37,088 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5919, 3.6448, 3.4566, 3.1295, 3.3051, 3.5482, 3.3166, 3.4229], device='cuda:0'), covar=tensor([0.0506, 0.0489, 0.0268, 0.0249, 0.0462, 0.0408, 0.1279, 0.0432], device='cuda:0'), in_proj_covar=tensor([0.0265, 0.0375, 0.0311, 0.0299, 0.0318, 0.0348, 0.0215, 0.0368], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:27:27,072 INFO [train.py:904] (0/8) Epoch 18, batch 9550, loss[loss=0.1765, simple_loss=0.2605, pruned_loss=0.04628, over 12794.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.262, pruned_loss=0.03717, over 3041285.73 frames. ], batch size: 250, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:29:06,828 INFO [train.py:904] (0/8) Epoch 18, batch 9600, loss[loss=0.1871, simple_loss=0.2876, pruned_loss=0.04332, over 15357.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2635, pruned_loss=0.03792, over 3043145.87 frames. ], batch size: 191, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:29:07,681 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1500, 2.0101, 2.2309, 3.8039, 1.9102, 2.3296, 2.1334, 2.1612], device='cuda:0'), covar=tensor([0.1454, 0.4636, 0.3233, 0.0617, 0.5617, 0.3072, 0.4173, 0.4039], device='cuda:0'), in_proj_covar=tensor([0.0378, 0.0420, 0.0349, 0.0310, 0.0421, 0.0479, 0.0391, 0.0487], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:29:15,386 INFO [optim.py:368] (0/8) 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 19:30:14,205 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182186.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:30:41,429 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9677, 2.7411, 2.9086, 2.0252, 2.6820, 2.1220, 2.6768, 2.8764], device='cuda:0'), covar=tensor([0.0248, 0.0751, 0.0514, 0.1756, 0.0771, 0.0921, 0.0627, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0150, 0.0159, 0.0145, 0.0137, 0.0123, 0.0136, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 19:30:52,035 INFO [train.py:904] (0/8) Epoch 18, batch 9650, loss[loss=0.183, simple_loss=0.279, pruned_loss=0.04348, over 16388.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2651, pruned_loss=0.0384, over 3026828.04 frames. ], batch size: 146, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:00,669 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182234.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:32:37,732 INFO [train.py:904] (0/8) Epoch 18, batch 9700, loss[loss=0.17, simple_loss=0.2646, pruned_loss=0.03771, over 16328.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2641, pruned_loss=0.03819, over 3022848.95 frames. ], batch size: 146, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:32:45,797 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.275e+02 2.741e+02 3.246e+02 5.302e+02, threshold=5.483e+02, percent-clipped=0.0 2023-04-30 19:33:37,414 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182281.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:34:05,794 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7161, 2.4675, 2.5374, 4.5932, 2.3638, 2.7835, 2.5266, 2.7604], device='cuda:0'), covar=tensor([0.0966, 0.3551, 0.2681, 0.0329, 0.3902, 0.2444, 0.3563, 0.2871], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0418, 0.0347, 0.0309, 0.0419, 0.0477, 0.0389, 0.0484], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:34:17,895 INFO [train.py:904] (0/8) Epoch 18, batch 9750, loss[loss=0.1677, simple_loss=0.2694, pruned_loss=0.03301, over 15394.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2626, pruned_loss=0.03822, over 3011705.40 frames. ], batch size: 191, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:35:00,175 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-04-30 19:35:24,333 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182334.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:39,031 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182342.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:35:56,062 INFO [train.py:904] (0/8) Epoch 18, batch 9800, loss[loss=0.173, simple_loss=0.276, pruned_loss=0.03494, over 16903.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2628, pruned_loss=0.03735, over 3033262.26 frames. ], batch size: 116, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:36:05,431 INFO [optim.py:368] (0/8) 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 19:37:39,088 INFO [train.py:904] (0/8) Epoch 18, batch 9850, loss[loss=0.1534, simple_loss=0.2433, pruned_loss=0.03176, over 12732.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2641, pruned_loss=0.03695, over 3056718.32 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:14,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0525, 5.0928, 5.4977, 5.4270, 5.4558, 5.2005, 5.1036, 4.9005], device='cuda:0'), covar=tensor([0.0375, 0.0660, 0.0453, 0.0595, 0.0545, 0.0511, 0.1046, 0.0477], device='cuda:0'), in_proj_covar=tensor([0.0360, 0.0391, 0.0382, 0.0363, 0.0427, 0.0403, 0.0488, 0.0325], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 19:39:29,926 INFO [train.py:904] (0/8) Epoch 18, batch 9900, loss[loss=0.1778, simple_loss=0.2566, pruned_loss=0.0495, over 12376.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2646, pruned_loss=0.03719, over 3049469.10 frames. ], batch size: 248, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:39:40,683 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.082e+02 2.451e+02 2.864e+02 7.377e+02, threshold=4.903e+02, percent-clipped=2.0 2023-04-30 19:41:28,677 INFO [train.py:904] (0/8) Epoch 18, batch 9950, loss[loss=0.1623, simple_loss=0.2564, pruned_loss=0.03414, over 16569.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2658, pruned_loss=0.03712, over 3053667.40 frames. ], batch size: 57, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:41:58,426 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:42:26,534 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-04-30 19:42:41,350 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8659, 3.2646, 3.1971, 2.1117, 2.9936, 3.2004, 3.0660, 1.9633], device='cuda:0'), covar=tensor([0.0554, 0.0040, 0.0048, 0.0372, 0.0096, 0.0077, 0.0070, 0.0435], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0075, 0.0075, 0.0127, 0.0090, 0.0099, 0.0087, 0.0121], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 19:43:17,826 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 19:43:31,289 INFO [train.py:904] (0/8) Epoch 18, batch 10000, loss[loss=0.1455, simple_loss=0.2461, pruned_loss=0.0224, over 16744.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2647, pruned_loss=0.03654, over 3075494.43 frames. ], batch size: 83, lr: 3.72e-03, grad_scale: 8.0 2023-04-30 19:43:42,215 INFO [optim.py:368] (0/8) 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 19:44:16,822 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182575.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:45:14,027 INFO [train.py:904] (0/8) Epoch 18, batch 10050, loss[loss=0.1773, simple_loss=0.2689, pruned_loss=0.04282, over 16362.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2647, pruned_loss=0.03633, over 3082207.80 frames. ], batch size: 146, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:45:51,314 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.83 vs. limit=5.0 2023-04-30 19:46:15,045 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182634.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:18,420 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6320, 3.6816, 3.4748, 3.1424, 3.3281, 3.5887, 3.3367, 3.4423], device='cuda:0'), covar=tensor([0.0529, 0.0590, 0.0264, 0.0236, 0.0486, 0.0471, 0.1334, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0260, 0.0368, 0.0307, 0.0293, 0.0312, 0.0341, 0.0208, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-04-30 19:46:20,795 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182637.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:46:25,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4792, 3.3414, 2.7087, 2.2124, 2.1928, 2.3362, 3.4314, 2.9609], device='cuda:0'), covar=tensor([0.2721, 0.0661, 0.1650, 0.2673, 0.2613, 0.1912, 0.0400, 0.1351], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0251, 0.0286, 0.0290, 0.0273, 0.0238, 0.0275, 0.0309], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:46:46,728 INFO [train.py:904] (0/8) Epoch 18, batch 10100, loss[loss=0.1617, simple_loss=0.2482, pruned_loss=0.03758, over 12454.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2653, pruned_loss=0.03672, over 3068620.31 frames. ], batch size: 246, lr: 3.71e-03, grad_scale: 8.0 2023-04-30 19:46:55,748 INFO [optim.py:368] (0/8) 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 19:47:44,813 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182682.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:47:58,940 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9800, 4.2120, 4.5208, 4.4829, 4.4869, 4.2046, 3.9700, 4.1241], device='cuda:0'), covar=tensor([0.0610, 0.0856, 0.0500, 0.0625, 0.0634, 0.0668, 0.1324, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0364, 0.0397, 0.0388, 0.0368, 0.0431, 0.0409, 0.0493, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 19:48:05,619 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8364, 1.2909, 1.7837, 1.6828, 1.8668, 1.9658, 1.6637, 1.8396], device='cuda:0'), covar=tensor([0.0212, 0.0402, 0.0189, 0.0284, 0.0261, 0.0211, 0.0372, 0.0128], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0180, 0.0169, 0.0168, 0.0179, 0.0138, 0.0183, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:48:08,595 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-18.pt 2023-04-30 19:48:32,146 INFO [train.py:904] (0/8) Epoch 19, batch 0, loss[loss=0.2806, simple_loss=0.3328, pruned_loss=0.1142, over 16769.00 frames. ], tot_loss[loss=0.2806, simple_loss=0.3328, pruned_loss=0.1142, over 16769.00 frames. ], batch size: 124, lr: 3.61e-03, grad_scale: 8.0 2023-04-30 19:48:32,147 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 19:48:39,772 INFO [train.py:938] (0/8) Epoch 19, validation: loss=0.1468, simple_loss=0.2504, pruned_loss=0.0216, over 944034.00 frames. 2023-04-30 19:48:39,773 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 19:49:08,384 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0101, 4.3208, 2.7816, 2.4609, 2.7578, 2.2651, 4.4781, 3.5254], device='cuda:0'), covar=tensor([0.2846, 0.0641, 0.2230, 0.2890, 0.3192, 0.2413, 0.0520, 0.1411], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0253, 0.0288, 0.0292, 0.0275, 0.0239, 0.0276, 0.0311], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:49:26,610 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0601, 2.5203, 1.9758, 2.3282, 2.8481, 2.6978, 3.0192, 2.9466], device='cuda:0'), covar=tensor([0.0165, 0.0378, 0.0529, 0.0437, 0.0262, 0.0343, 0.0203, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0225, 0.0218, 0.0218, 0.0225, 0.0223, 0.0222, 0.0215], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:49:33,019 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8534, 2.9140, 2.6173, 4.9137, 3.9520, 4.2907, 1.5825, 3.1251], device='cuda:0'), covar=tensor([0.1358, 0.0748, 0.1232, 0.0201, 0.0277, 0.0439, 0.1642, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0166, 0.0187, 0.0173, 0.0193, 0.0208, 0.0192, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 19:49:50,021 INFO [train.py:904] (0/8) Epoch 19, batch 50, loss[loss=0.1944, simple_loss=0.2726, pruned_loss=0.05809, over 16821.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2724, pruned_loss=0.05225, over 759399.75 frames. ], batch size: 124, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:49:54,909 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4087, 3.4733, 3.7962, 2.6550, 3.4469, 3.8117, 3.5751, 2.2251], device='cuda:0'), covar=tensor([0.0498, 0.0183, 0.0045, 0.0350, 0.0109, 0.0080, 0.0086, 0.0441], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0076, 0.0076, 0.0128, 0.0091, 0.0100, 0.0088, 0.0122], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 19:49:59,029 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.537e+02 2.571e+02 3.042e+02 3.790e+02 6.505e+02, threshold=6.085e+02, percent-clipped=6.0 2023-04-30 19:50:55,408 INFO [train.py:904] (0/8) Epoch 19, batch 100, loss[loss=0.1945, simple_loss=0.2709, pruned_loss=0.05905, over 16885.00 frames. ], tot_loss[loss=0.185, simple_loss=0.2699, pruned_loss=0.0501, over 1328784.63 frames. ], batch size: 116, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:03,146 INFO [train.py:904] (0/8) Epoch 19, batch 150, loss[loss=0.1653, simple_loss=0.2447, pruned_loss=0.04299, over 16890.00 frames. ], tot_loss[loss=0.1822, simple_loss=0.2676, pruned_loss=0.04837, over 1767234.09 frames. ], batch size: 96, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:52:09,679 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182856.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:15,181 INFO [optim.py:368] (0/8) 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 19:52:18,070 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=182862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:28,097 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=182870.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:52:28,295 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1701, 4.0572, 4.5646, 2.4108, 4.6935, 4.8109, 3.5590, 3.5811], device='cuda:0'), covar=tensor([0.0671, 0.0255, 0.0171, 0.1147, 0.0055, 0.0130, 0.0388, 0.0388], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0105, 0.0092, 0.0138, 0.0075, 0.0118, 0.0124, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 19:52:38,528 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2955, 3.4899, 3.8707, 2.1291, 3.1218, 2.4665, 3.7411, 3.7382], device='cuda:0'), covar=tensor([0.0229, 0.0896, 0.0473, 0.1889, 0.0814, 0.0929, 0.0559, 0.0871], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0151, 0.0160, 0.0147, 0.0138, 0.0124, 0.0137, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 19:53:13,455 INFO [train.py:904] (0/8) Epoch 19, batch 200, loss[loss=0.1683, simple_loss=0.2715, pruned_loss=0.03261, over 17156.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.269, pruned_loss=0.04882, over 2104834.58 frames. ], batch size: 48, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:53:34,307 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182917.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:53:42,588 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=182923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:00,404 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=182937.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:54:03,327 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1333, 5.6978, 5.7994, 5.5163, 5.5845, 6.1895, 5.6942, 5.4301], device='cuda:0'), covar=tensor([0.0853, 0.2012, 0.2272, 0.2033, 0.2543, 0.0963, 0.1514, 0.2357], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0550, 0.0613, 0.0464, 0.0617, 0.0642, 0.0482, 0.0620], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 19:54:21,122 INFO [train.py:904] (0/8) Epoch 19, batch 250, loss[loss=0.1451, simple_loss=0.2361, pruned_loss=0.02701, over 17251.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2663, pruned_loss=0.04749, over 2374359.64 frames. ], batch size: 45, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:54:27,727 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6629, 2.8807, 2.7018, 5.0667, 4.1110, 4.5459, 1.6731, 3.2800], device='cuda:0'), covar=tensor([0.1415, 0.0761, 0.1159, 0.0155, 0.0204, 0.0345, 0.1555, 0.0680], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0168, 0.0190, 0.0176, 0.0197, 0.0210, 0.0194, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 19:54:32,990 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.641e+02 2.306e+02 2.752e+02 3.234e+02 1.372e+03, threshold=5.503e+02, percent-clipped=2.0 2023-04-30 19:55:04,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0005, 4.0946, 2.5269, 4.6907, 3.2909, 4.6019, 2.8403, 3.4682], device='cuda:0'), covar=tensor([0.0250, 0.0371, 0.1556, 0.0234, 0.0744, 0.0552, 0.1311, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0169, 0.0189, 0.0150, 0.0172, 0.0206, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 19:55:08,839 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=182985.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:55:30,557 INFO [train.py:904] (0/8) Epoch 19, batch 300, loss[loss=0.1756, simple_loss=0.257, pruned_loss=0.04705, over 16770.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2633, pruned_loss=0.04601, over 2594762.20 frames. ], batch size: 102, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:22,936 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4055, 3.5272, 4.0306, 2.1604, 3.2641, 2.4949, 3.8196, 3.7473], device='cuda:0'), covar=tensor([0.0275, 0.0949, 0.0455, 0.1985, 0.0771, 0.1011, 0.0635, 0.1091], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0154, 0.0161, 0.0148, 0.0140, 0.0125, 0.0139, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 19:56:41,139 INFO [train.py:904] (0/8) Epoch 19, batch 350, loss[loss=0.1864, simple_loss=0.2612, pruned_loss=0.05583, over 16410.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2604, pruned_loss=0.04494, over 2753918.98 frames. ], batch size: 146, lr: 3.61e-03, grad_scale: 1.0 2023-04-30 19:56:52,342 INFO [optim.py:368] (0/8) 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 19:57:37,508 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-04-30 19:57:51,266 INFO [train.py:904] (0/8) Epoch 19, batch 400, loss[loss=0.189, simple_loss=0.2583, pruned_loss=0.05979, over 16787.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2584, pruned_loss=0.0455, over 2879071.77 frames. ], batch size: 124, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:58:32,699 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5647, 2.2366, 2.3486, 4.2295, 2.2400, 2.6065, 2.3749, 2.3697], device='cuda:0'), covar=tensor([0.1135, 0.3775, 0.2921, 0.0502, 0.4272, 0.2739, 0.3545, 0.3932], device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0431, 0.0357, 0.0320, 0.0429, 0.0491, 0.0401, 0.0500], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 19:58:39,083 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1488, 4.1085, 4.4882, 4.4631, 4.4997, 4.2261, 4.2578, 4.1612], device='cuda:0'), covar=tensor([0.0406, 0.0725, 0.0415, 0.0431, 0.0497, 0.0451, 0.0799, 0.0549], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0421, 0.0410, 0.0388, 0.0454, 0.0432, 0.0521, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 19:59:02,561 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0130, 4.4613, 3.1041, 2.4189, 2.7744, 2.5288, 4.6686, 3.6617], device='cuda:0'), covar=tensor([0.2637, 0.0515, 0.1764, 0.2801, 0.2883, 0.2152, 0.0387, 0.1293], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0261, 0.0296, 0.0301, 0.0286, 0.0247, 0.0284, 0.0323], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 19:59:03,159 INFO [train.py:904] (0/8) Epoch 19, batch 450, loss[loss=0.1602, simple_loss=0.2538, pruned_loss=0.03329, over 17123.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2568, pruned_loss=0.04423, over 2973622.09 frames. ], batch size: 48, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 19:59:14,100 INFO [optim.py:368] (0/8) 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 19:59:28,270 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183170.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 19:59:42,118 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1309, 4.9279, 5.1628, 5.3675, 5.5578, 4.7501, 5.5091, 5.5480], device='cuda:0'), covar=tensor([0.1854, 0.1318, 0.1701, 0.0823, 0.0516, 0.0981, 0.0596, 0.0632], device='cuda:0'), in_proj_covar=tensor([0.0603, 0.0742, 0.0865, 0.0760, 0.0564, 0.0594, 0.0617, 0.0710], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:00:12,290 INFO [train.py:904] (0/8) Epoch 19, batch 500, loss[loss=0.1839, simple_loss=0.2651, pruned_loss=0.05141, over 16830.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2553, pruned_loss=0.04339, over 3045234.94 frames. ], batch size: 102, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:00:27,757 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,917 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:35,934 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183218.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:00:46,516 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3175, 5.2650, 5.1485, 4.5752, 4.7615, 5.2294, 5.1932, 4.7685], device='cuda:0'), covar=tensor([0.0582, 0.0464, 0.0291, 0.0360, 0.1109, 0.0411, 0.0304, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0400, 0.0330, 0.0318, 0.0338, 0.0371, 0.0226, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:01:23,254 INFO [train.py:904] (0/8) Epoch 19, batch 550, loss[loss=0.1663, simple_loss=0.2561, pruned_loss=0.03822, over 17059.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2552, pruned_loss=0.04318, over 3093884.45 frames. ], batch size: 50, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:01:34,919 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.382e+02 2.167e+02 2.560e+02 2.885e+02 5.610e+02, threshold=5.120e+02, percent-clipped=1.0 2023-04-30 20:02:32,735 INFO [train.py:904] (0/8) Epoch 19, batch 600, loss[loss=0.1649, simple_loss=0.2419, pruned_loss=0.04393, over 12491.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.255, pruned_loss=0.04317, over 3142247.65 frames. ], batch size: 246, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:13,268 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3720, 3.1975, 3.4701, 1.7488, 3.5120, 3.5805, 2.9350, 2.7475], device='cuda:0'), covar=tensor([0.0774, 0.0232, 0.0177, 0.1287, 0.0128, 0.0189, 0.0437, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0106, 0.0094, 0.0140, 0.0077, 0.0121, 0.0126, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 20:03:40,859 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3870, 3.5314, 3.6633, 3.6294, 3.6394, 3.4904, 3.5197, 3.5329], device='cuda:0'), covar=tensor([0.0422, 0.0736, 0.0517, 0.0511, 0.0575, 0.0537, 0.0732, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0392, 0.0429, 0.0418, 0.0394, 0.0464, 0.0441, 0.0530, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 20:03:42,740 INFO [train.py:904] (0/8) Epoch 19, batch 650, loss[loss=0.1585, simple_loss=0.2506, pruned_loss=0.03326, over 17194.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2534, pruned_loss=0.04297, over 3178096.13 frames. ], batch size: 46, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:03:54,610 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.479e+02 2.241e+02 2.597e+02 3.196e+02 6.464e+02, threshold=5.194e+02, percent-clipped=2.0 2023-04-30 20:04:32,311 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-04-30 20:04:53,083 INFO [train.py:904] (0/8) Epoch 19, batch 700, loss[loss=0.1711, simple_loss=0.2628, pruned_loss=0.03972, over 17132.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2532, pruned_loss=0.04222, over 3207196.74 frames. ], batch size: 47, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:05:01,034 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:05:50,887 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-04-30 20:05:59,950 INFO [train.py:904] (0/8) Epoch 19, batch 750, loss[loss=0.149, simple_loss=0.2447, pruned_loss=0.02661, over 17126.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2519, pruned_loss=0.04135, over 3236643.33 frames. ], batch size: 47, lr: 3.61e-03, grad_scale: 2.0 2023-04-30 20:06:11,899 INFO [optim.py:368] (0/8) 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 20:06:25,296 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183469.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:06:39,733 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4817, 5.8461, 5.6095, 5.6257, 5.2923, 5.2733, 5.2688, 5.9802], device='cuda:0'), covar=tensor([0.1377, 0.0946, 0.0989, 0.0822, 0.0870, 0.0680, 0.1096, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0654, 0.0801, 0.0658, 0.0593, 0.0502, 0.0511, 0.0665, 0.0619], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:07:10,066 INFO [train.py:904] (0/8) Epoch 19, batch 800, loss[loss=0.166, simple_loss=0.2625, pruned_loss=0.03473, over 17275.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2525, pruned_loss=0.04169, over 3261764.66 frames. ], batch size: 52, lr: 3.61e-03, grad_scale: 4.0 2023-04-30 20:07:24,268 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:07:32,567 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=183518.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:19,011 INFO [train.py:904] (0/8) Epoch 19, batch 850, loss[loss=0.1457, simple_loss=0.2373, pruned_loss=0.02705, over 17188.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2518, pruned_loss=0.04122, over 3273957.33 frames. ], batch size: 46, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:08:20,565 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7295, 2.7662, 2.5065, 4.1411, 3.3236, 4.0519, 1.5395, 2.7963], device='cuda:0'), covar=tensor([0.1480, 0.0752, 0.1228, 0.0209, 0.0209, 0.0442, 0.1704, 0.0916], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0170, 0.0191, 0.0180, 0.0200, 0.0213, 0.0196, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 20:08:29,789 INFO [optim.py:368] (0/8) 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 20:08:30,104 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183560.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:08:38,502 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=183566.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:09:28,468 INFO [train.py:904] (0/8) Epoch 19, batch 900, loss[loss=0.1705, simple_loss=0.2653, pruned_loss=0.03787, over 16986.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2518, pruned_loss=0.0411, over 3279089.19 frames. ], batch size: 53, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:09:44,193 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183614.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:09:54,416 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5474, 3.4750, 3.9212, 1.8584, 4.1437, 4.1327, 3.2058, 2.9831], device='cuda:0'), covar=tensor([0.0877, 0.0326, 0.0274, 0.1390, 0.0091, 0.0197, 0.0408, 0.0520], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0107, 0.0095, 0.0140, 0.0077, 0.0122, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 20:09:59,073 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 20:10:35,713 INFO [train.py:904] (0/8) Epoch 19, batch 950, loss[loss=0.1767, simple_loss=0.2623, pruned_loss=0.0456, over 16678.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2519, pruned_loss=0.04173, over 3291779.30 frames. ], batch size: 57, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:10:45,895 INFO [optim.py:368] (0/8) 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 20:11:08,249 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:11:09,452 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5147, 2.2507, 2.2902, 4.2553, 2.3213, 2.6158, 2.3401, 2.3965], device='cuda:0'), covar=tensor([0.1191, 0.3889, 0.3096, 0.0522, 0.4074, 0.2757, 0.3644, 0.3999], device='cuda:0'), in_proj_covar=tensor([0.0392, 0.0434, 0.0360, 0.0324, 0.0433, 0.0498, 0.0404, 0.0507], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:11:42,847 INFO [train.py:904] (0/8) Epoch 19, batch 1000, loss[loss=0.1574, simple_loss=0.2345, pruned_loss=0.04009, over 15513.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2517, pruned_loss=0.0413, over 3293757.18 frames. ], batch size: 190, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:12:50,598 INFO [train.py:904] (0/8) Epoch 19, batch 1050, loss[loss=0.1677, simple_loss=0.2611, pruned_loss=0.03717, over 17121.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2524, pruned_loss=0.04142, over 3301588.29 frames. ], batch size: 47, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:13:01,769 INFO [optim.py:368] (0/8) 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 20:13:06,566 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:14:00,554 INFO [train.py:904] (0/8) Epoch 19, batch 1100, loss[loss=0.159, simple_loss=0.2409, pruned_loss=0.03853, over 16241.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2519, pruned_loss=0.04165, over 3301804.13 frames. ], batch size: 165, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:14:37,211 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-04-30 20:14:55,809 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=183842.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:15:08,305 INFO [train.py:904] (0/8) Epoch 19, batch 1150, loss[loss=0.1702, simple_loss=0.248, pruned_loss=0.04626, over 16340.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2507, pruned_loss=0.04132, over 3303705.00 frames. ], batch size: 165, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:15:20,304 INFO [optim.py:368] (0/8) 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 20:16:18,553 INFO [train.py:904] (0/8) Epoch 19, batch 1200, loss[loss=0.1846, simple_loss=0.26, pruned_loss=0.05465, over 15474.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2506, pruned_loss=0.0408, over 3305671.87 frames. ], batch size: 190, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:16:19,204 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 20:16:20,038 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=183903.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:17:25,284 INFO [train.py:904] (0/8) Epoch 19, batch 1250, loss[loss=0.1866, simple_loss=0.2574, pruned_loss=0.05788, over 16731.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2511, pruned_loss=0.04173, over 3317592.02 frames. ], batch size: 124, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:17:35,903 INFO [optim.py:368] (0/8) 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 20:17:50,869 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=183970.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:17:51,111 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1498, 3.1881, 3.3985, 2.2318, 2.8986, 2.3128, 3.5638, 3.4706], device='cuda:0'), covar=tensor([0.0248, 0.0928, 0.0549, 0.1811, 0.0846, 0.1008, 0.0579, 0.0963], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0158, 0.0164, 0.0150, 0.0142, 0.0127, 0.0142, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 20:18:02,996 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9618, 2.4423, 1.9726, 2.2880, 2.8958, 2.6757, 2.9664, 3.0050], device='cuda:0'), covar=tensor([0.0234, 0.0422, 0.0521, 0.0435, 0.0220, 0.0337, 0.0238, 0.0260], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0235, 0.0226, 0.0226, 0.0236, 0.0233, 0.0237, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:18:31,204 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-184000.pt 2023-04-30 20:18:37,495 INFO [train.py:904] (0/8) Epoch 19, batch 1300, loss[loss=0.1661, simple_loss=0.264, pruned_loss=0.03409, over 17251.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.25, pruned_loss=0.04144, over 3320408.74 frames. ], batch size: 52, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:44,286 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184050.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:19:46,329 INFO [train.py:904] (0/8) Epoch 19, batch 1350, loss[loss=0.1821, simple_loss=0.2581, pruned_loss=0.05312, over 16400.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2498, pruned_loss=0.04087, over 3324916.94 frames. ], batch size: 146, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:19:55,441 INFO [optim.py:368] (0/8) 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 20:20:03,908 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:20:55,162 INFO [train.py:904] (0/8) Epoch 19, batch 1400, loss[loss=0.1751, simple_loss=0.272, pruned_loss=0.03905, over 17122.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2501, pruned_loss=0.04084, over 3314973.69 frames. ], batch size: 48, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:21:01,468 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 20:21:09,588 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184111.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:10,485 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184112.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:21:30,138 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3733, 5.3828, 5.1132, 4.5961, 5.1735, 2.1342, 4.9354, 5.1509], device='cuda:0'), covar=tensor([0.0075, 0.0082, 0.0191, 0.0413, 0.0107, 0.2575, 0.0150, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0148, 0.0193, 0.0176, 0.0170, 0.0206, 0.0185, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:22:05,091 INFO [train.py:904] (0/8) Epoch 19, batch 1450, loss[loss=0.1646, simple_loss=0.2405, pruned_loss=0.04434, over 15924.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2489, pruned_loss=0.04031, over 3319439.00 frames. ], batch size: 35, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:22:15,461 INFO [optim.py:368] (0/8) 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 20:22:30,546 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5305, 4.4109, 4.4227, 4.1422, 4.1843, 4.4836, 4.2706, 4.2364], device='cuda:0'), covar=tensor([0.0578, 0.0633, 0.0301, 0.0283, 0.0748, 0.0445, 0.0486, 0.0606], device='cuda:0'), in_proj_covar=tensor([0.0292, 0.0417, 0.0344, 0.0332, 0.0352, 0.0388, 0.0235, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:23:08,694 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184198.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:23:13,290 INFO [train.py:904] (0/8) Epoch 19, batch 1500, loss[loss=0.166, simple_loss=0.2634, pruned_loss=0.0343, over 17025.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2491, pruned_loss=0.04078, over 3325178.81 frames. ], batch size: 50, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:22,400 INFO [train.py:904] (0/8) Epoch 19, batch 1550, loss[loss=0.1747, simple_loss=0.27, pruned_loss=0.03969, over 17110.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2504, pruned_loss=0.04198, over 3330840.54 frames. ], batch size: 49, lr: 3.60e-03, grad_scale: 4.0 2023-04-30 20:24:34,829 INFO [optim.py:368] (0/8) 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 20:24:47,405 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184270.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:25:31,176 INFO [train.py:904] (0/8) Epoch 19, batch 1600, loss[loss=0.1504, simple_loss=0.2298, pruned_loss=0.0355, over 16786.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.253, pruned_loss=0.04238, over 3332899.27 frames. ], batch size: 39, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:25:53,348 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184318.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:39,962 INFO [train.py:904] (0/8) Epoch 19, batch 1650, loss[loss=0.1449, simple_loss=0.2301, pruned_loss=0.0299, over 16989.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2535, pruned_loss=0.0429, over 3326471.42 frames. ], batch size: 41, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:26:47,079 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6094, 3.3231, 3.6617, 2.0061, 3.7865, 3.7655, 3.1412, 2.7969], device='cuda:0'), covar=tensor([0.0669, 0.0219, 0.0177, 0.1051, 0.0077, 0.0194, 0.0345, 0.0440], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0107, 0.0095, 0.0139, 0.0077, 0.0123, 0.0125, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 20:26:50,245 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184359.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:26:52,318 INFO [optim.py:368] (0/8) 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 20:27:43,178 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-04-30 20:27:49,677 INFO [train.py:904] (0/8) Epoch 19, batch 1700, loss[loss=0.1808, simple_loss=0.276, pruned_loss=0.04281, over 17046.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2555, pruned_loss=0.04308, over 3331917.83 frames. ], batch size: 53, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:27:55,063 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184406.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:14,034 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:28:23,302 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-04-30 20:28:58,453 INFO [train.py:904] (0/8) Epoch 19, batch 1750, loss[loss=0.1824, simple_loss=0.2673, pruned_loss=0.04877, over 16522.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2559, pruned_loss=0.04311, over 3326155.84 frames. ], batch size: 68, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:29:10,992 INFO [optim.py:368] (0/8) 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 20:29:45,759 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184486.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:02,721 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184498.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:30:07,532 INFO [train.py:904] (0/8) Epoch 19, batch 1800, loss[loss=0.1535, simple_loss=0.2422, pruned_loss=0.03237, over 15794.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2562, pruned_loss=0.04244, over 3329588.88 frames. ], batch size: 35, lr: 3.60e-03, grad_scale: 8.0 2023-04-30 20:31:05,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2205, 4.7112, 4.6733, 3.4787, 3.8786, 4.6278, 4.0346, 2.7961], device='cuda:0'), covar=tensor([0.0379, 0.0051, 0.0032, 0.0287, 0.0118, 0.0070, 0.0075, 0.0416], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0082, 0.0081, 0.0133, 0.0096, 0.0106, 0.0093, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 20:31:07,263 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184546.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:08,666 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184547.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:31:16,707 INFO [train.py:904] (0/8) Epoch 19, batch 1850, loss[loss=0.1571, simple_loss=0.2385, pruned_loss=0.03781, over 16879.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2568, pruned_loss=0.04267, over 3322234.90 frames. ], batch size: 96, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:31:29,854 INFO [optim.py:368] (0/8) 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 20:31:48,026 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0303, 5.0640, 4.8451, 4.3491, 4.9354, 1.9038, 4.6578, 4.7745], device='cuda:0'), covar=tensor([0.0100, 0.0091, 0.0207, 0.0378, 0.0108, 0.2774, 0.0150, 0.0217], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0150, 0.0194, 0.0177, 0.0171, 0.0206, 0.0186, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:32:26,196 INFO [train.py:904] (0/8) Epoch 19, batch 1900, loss[loss=0.151, simple_loss=0.2341, pruned_loss=0.03392, over 17017.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2565, pruned_loss=0.04204, over 3312120.74 frames. ], batch size: 41, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:35,924 INFO [train.py:904] (0/8) Epoch 19, batch 1950, loss[loss=0.1413, simple_loss=0.2305, pruned_loss=0.02603, over 17225.00 frames. ], tot_loss[loss=0.17, simple_loss=0.256, pruned_loss=0.04205, over 3316361.79 frames. ], batch size: 45, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:33:48,872 INFO [optim.py:368] (0/8) 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 20:34:47,057 INFO [train.py:904] (0/8) Epoch 19, batch 2000, loss[loss=0.1634, simple_loss=0.2513, pruned_loss=0.0378, over 17206.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2554, pruned_loss=0.04157, over 3316483.62 frames. ], batch size: 46, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:34:53,209 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=184706.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:06,425 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184715.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:35:09,125 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-04-30 20:35:10,247 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 20:35:49,230 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-04-30 20:35:56,930 INFO [train.py:904] (0/8) Epoch 19, batch 2050, loss[loss=0.1953, simple_loss=0.2783, pruned_loss=0.05613, over 16312.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.256, pruned_loss=0.04185, over 3325290.40 frames. ], batch size: 165, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:35:59,927 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=184754.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:36:10,013 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.173e+02 2.577e+02 2.933e+02 5.913e+02, threshold=5.153e+02, percent-clipped=1.0 2023-04-30 20:37:10,296 INFO [train.py:904] (0/8) Epoch 19, batch 2100, loss[loss=0.1985, simple_loss=0.2814, pruned_loss=0.05782, over 12282.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2571, pruned_loss=0.04242, over 3318145.84 frames. ], batch size: 246, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:37:36,160 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8132, 2.6798, 2.3984, 2.5580, 3.1012, 2.8501, 3.4623, 3.3364], device='cuda:0'), covar=tensor([0.0121, 0.0391, 0.0476, 0.0435, 0.0275, 0.0385, 0.0218, 0.0236], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0236, 0.0225, 0.0227, 0.0236, 0.0236, 0.0239, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:38:05,014 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=184842.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:38:18,978 INFO [train.py:904] (0/8) Epoch 19, batch 2150, loss[loss=0.1465, simple_loss=0.2417, pruned_loss=0.02564, over 17223.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2574, pruned_loss=0.04293, over 3326227.11 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:38:30,744 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.154e+02 2.741e+02 3.179e+02 7.758e+02, threshold=5.482e+02, percent-clipped=2.0 2023-04-30 20:38:38,228 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184866.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:39:27,113 INFO [train.py:904] (0/8) Epoch 19, batch 2200, loss[loss=0.1867, simple_loss=0.274, pruned_loss=0.04972, over 16430.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2584, pruned_loss=0.04352, over 3327068.21 frames. ], batch size: 75, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:01,231 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=184927.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:17,939 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=184939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:40:35,954 INFO [train.py:904] (0/8) Epoch 19, batch 2250, loss[loss=0.1773, simple_loss=0.2547, pruned_loss=0.04998, over 16462.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2594, pruned_loss=0.04373, over 3325778.41 frames. ], batch size: 75, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:40:48,434 INFO [optim.py:368] (0/8) 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 20:41:27,726 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 20:41:36,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2611, 2.1629, 2.3054, 3.9811, 2.2288, 2.5867, 2.2671, 2.3482], device='cuda:0'), covar=tensor([0.1408, 0.3602, 0.2843, 0.0611, 0.3839, 0.2580, 0.3809, 0.3178], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0441, 0.0363, 0.0327, 0.0435, 0.0506, 0.0409, 0.0514], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:41:38,110 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.11 vs. limit=5.0 2023-04-30 20:41:45,029 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185000.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:41:46,814 INFO [train.py:904] (0/8) Epoch 19, batch 2300, loss[loss=0.1812, simple_loss=0.2569, pruned_loss=0.05274, over 16831.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.259, pruned_loss=0.04371, over 3327408.95 frames. ], batch size: 96, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:42:05,822 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185015.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:42:15,454 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-04-30 20:42:16,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3782, 4.0660, 4.1201, 4.5106, 4.6751, 4.2071, 4.5260, 4.6251], device='cuda:0'), covar=tensor([0.1573, 0.1487, 0.2042, 0.1028, 0.0826, 0.1672, 0.2139, 0.1236], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0801, 0.0936, 0.0816, 0.0607, 0.0640, 0.0656, 0.0761], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:42:56,749 INFO [train.py:904] (0/8) Epoch 19, batch 2350, loss[loss=0.2087, simple_loss=0.2818, pruned_loss=0.06781, over 16290.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2588, pruned_loss=0.04362, over 3328615.65 frames. ], batch size: 165, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:43:08,895 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.650e+02 2.245e+02 2.599e+02 3.315e+02 5.469e+02, threshold=5.198e+02, percent-clipped=1.0 2023-04-30 20:43:12,101 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185063.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:43:37,451 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3378, 3.1771, 3.4430, 1.7891, 3.5156, 3.5531, 2.9483, 2.7361], device='cuda:0'), covar=tensor([0.0829, 0.0227, 0.0188, 0.1273, 0.0120, 0.0218, 0.0450, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0109, 0.0097, 0.0142, 0.0079, 0.0126, 0.0129, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 20:44:06,546 INFO [train.py:904] (0/8) Epoch 19, batch 2400, loss[loss=0.1781, simple_loss=0.2741, pruned_loss=0.04109, over 17073.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2598, pruned_loss=0.0439, over 3330862.61 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:44:33,362 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185121.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 20:44:43,169 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:45:02,142 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 19, batch 2450, loss[loss=0.1812, simple_loss=0.2718, pruned_loss=0.04528, over 17054.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.261, pruned_loss=0.04403, over 3334788.10 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:45:27,054 INFO [optim.py:368] (0/8) 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 20:45:33,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6200, 2.3304, 2.3113, 4.5022, 2.3406, 2.8482, 2.4624, 2.5243], device='cuda:0'), covar=tensor([0.1167, 0.3792, 0.2988, 0.0435, 0.4070, 0.2531, 0.3370, 0.3528], device='cuda:0'), in_proj_covar=tensor([0.0396, 0.0439, 0.0362, 0.0327, 0.0434, 0.0505, 0.0408, 0.0512], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:46:08,236 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:08,451 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185190.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:46:23,977 INFO [train.py:904] (0/8) Epoch 19, batch 2500, loss[loss=0.1762, simple_loss=0.267, pruned_loss=0.04268, over 17265.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.261, pruned_loss=0.04387, over 3323933.90 frames. ], batch size: 52, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:46:51,194 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185222.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:47:32,279 INFO [train.py:904] (0/8) Epoch 19, batch 2550, loss[loss=0.1445, simple_loss=0.2354, pruned_loss=0.02681, over 17218.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.261, pruned_loss=0.04395, over 3322639.89 frames. ], batch size: 44, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:47:36,317 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7700, 3.0112, 3.1145, 2.1629, 2.7596, 2.1555, 3.2347, 3.2951], device='cuda:0'), covar=tensor([0.0259, 0.0916, 0.0609, 0.1737, 0.0849, 0.1022, 0.0598, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0161, 0.0165, 0.0151, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 20:47:44,039 INFO [optim.py:368] (0/8) 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 20:47:48,062 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7668, 4.8846, 5.0790, 4.8459, 4.8844, 5.5432, 5.0428, 4.7248], device='cuda:0'), covar=tensor([0.1401, 0.2320, 0.2433, 0.2468, 0.3023, 0.1208, 0.1880, 0.3065], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0596, 0.0659, 0.0496, 0.0667, 0.0699, 0.0515, 0.0667], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 20:48:04,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8847, 4.2973, 4.2413, 2.9644, 3.6877, 4.2185, 3.8868, 2.0784], device='cuda:0'), covar=tensor([0.0493, 0.0081, 0.0070, 0.0435, 0.0151, 0.0140, 0.0161, 0.0677], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0134, 0.0096, 0.0106, 0.0093, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 20:48:31,916 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185295.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:48:40,166 INFO [train.py:904] (0/8) Epoch 19, batch 2600, loss[loss=0.1571, simple_loss=0.2577, pruned_loss=0.02827, over 17119.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2615, pruned_loss=0.04388, over 3317261.72 frames. ], batch size: 47, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:49:50,078 INFO [train.py:904] (0/8) Epoch 19, batch 2650, loss[loss=0.1853, simple_loss=0.287, pruned_loss=0.04186, over 17021.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2616, pruned_loss=0.04332, over 3318546.94 frames. ], batch size: 50, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:50:03,486 INFO [optim.py:368] (0/8) 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 20:50:58,259 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7683, 3.6628, 4.2165, 1.9936, 4.5095, 4.5337, 3.1013, 3.5018], device='cuda:0'), covar=tensor([0.0785, 0.0293, 0.0219, 0.1258, 0.0072, 0.0176, 0.0495, 0.0380], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0109, 0.0097, 0.0141, 0.0079, 0.0126, 0.0128, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 20:50:59,020 INFO [train.py:904] (0/8) Epoch 19, batch 2700, loss[loss=0.1825, simple_loss=0.2837, pruned_loss=0.04066, over 17071.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2619, pruned_loss=0.04269, over 3320802.88 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:51:19,124 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185416.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 20:51:56,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4655, 5.7818, 5.5438, 5.6049, 5.2115, 5.1446, 5.1837, 5.9444], device='cuda:0'), covar=tensor([0.1253, 0.0979, 0.1155, 0.0879, 0.0902, 0.0850, 0.1391, 0.0985], device='cuda:0'), in_proj_covar=tensor([0.0666, 0.0820, 0.0677, 0.0609, 0.0514, 0.0522, 0.0682, 0.0636], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:52:08,642 INFO [train.py:904] (0/8) Epoch 19, batch 2750, loss[loss=0.1482, simple_loss=0.2385, pruned_loss=0.02897, over 16754.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2622, pruned_loss=0.04274, over 3324181.57 frames. ], batch size: 39, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:52:20,529 INFO [optim.py:368] (0/8) 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 20:52:30,882 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185468.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:52:54,432 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185485.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:08,029 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 20:53:18,344 INFO [train.py:904] (0/8) Epoch 19, batch 2800, loss[loss=0.1722, simple_loss=0.2678, pruned_loss=0.03831, over 17063.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.262, pruned_loss=0.04296, over 3324931.82 frames. ], batch size: 55, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:53:46,667 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185522.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:53:55,703 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185529.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:54:28,294 INFO [train.py:904] (0/8) Epoch 19, batch 2850, loss[loss=0.1516, simple_loss=0.2341, pruned_loss=0.03452, over 16824.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2619, pruned_loss=0.0431, over 3323674.98 frames. ], batch size: 102, lr: 3.59e-03, grad_scale: 8.0 2023-04-30 20:54:41,480 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.535e+02 2.166e+02 2.620e+02 3.152e+02 5.145e+02, threshold=5.240e+02, percent-clipped=1.0 2023-04-30 20:54:53,419 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185570.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:28,820 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185595.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:55:39,065 INFO [train.py:904] (0/8) Epoch 19, batch 2900, loss[loss=0.213, simple_loss=0.2794, pruned_loss=0.07329, over 15633.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2609, pruned_loss=0.04374, over 3323036.00 frames. ], batch size: 191, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:56:08,696 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6612, 4.4642, 4.6318, 4.8333, 4.9610, 4.4095, 4.8576, 4.9375], device='cuda:0'), covar=tensor([0.1610, 0.1449, 0.1682, 0.0923, 0.0733, 0.1200, 0.1425, 0.1065], device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0799, 0.0938, 0.0814, 0.0607, 0.0640, 0.0655, 0.0764], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 20:56:36,418 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185643.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:56:47,752 INFO [train.py:904] (0/8) Epoch 19, batch 2950, loss[loss=0.1682, simple_loss=0.2659, pruned_loss=0.03526, over 17054.00 frames. ], tot_loss[loss=0.174, simple_loss=0.26, pruned_loss=0.04403, over 3321108.41 frames. ], batch size: 53, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:00,936 INFO [optim.py:368] (0/8) 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 20:57:58,864 INFO [train.py:904] (0/8) Epoch 19, batch 3000, loss[loss=0.1891, simple_loss=0.2673, pruned_loss=0.05539, over 16879.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2608, pruned_loss=0.04443, over 3313499.42 frames. ], batch size: 109, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:57:58,865 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 20:58:07,638 INFO [train.py:938] (0/8) Epoch 19, validation: loss=0.1362, simple_loss=0.2416, pruned_loss=0.01538, over 944034.00 frames. 2023-04-30 20:58:07,639 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 20:58:12,771 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 20:58:27,599 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185716.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 20:59:16,288 INFO [train.py:904] (0/8) Epoch 19, batch 3050, loss[loss=0.154, simple_loss=0.2454, pruned_loss=0.03136, over 16806.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2594, pruned_loss=0.04409, over 3320345.65 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 20:59:28,120 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185759.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 20:59:29,999 INFO [optim.py:368] (0/8) 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 20:59:35,054 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185764.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:03,104 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9589, 4.2338, 3.0899, 2.3831, 2.7153, 2.5162, 4.6735, 3.5990], device='cuda:0'), covar=tensor([0.2582, 0.0630, 0.1749, 0.2848, 0.2860, 0.2023, 0.0362, 0.1374], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0265, 0.0300, 0.0303, 0.0293, 0.0251, 0.0288, 0.0331], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 21:00:03,995 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=185785.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:00:23,228 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=185799.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:00:26,861 INFO [train.py:904] (0/8) Epoch 19, batch 3100, loss[loss=0.1878, simple_loss=0.2598, pruned_loss=0.05789, over 16345.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2596, pruned_loss=0.04492, over 3317645.08 frames. ], batch size: 165, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:00:51,339 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185820.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 21:00:56,531 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=185824.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:09,060 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=185833.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:01:34,500 INFO [train.py:904] (0/8) Epoch 19, batch 3150, loss[loss=0.1963, simple_loss=0.2781, pruned_loss=0.05729, over 15438.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2584, pruned_loss=0.0439, over 3314020.82 frames. ], batch size: 190, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:01:46,039 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=185860.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:01:47,107 INFO [optim.py:368] (0/8) 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 21:02:32,166 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8053, 4.3486, 3.0171, 2.1985, 2.6705, 2.4478, 4.6575, 3.6230], device='cuda:0'), covar=tensor([0.2804, 0.0432, 0.1723, 0.3003, 0.3103, 0.2116, 0.0306, 0.1262], device='cuda:0'), in_proj_covar=tensor([0.0322, 0.0267, 0.0302, 0.0306, 0.0296, 0.0253, 0.0290, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 21:02:40,135 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8533, 4.9937, 5.3366, 5.3445, 5.4093, 5.0732, 4.8925, 4.8036], device='cuda:0'), covar=tensor([0.0497, 0.0644, 0.0541, 0.0588, 0.0628, 0.0524, 0.1232, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0408, 0.0448, 0.0433, 0.0407, 0.0482, 0.0458, 0.0554, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 21:02:43,074 INFO [train.py:904] (0/8) Epoch 19, batch 3200, loss[loss=0.2009, simple_loss=0.275, pruned_loss=0.06341, over 16870.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2572, pruned_loss=0.04313, over 3318036.41 frames. ], batch size: 109, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:03:53,545 INFO [train.py:904] (0/8) Epoch 19, batch 3250, loss[loss=0.1413, simple_loss=0.2251, pruned_loss=0.02873, over 16810.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2565, pruned_loss=0.04285, over 3324785.04 frames. ], batch size: 39, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:04:06,316 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.264e+02 2.586e+02 3.072e+02 8.632e+02, threshold=5.173e+02, percent-clipped=3.0 2023-04-30 21:04:09,189 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2182, 4.1986, 4.1342, 3.8871, 3.9154, 4.2228, 3.9062, 3.9799], device='cuda:0'), covar=tensor([0.0644, 0.0637, 0.0305, 0.0267, 0.0710, 0.0454, 0.0813, 0.0616], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0428, 0.0352, 0.0343, 0.0362, 0.0398, 0.0241, 0.0418], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 21:05:01,616 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-186000.pt 2023-04-30 21:05:07,485 INFO [train.py:904] (0/8) Epoch 19, batch 3300, loss[loss=0.1944, simple_loss=0.275, pruned_loss=0.05695, over 16719.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2576, pruned_loss=0.04333, over 3329938.29 frames. ], batch size: 83, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:06:08,413 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5918, 4.5629, 4.4801, 4.2092, 4.2298, 4.5694, 4.3032, 4.2819], device='cuda:0'), covar=tensor([0.0604, 0.0613, 0.0290, 0.0269, 0.0722, 0.0443, 0.0574, 0.0605], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0427, 0.0351, 0.0342, 0.0361, 0.0397, 0.0240, 0.0417], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 21:06:12,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5464, 4.5947, 4.9299, 4.9485, 4.9620, 4.6185, 4.6197, 4.5032], device='cuda:0'), covar=tensor([0.0366, 0.0581, 0.0392, 0.0344, 0.0447, 0.0415, 0.0814, 0.0507], device='cuda:0'), in_proj_covar=tensor([0.0409, 0.0449, 0.0435, 0.0407, 0.0481, 0.0459, 0.0554, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 21:06:15,819 INFO [train.py:904] (0/8) Epoch 19, batch 3350, loss[loss=0.198, simple_loss=0.2712, pruned_loss=0.06238, over 16716.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2589, pruned_loss=0.04358, over 3334225.16 frames. ], batch size: 134, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:06:20,395 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2482, 3.6682, 3.6748, 2.2674, 3.1452, 2.5723, 3.7190, 3.8642], device='cuda:0'), covar=tensor([0.0290, 0.0735, 0.0529, 0.1792, 0.0788, 0.0869, 0.0542, 0.0876], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0162, 0.0166, 0.0152, 0.0144, 0.0128, 0.0144, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 21:06:28,002 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.170e+02 2.521e+02 2.996e+02 5.638e+02, threshold=5.041e+02, percent-clipped=4.0 2023-04-30 21:07:23,760 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186100.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:07:25,699 INFO [train.py:904] (0/8) Epoch 19, batch 3400, loss[loss=0.1867, simple_loss=0.2557, pruned_loss=0.05882, over 16664.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2588, pruned_loss=0.04375, over 3328152.86 frames. ], batch size: 134, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:07:44,050 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186115.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:07:57,604 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186124.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:34,735 INFO [train.py:904] (0/8) Epoch 19, batch 3450, loss[loss=0.1483, simple_loss=0.2397, pruned_loss=0.02849, over 17236.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2582, pruned_loss=0.043, over 3325705.97 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 8.0 2023-04-30 21:08:38,537 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186155.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:08:43,408 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186158.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:47,861 INFO [optim.py:368] (0/8) 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 21:08:48,270 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186161.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:08:48,711 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-04-30 21:09:02,956 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186172.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:09:32,074 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3012, 4.2721, 4.2341, 3.7080, 4.2556, 1.7428, 4.0246, 3.8076], device='cuda:0'), covar=tensor([0.0142, 0.0123, 0.0175, 0.0278, 0.0096, 0.2683, 0.0145, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0150, 0.0195, 0.0179, 0.0173, 0.0205, 0.0188, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:09:45,142 INFO [train.py:904] (0/8) Epoch 19, batch 3500, loss[loss=0.1574, simple_loss=0.2464, pruned_loss=0.03418, over 17223.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2565, pruned_loss=0.04276, over 3320142.94 frames. ], batch size: 45, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:10:08,287 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:10:55,187 INFO [train.py:904] (0/8) Epoch 19, batch 3550, loss[loss=0.1644, simple_loss=0.2566, pruned_loss=0.03615, over 17105.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2558, pruned_loss=0.04246, over 3314629.65 frames. ], batch size: 47, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:11:04,595 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186259.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:11:04,899 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-04-30 21:11:06,983 INFO [optim.py:368] (0/8) 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 21:11:29,181 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8266, 2.7010, 2.8370, 2.1141, 2.6748, 2.1500, 2.7175, 2.9081], device='cuda:0'), covar=tensor([0.0271, 0.0814, 0.0428, 0.1814, 0.0784, 0.0905, 0.0565, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0162, 0.0166, 0.0152, 0.0144, 0.0128, 0.0143, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 21:12:03,398 INFO [train.py:904] (0/8) Epoch 19, batch 3600, loss[loss=0.1724, simple_loss=0.2674, pruned_loss=0.03869, over 16700.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2545, pruned_loss=0.04189, over 3319713.75 frames. ], batch size: 57, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:12:28,420 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186320.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:12:34,759 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 21:13:13,928 INFO [train.py:904] (0/8) Epoch 19, batch 3650, loss[loss=0.1577, simple_loss=0.2516, pruned_loss=0.03194, over 17131.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2539, pruned_loss=0.04264, over 3309913.32 frames. ], batch size: 46, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:13:27,665 INFO [optim.py:368] (0/8) 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 21:14:28,525 INFO [train.py:904] (0/8) Epoch 19, batch 3700, loss[loss=0.1718, simple_loss=0.2504, pruned_loss=0.04656, over 16707.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2522, pruned_loss=0.04388, over 3283126.73 frames. ], batch size: 124, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:14:48,801 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186415.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:14:54,566 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8399, 5.0924, 5.2061, 5.0375, 5.1186, 5.6394, 5.1495, 4.8574], device='cuda:0'), covar=tensor([0.1281, 0.1982, 0.1778, 0.1932, 0.2236, 0.0915, 0.1469, 0.2301], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0587, 0.0649, 0.0491, 0.0656, 0.0686, 0.0508, 0.0657], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 21:14:57,083 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186421.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:15:41,470 INFO [train.py:904] (0/8) Epoch 19, batch 3750, loss[loss=0.1792, simple_loss=0.2488, pruned_loss=0.05483, over 16759.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2527, pruned_loss=0.04505, over 3260478.69 frames. ], batch size: 124, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:15:45,890 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186455.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:15:47,766 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186456.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:15:55,029 INFO [optim.py:368] (0/8) 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 21:15:57,664 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186463.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:15,879 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186475.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:17,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0599, 2.9124, 2.7131, 4.5958, 3.6989, 4.2879, 1.7083, 3.1077], device='cuda:0'), covar=tensor([0.1217, 0.0707, 0.1065, 0.0180, 0.0308, 0.0354, 0.1483, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0172, 0.0192, 0.0185, 0.0205, 0.0216, 0.0197, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 21:16:26,549 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186482.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:16:53,855 INFO [train.py:904] (0/8) Epoch 19, batch 3800, loss[loss=0.1875, simple_loss=0.2596, pruned_loss=0.05776, over 16374.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2538, pruned_loss=0.04652, over 3264556.19 frames. ], batch size: 146, lr: 3.58e-03, grad_scale: 16.0 2023-04-30 21:16:55,940 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186503.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:17:07,276 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-04-30 21:17:11,602 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186514.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:15,548 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186517.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:42,904 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186536.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:17:47,953 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186539.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:05,825 INFO [train.py:904] (0/8) Epoch 19, batch 3850, loss[loss=0.1647, simple_loss=0.2463, pruned_loss=0.04154, over 16892.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2539, pruned_loss=0.04715, over 3259782.99 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:18:16,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1536, 3.9448, 4.5651, 2.6244, 4.8858, 4.8255, 3.3736, 3.8861], device='cuda:0'), covar=tensor([0.0647, 0.0257, 0.0144, 0.0952, 0.0036, 0.0072, 0.0366, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0107, 0.0097, 0.0138, 0.0078, 0.0125, 0.0127, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 21:18:22,161 INFO [optim.py:368] (0/8) 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 21:18:28,286 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:45,006 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186578.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:18:56,106 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186586.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:06,987 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 21:19:16,674 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 19, batch 3900, loss[loss=0.1599, simple_loss=0.2413, pruned_loss=0.0393, over 16847.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2535, pruned_loss=0.04727, over 3258339.34 frames. ], batch size: 96, lr: 3.58e-03, grad_scale: 4.0 2023-04-30 21:19:36,756 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:19:55,034 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186628.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:20,742 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186647.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:20:28,334 INFO [train.py:904] (0/8) Epoch 19, batch 3950, loss[loss=0.1791, simple_loss=0.2586, pruned_loss=0.0498, over 16256.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2539, pruned_loss=0.04811, over 3253869.27 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 4.0 2023-04-30 21:20:42,974 INFO [optim.py:368] (0/8) 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 21:21:13,171 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.47 vs. limit=5.0 2023-04-30 21:21:39,186 INFO [train.py:904] (0/8) Epoch 19, batch 4000, loss[loss=0.1835, simple_loss=0.2528, pruned_loss=0.05715, over 16859.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.254, pruned_loss=0.04829, over 3263664.23 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:21:48,996 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186709.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:22:45,693 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0906, 2.2061, 2.3042, 3.7190, 2.1401, 2.5107, 2.3010, 2.3369], device='cuda:0'), covar=tensor([0.1336, 0.3343, 0.2780, 0.0618, 0.3894, 0.2391, 0.3473, 0.3106], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0441, 0.0362, 0.0327, 0.0433, 0.0509, 0.0410, 0.0516], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:22:49,924 INFO [train.py:904] (0/8) Epoch 19, batch 4050, loss[loss=0.1703, simple_loss=0.2548, pruned_loss=0.04294, over 17187.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2546, pruned_loss=0.04744, over 3277289.69 frames. ], batch size: 46, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:22:57,387 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186756.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:06,014 INFO [optim.py:368] (0/8) 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 21:23:16,281 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186770.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:23:26,470 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186777.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:03,616 INFO [train.py:904] (0/8) Epoch 19, batch 4100, loss[loss=0.1668, simple_loss=0.2547, pruned_loss=0.03948, over 16832.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2557, pruned_loss=0.04676, over 3266430.52 frames. ], batch size: 42, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:24:06,500 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186804.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:21,814 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186814.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:46,682 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186831.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:24:59,827 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186840.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:17,901 INFO [train.py:904] (0/8) Epoch 19, batch 4150, loss[loss=0.1881, simple_loss=0.2795, pruned_loss=0.04832, over 17227.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2634, pruned_loss=0.04925, over 3240073.42 frames. ], batch size: 45, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:25:33,652 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186862.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:25:34,925 INFO [optim.py:368] (0/8) 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 21:25:49,714 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186873.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:22,980 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186895.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:31,948 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186901.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:26:32,672 INFO [train.py:904] (0/8) Epoch 19, batch 4200, loss[loss=0.1903, simple_loss=0.2846, pruned_loss=0.04803, over 16662.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2701, pruned_loss=0.05078, over 3213412.26 frames. ], batch size: 89, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:26:47,130 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=186911.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:26:53,374 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=186915.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:06,341 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186923.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:34,747 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=186942.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:27:49,662 INFO [train.py:904] (0/8) Epoch 19, batch 4250, loss[loss=0.1854, simple_loss=0.2836, pruned_loss=0.04366, over 16201.00 frames. ], tot_loss[loss=0.1877, simple_loss=0.2739, pruned_loss=0.05075, over 3198516.42 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:27:50,209 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0715, 4.1028, 3.2231, 2.5438, 3.0637, 2.8135, 4.6050, 3.8201], device='cuda:0'), covar=tensor([0.2487, 0.0705, 0.1658, 0.2593, 0.2456, 0.1730, 0.0498, 0.1066], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0265, 0.0299, 0.0303, 0.0296, 0.0251, 0.0288, 0.0329], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 21:28:05,623 INFO [optim.py:368] (0/8) 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 21:28:05,996 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=186963.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:28:06,086 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4190, 5.7216, 5.4599, 5.5280, 5.1928, 5.0790, 5.1511, 5.7966], device='cuda:0'), covar=tensor([0.0997, 0.0778, 0.0925, 0.0716, 0.0783, 0.0695, 0.1014, 0.0855], device='cuda:0'), in_proj_covar=tensor([0.0654, 0.0805, 0.0663, 0.0602, 0.0506, 0.0515, 0.0673, 0.0619], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:28:09,926 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7734, 4.9518, 4.6725, 4.3946, 4.1198, 4.8397, 4.6836, 4.3850], device='cuda:0'), covar=tensor([0.0642, 0.0524, 0.0399, 0.0368, 0.1230, 0.0453, 0.0382, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0290, 0.0413, 0.0341, 0.0331, 0.0351, 0.0383, 0.0232, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:28:19,150 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=186972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:02,917 INFO [train.py:904] (0/8) Epoch 19, batch 4300, loss[loss=0.2027, simple_loss=0.2886, pruned_loss=0.05842, over 16753.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2752, pruned_loss=0.05062, over 3183757.70 frames. ], batch size: 62, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:29:37,028 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187025.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:29:56,618 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187038.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:16,628 INFO [train.py:904] (0/8) Epoch 19, batch 4350, loss[loss=0.224, simple_loss=0.2946, pruned_loss=0.07668, over 11659.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2783, pruned_loss=0.05149, over 3175960.46 frames. ], batch size: 248, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:30:32,932 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.681e+02 2.200e+02 2.562e+02 2.922e+02 5.186e+02, threshold=5.123e+02, percent-clipped=1.0 2023-04-30 21:30:35,929 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187065.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:30:46,871 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-04-30 21:30:53,873 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187077.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:07,455 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187086.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:31:26,240 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187099.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:31:31,220 INFO [train.py:904] (0/8) Epoch 19, batch 4400, loss[loss=0.1878, simple_loss=0.275, pruned_loss=0.05026, over 16705.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2804, pruned_loss=0.05265, over 3170813.64 frames. ], batch size: 134, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:32:04,911 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:13,532 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187131.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:32:41,916 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-04-30 21:32:43,362 INFO [train.py:904] (0/8) Epoch 19, batch 4450, loss[loss=0.1959, simple_loss=0.2773, pruned_loss=0.05722, over 11705.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2834, pruned_loss=0.05379, over 3163317.36 frames. ], batch size: 247, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:32:54,236 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3566, 4.4537, 4.2468, 3.9783, 3.9458, 4.3645, 4.0231, 4.0649], device='cuda:0'), covar=tensor([0.0499, 0.0263, 0.0225, 0.0217, 0.0681, 0.0306, 0.0634, 0.0510], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0405, 0.0336, 0.0325, 0.0346, 0.0377, 0.0230, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:33:00,622 INFO [optim.py:368] (0/8) 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 21:33:14,969 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187173.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:24,161 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:47,383 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187195.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:33:49,268 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187196.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 21:33:53,667 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7996, 4.8670, 5.1303, 5.1092, 5.1496, 4.8105, 4.7887, 4.5732], device='cuda:0'), covar=tensor([0.0254, 0.0330, 0.0281, 0.0315, 0.0344, 0.0311, 0.0796, 0.0399], device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0429, 0.0418, 0.0390, 0.0459, 0.0440, 0.0531, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 21:33:57,520 INFO [train.py:904] (0/8) Epoch 19, batch 4500, loss[loss=0.2141, simple_loss=0.2969, pruned_loss=0.06562, over 16912.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2841, pruned_loss=0.05437, over 3186041.68 frames. ], batch size: 109, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:34:21,062 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 21:34:24,640 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:27,214 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:38,086 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4765, 2.5136, 2.0212, 2.2421, 2.9214, 2.5555, 3.1055, 3.1270], device='cuda:0'), covar=tensor([0.0086, 0.0369, 0.0501, 0.0431, 0.0219, 0.0329, 0.0166, 0.0206], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0227, 0.0218, 0.0220, 0.0229, 0.0228, 0.0232, 0.0223], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:34:54,254 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187242.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:34:55,417 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187243.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:08,497 INFO [train.py:904] (0/8) Epoch 19, batch 4550, loss[loss=0.217, simple_loss=0.3046, pruned_loss=0.06466, over 17036.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2853, pruned_loss=0.05556, over 3198125.48 frames. ], batch size: 50, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:35:24,437 INFO [optim.py:368] (0/8) 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 21:35:30,229 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187267.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:35:36,704 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187271.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:36:04,449 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187290.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:36:20,311 INFO [train.py:904] (0/8) Epoch 19, batch 4600, loss[loss=0.1892, simple_loss=0.2768, pruned_loss=0.05086, over 16863.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2861, pruned_loss=0.05562, over 3219758.06 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:33,044 INFO [train.py:904] (0/8) Epoch 19, batch 4650, loss[loss=0.223, simple_loss=0.2995, pruned_loss=0.0733, over 11716.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2852, pruned_loss=0.05569, over 3211198.26 frames. ], batch size: 246, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:37:49,313 INFO [optim.py:368] (0/8) 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 21:37:52,250 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:01,517 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187372.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:05,242 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4335, 3.4309, 1.9300, 3.9125, 2.5931, 3.9094, 2.2081, 2.7598], device='cuda:0'), covar=tensor([0.0272, 0.0382, 0.1834, 0.0140, 0.0944, 0.0398, 0.1513, 0.0799], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0156, 0.0175, 0.0215, 0.0198, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 21:38:14,870 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187381.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 21:38:33,241 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187394.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:38:44,858 INFO [train.py:904] (0/8) Epoch 19, batch 4700, loss[loss=0.1733, simple_loss=0.2642, pruned_loss=0.04124, over 16848.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2815, pruned_loss=0.05402, over 3218001.33 frames. ], batch size: 116, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:39:01,717 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187413.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:30,042 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187433.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:39:55,445 INFO [train.py:904] (0/8) Epoch 19, batch 4750, loss[loss=0.1661, simple_loss=0.25, pruned_loss=0.04116, over 16462.00 frames. ], tot_loss[loss=0.19, simple_loss=0.277, pruned_loss=0.05146, over 3216928.06 frames. ], batch size: 75, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:40:10,496 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9689, 3.1927, 3.3724, 1.9233, 2.7401, 2.1156, 3.4324, 3.3737], device='cuda:0'), covar=tensor([0.0271, 0.0801, 0.0603, 0.2049, 0.0937, 0.0983, 0.0610, 0.1013], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0160, 0.0164, 0.0150, 0.0142, 0.0127, 0.0141, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 21:40:11,094 INFO [optim.py:368] (0/8) 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 21:40:57,717 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187496.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:00,704 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3409, 3.2952, 3.5071, 1.7382, 3.7233, 3.7559, 2.9543, 2.6678], device='cuda:0'), covar=tensor([0.0883, 0.0260, 0.0196, 0.1292, 0.0076, 0.0148, 0.0416, 0.0553], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0107, 0.0096, 0.0138, 0.0077, 0.0123, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 21:41:05,660 INFO [train.py:904] (0/8) Epoch 19, batch 4800, loss[loss=0.208, simple_loss=0.2866, pruned_loss=0.06468, over 11614.00 frames. ], tot_loss[loss=0.1866, simple_loss=0.2738, pruned_loss=0.04969, over 3197355.77 frames. ], batch size: 248, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:41:20,843 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187512.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:30,384 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187519.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:37,688 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=187523.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:41:57,675 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2442, 3.5300, 3.5942, 2.0227, 3.0254, 2.4688, 3.6511, 3.5996], device='cuda:0'), covar=tensor([0.0258, 0.0704, 0.0556, 0.1970, 0.0801, 0.0871, 0.0569, 0.0948], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0160, 0.0165, 0.0150, 0.0142, 0.0127, 0.0142, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 21:42:08,456 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187544.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:19,534 INFO [train.py:904] (0/8) Epoch 19, batch 4850, loss[loss=0.1998, simple_loss=0.2902, pruned_loss=0.05469, over 16747.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2743, pruned_loss=0.04879, over 3200910.50 frames. ], batch size: 124, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:42:30,844 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0416, 2.0645, 1.7893, 1.7298, 2.2938, 1.9489, 1.9535, 2.3835], device='cuda:0'), covar=tensor([0.0163, 0.0325, 0.0439, 0.0436, 0.0203, 0.0326, 0.0161, 0.0220], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0227, 0.0219, 0.0221, 0.0229, 0.0228, 0.0231, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:42:36,213 INFO [optim.py:368] (0/8) 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 21:42:41,908 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187567.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:42:50,150 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187573.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:01,980 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187580.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:07,607 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=187584.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:43:32,171 INFO [train.py:904] (0/8) Epoch 19, batch 4900, loss[loss=0.1794, simple_loss=0.277, pruned_loss=0.04092, over 15248.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2738, pruned_loss=0.04776, over 3199188.31 frames. ], batch size: 190, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:43:51,072 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187615.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:44:40,255 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-04-30 21:44:43,113 INFO [train.py:904] (0/8) Epoch 19, batch 4950, loss[loss=0.1913, simple_loss=0.284, pruned_loss=0.0493, over 16157.00 frames. ], tot_loss[loss=0.1841, simple_loss=0.2736, pruned_loss=0.04732, over 3203402.15 frames. ], batch size: 165, lr: 3.57e-03, grad_scale: 8.0 2023-04-30 21:44:58,342 INFO [optim.py:368] (0/8) 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 21:45:24,097 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=187681.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 21:45:44,051 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 19, batch 5000, loss[loss=0.1829, simple_loss=0.2736, pruned_loss=0.0461, over 16631.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2751, pruned_loss=0.04733, over 3197551.72 frames. ], batch size: 62, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:46:16,905 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8565, 4.8922, 4.7782, 3.4579, 4.1123, 4.8052, 4.0045, 2.6436], device='cuda:0'), covar=tensor([0.0489, 0.0028, 0.0027, 0.0312, 0.0080, 0.0068, 0.0083, 0.0440], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0078, 0.0079, 0.0130, 0.0093, 0.0103, 0.0090, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 21:46:25,224 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5802, 3.6408, 3.3950, 3.0966, 3.2152, 3.4925, 3.3245, 3.3001], device='cuda:0'), covar=tensor([0.0530, 0.0493, 0.0289, 0.0265, 0.0556, 0.0414, 0.1522, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0402, 0.0331, 0.0322, 0.0341, 0.0375, 0.0227, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:46:32,223 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:33,450 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187729.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:46:52,595 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=187742.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:47:07,090 INFO [train.py:904] (0/8) Epoch 19, batch 5050, loss[loss=0.1964, simple_loss=0.2852, pruned_loss=0.05378, over 15381.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2755, pruned_loss=0.04719, over 3207404.19 frames. ], batch size: 191, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:47:21,708 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.150e+02 2.460e+02 2.787e+02 5.516e+02, threshold=4.919e+02, percent-clipped=2.0 2023-04-30 21:48:02,939 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1146, 5.2250, 5.5099, 5.4121, 5.5761, 5.2245, 5.0643, 4.9269], device='cuda:0'), covar=tensor([0.0380, 0.0531, 0.0379, 0.0553, 0.0533, 0.0425, 0.1211, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0384, 0.0423, 0.0412, 0.0384, 0.0456, 0.0433, 0.0519, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 21:48:17,855 INFO [train.py:904] (0/8) Epoch 19, batch 5100, loss[loss=0.1877, simple_loss=0.2794, pruned_loss=0.04796, over 17004.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.274, pruned_loss=0.04679, over 3204376.13 frames. ], batch size: 53, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:20,259 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-04-30 21:49:30,697 INFO [train.py:904] (0/8) Epoch 19, batch 5150, loss[loss=0.1817, simple_loss=0.2741, pruned_loss=0.04459, over 16999.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2736, pruned_loss=0.04644, over 3186635.55 frames. ], batch size: 55, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:49:47,461 INFO [optim.py:368] (0/8) 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 21:49:54,404 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187868.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:04,813 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187875.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:11,341 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=187879.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:50:41,949 INFO [train.py:904] (0/8) Epoch 19, batch 5200, loss[loss=0.1639, simple_loss=0.2472, pruned_loss=0.04033, over 17237.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2722, pruned_loss=0.04582, over 3199947.28 frames. ], batch size: 52, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:51:53,337 INFO [train.py:904] (0/8) Epoch 19, batch 5250, loss[loss=0.1776, simple_loss=0.2703, pruned_loss=0.04246, over 15290.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.27, pruned_loss=0.04565, over 3197818.28 frames. ], batch size: 190, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:52:04,071 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2983, 4.1914, 4.3726, 4.5324, 4.7092, 4.2848, 4.6462, 4.7384], device='cuda:0'), covar=tensor([0.1879, 0.1265, 0.1571, 0.0753, 0.0493, 0.1098, 0.0636, 0.0578], device='cuda:0'), in_proj_covar=tensor([0.0616, 0.0759, 0.0892, 0.0779, 0.0580, 0.0607, 0.0622, 0.0727], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:52:08,314 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.924e+02 2.321e+02 2.643e+02 4.137e+02, threshold=4.643e+02, percent-clipped=0.0 2023-04-30 21:52:51,950 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-04-30 21:53:02,255 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-04-30 21:53:03,082 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-188000.pt 2023-04-30 21:53:08,608 INFO [train.py:904] (0/8) Epoch 19, batch 5300, loss[loss=0.1751, simple_loss=0.2576, pruned_loss=0.04632, over 12417.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2666, pruned_loss=0.04458, over 3197284.82 frames. ], batch size: 246, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:53:17,980 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 21:53:46,197 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188028.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:54:21,274 INFO [train.py:904] (0/8) Epoch 19, batch 5350, loss[loss=0.176, simple_loss=0.27, pruned_loss=0.04101, over 16865.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2646, pruned_loss=0.0437, over 3206976.15 frames. ], batch size: 96, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:54:25,950 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5159, 5.5028, 5.3293, 4.6004, 5.4241, 2.1770, 5.1365, 5.1197], device='cuda:0'), covar=tensor([0.0062, 0.0063, 0.0141, 0.0412, 0.0071, 0.2468, 0.0094, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0144, 0.0188, 0.0173, 0.0165, 0.0199, 0.0180, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:54:38,015 INFO [optim.py:368] (0/8) 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 21:54:57,076 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188076.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:55:35,754 INFO [train.py:904] (0/8) Epoch 19, batch 5400, loss[loss=0.191, simple_loss=0.2842, pruned_loss=0.0489, over 16804.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2677, pruned_loss=0.04444, over 3218250.92 frames. ], batch size: 124, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:55:58,525 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5394, 3.5972, 2.0321, 4.1200, 2.7281, 4.0748, 2.3230, 2.9306], device='cuda:0'), covar=tensor([0.0265, 0.0358, 0.1730, 0.0160, 0.0845, 0.0484, 0.1473, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0175, 0.0194, 0.0155, 0.0175, 0.0213, 0.0200, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 21:55:59,783 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1802, 2.1663, 2.2111, 3.7818, 2.1276, 2.6403, 2.3035, 2.3622], device='cuda:0'), covar=tensor([0.1284, 0.3484, 0.2818, 0.0517, 0.3994, 0.2304, 0.3385, 0.2977], device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0432, 0.0357, 0.0320, 0.0427, 0.0499, 0.0403, 0.0503], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 21:56:37,884 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 21:56:54,052 INFO [train.py:904] (0/8) Epoch 19, batch 5450, loss[loss=0.1855, simple_loss=0.2775, pruned_loss=0.04668, over 16826.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2712, pruned_loss=0.04567, over 3241763.89 frames. ], batch size: 102, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:57:11,922 INFO [optim.py:368] (0/8) 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 21:57:19,814 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188168.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:32,208 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188175.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:57:37,820 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188179.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:14,507 INFO [train.py:904] (0/8) Epoch 19, batch 5500, loss[loss=0.2487, simple_loss=0.3346, pruned_loss=0.08139, over 16254.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2782, pruned_loss=0.04957, over 3214022.52 frames. ], batch size: 165, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:58:37,954 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:48,996 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188223.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:58:55,241 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188227.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 21:59:35,425 INFO [train.py:904] (0/8) Epoch 19, batch 5550, loss[loss=0.2083, simple_loss=0.2944, pruned_loss=0.06114, over 16694.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2846, pruned_loss=0.05434, over 3180492.56 frames. ], batch size: 124, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 21:59:53,553 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.484e+02 3.263e+02 3.931e+02 4.573e+02 1.048e+03, threshold=7.861e+02, percent-clipped=6.0 2023-04-30 22:00:26,007 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188283.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:00:55,105 INFO [train.py:904] (0/8) Epoch 19, batch 5600, loss[loss=0.2137, simple_loss=0.2988, pruned_loss=0.06429, over 17023.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2894, pruned_loss=0.05842, over 3149367.16 frames. ], batch size: 55, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:01:12,987 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1951, 4.1884, 4.1104, 3.3336, 4.1617, 1.7361, 3.9463, 3.7188], device='cuda:0'), covar=tensor([0.0127, 0.0107, 0.0190, 0.0344, 0.0106, 0.2805, 0.0141, 0.0247], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0144, 0.0190, 0.0175, 0.0166, 0.0199, 0.0180, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:02:07,222 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:02:15,411 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 22:02:19,471 INFO [train.py:904] (0/8) Epoch 19, batch 5650, loss[loss=0.2202, simple_loss=0.3021, pruned_loss=0.06914, over 16392.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2943, pruned_loss=0.06241, over 3121464.56 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:02:36,991 INFO [optim.py:368] (0/8) 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 22:03:36,660 INFO [train.py:904] (0/8) Epoch 19, batch 5700, loss[loss=0.2125, simple_loss=0.3032, pruned_loss=0.06097, over 16360.00 frames. ], tot_loss[loss=0.212, simple_loss=0.2961, pruned_loss=0.06398, over 3111864.16 frames. ], batch size: 146, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:03:40,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3951, 5.6709, 5.3927, 5.4889, 5.1257, 4.9535, 5.1261, 5.7983], device='cuda:0'), covar=tensor([0.1193, 0.0825, 0.1208, 0.0944, 0.0880, 0.0831, 0.1127, 0.0862], device='cuda:0'), in_proj_covar=tensor([0.0634, 0.0782, 0.0643, 0.0581, 0.0490, 0.0497, 0.0646, 0.0601], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:04:55,862 INFO [train.py:904] (0/8) Epoch 19, batch 5750, loss[loss=0.2218, simple_loss=0.3037, pruned_loss=0.06991, over 15130.00 frames. ], tot_loss[loss=0.2158, simple_loss=0.2993, pruned_loss=0.06614, over 3067586.18 frames. ], batch size: 190, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:05:12,679 INFO [optim.py:368] (0/8) 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 22:05:36,489 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-04-30 22:05:45,302 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188481.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:06:19,322 INFO [train.py:904] (0/8) Epoch 19, batch 5800, loss[loss=0.2007, simple_loss=0.2841, pruned_loss=0.05862, over 16619.00 frames. ], tot_loss[loss=0.2141, simple_loss=0.2985, pruned_loss=0.06485, over 3072167.98 frames. ], batch size: 62, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:24,030 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188542.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:07:38,959 INFO [train.py:904] (0/8) Epoch 19, batch 5850, loss[loss=0.2179, simple_loss=0.3045, pruned_loss=0.06565, over 16930.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2963, pruned_loss=0.06354, over 3064864.63 frames. ], batch size: 109, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:07:46,736 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0490, 2.1387, 2.1766, 3.6502, 2.0576, 2.4876, 2.2255, 2.2851], device='cuda:0'), covar=tensor([0.1352, 0.3336, 0.2815, 0.0555, 0.3996, 0.2372, 0.3409, 0.3264], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0433, 0.0356, 0.0319, 0.0428, 0.0498, 0.0402, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:07:57,316 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.681e+02 3.202e+02 3.753e+02 8.228e+02, threshold=6.404e+02, percent-clipped=1.0 2023-04-30 22:08:59,602 INFO [train.py:904] (0/8) Epoch 19, batch 5900, loss[loss=0.1833, simple_loss=0.273, pruned_loss=0.04683, over 17155.00 frames. ], tot_loss[loss=0.2107, simple_loss=0.2959, pruned_loss=0.06272, over 3095090.49 frames. ], batch size: 46, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:10:01,861 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188639.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:10:21,859 INFO [train.py:904] (0/8) Epoch 19, batch 5950, loss[loss=0.21, simple_loss=0.2926, pruned_loss=0.06373, over 16571.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2964, pruned_loss=0.0609, over 3116041.99 frames. ], batch size: 62, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:10:40,537 INFO [optim.py:368] (0/8) 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 22:11:41,794 INFO [train.py:904] (0/8) Epoch 19, batch 6000, loss[loss=0.1967, simple_loss=0.2833, pruned_loss=0.05501, over 16727.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2953, pruned_loss=0.06062, over 3111913.67 frames. ], batch size: 124, lr: 3.56e-03, grad_scale: 8.0 2023-04-30 22:11:41,796 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 22:11:52,561 INFO [train.py:938] (0/8) Epoch 19, validation: loss=0.1523, simple_loss=0.2653, pruned_loss=0.01968, over 944034.00 frames. 2023-04-30 22:11:52,562 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 22:11:52,975 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2593, 4.3474, 4.4708, 4.2915, 4.3834, 4.8365, 4.4135, 4.1492], device='cuda:0'), covar=tensor([0.1734, 0.1930, 0.2381, 0.2103, 0.2251, 0.1029, 0.1577, 0.2485], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0574, 0.0633, 0.0482, 0.0639, 0.0666, 0.0494, 0.0644], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 22:12:14,120 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5065, 4.6657, 4.7672, 4.5604, 4.6489, 5.1570, 4.6700, 4.4101], device='cuda:0'), covar=tensor([0.1516, 0.1923, 0.2328, 0.2241, 0.2599, 0.1073, 0.1658, 0.2518], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0573, 0.0632, 0.0482, 0.0639, 0.0665, 0.0494, 0.0644], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 22:13:10,898 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0014, 4.1905, 2.4972, 4.8779, 3.2114, 4.7234, 2.5411, 3.2097], device='cuda:0'), covar=tensor([0.0264, 0.0290, 0.1673, 0.0170, 0.0745, 0.0448, 0.1716, 0.0771], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0175, 0.0194, 0.0156, 0.0175, 0.0215, 0.0202, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 22:13:13,868 INFO [train.py:904] (0/8) Epoch 19, batch 6050, loss[loss=0.179, simple_loss=0.2778, pruned_loss=0.04007, over 16752.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2939, pruned_loss=0.05974, over 3128314.69 frames. ], batch size: 83, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:13:23,375 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-04-30 22:13:33,109 INFO [optim.py:368] (0/8) 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 22:13:53,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6932, 1.8495, 1.5796, 1.5959, 1.9923, 1.6748, 1.7202, 1.9662], device='cuda:0'), covar=tensor([0.0165, 0.0248, 0.0365, 0.0308, 0.0187, 0.0260, 0.0163, 0.0187], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0228, 0.0222, 0.0222, 0.0230, 0.0229, 0.0230, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:14:18,971 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-04-30 22:14:30,807 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 22:14:32,941 INFO [train.py:904] (0/8) Epoch 19, batch 6100, loss[loss=0.257, simple_loss=0.319, pruned_loss=0.09753, over 11884.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.294, pruned_loss=0.05951, over 3115316.75 frames. ], batch size: 248, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:15:07,093 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-04-30 22:15:33,788 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=188837.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:15:57,524 INFO [train.py:904] (0/8) Epoch 19, batch 6150, loss[loss=0.1831, simple_loss=0.2733, pruned_loss=0.04643, over 16576.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2915, pruned_loss=0.05837, over 3123632.08 frames. ], batch size: 68, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:16:16,885 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.888e+02 3.402e+02 4.182e+02 1.052e+03, threshold=6.804e+02, percent-clipped=3.0 2023-04-30 22:16:22,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6010, 4.7203, 4.8655, 4.7173, 4.8072, 5.2776, 4.8096, 4.5592], device='cuda:0'), covar=tensor([0.1241, 0.1967, 0.2096, 0.2053, 0.2461, 0.1020, 0.1706, 0.2636], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0579, 0.0639, 0.0488, 0.0645, 0.0672, 0.0503, 0.0651], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 22:16:30,836 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-04-30 22:17:16,333 INFO [train.py:904] (0/8) Epoch 19, batch 6200, loss[loss=0.1882, simple_loss=0.2761, pruned_loss=0.05012, over 16686.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2903, pruned_loss=0.05884, over 3106641.03 frames. ], batch size: 76, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:17:48,487 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 22:17:58,771 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188929.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:01,735 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=188931.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:15,060 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=188939.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:18:34,123 INFO [train.py:904] (0/8) Epoch 19, batch 6250, loss[loss=0.1848, simple_loss=0.2825, pruned_loss=0.0436, over 16469.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2898, pruned_loss=0.05853, over 3115260.40 frames. ], batch size: 75, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:18:52,890 INFO [optim.py:368] (0/8) 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 22:19:28,071 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=188987.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:32,250 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188990.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:36,045 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=188992.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:19:51,031 INFO [train.py:904] (0/8) Epoch 19, batch 6300, loss[loss=0.2227, simple_loss=0.2965, pruned_loss=0.07442, over 11848.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.2898, pruned_loss=0.05831, over 3114256.64 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:20:35,805 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189030.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:20:48,255 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-04-30 22:21:09,041 INFO [train.py:904] (0/8) Epoch 19, batch 6350, loss[loss=0.2138, simple_loss=0.3003, pruned_loss=0.06365, over 16457.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2905, pruned_loss=0.05938, over 3106130.40 frames. ], batch size: 146, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:21:27,093 INFO [optim.py:368] (0/8) 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 22:21:28,308 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189064.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:21:51,861 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189079.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:10,396 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189091.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:22:26,886 INFO [train.py:904] (0/8) Epoch 19, batch 6400, loss[loss=0.2027, simple_loss=0.2828, pruned_loss=0.06128, over 16882.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.292, pruned_loss=0.06172, over 3078476.81 frames. ], batch size: 109, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:22:31,604 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1684, 3.3795, 3.5705, 2.0017, 3.0236, 2.3592, 3.5702, 3.6638], device='cuda:0'), covar=tensor([0.0252, 0.0764, 0.0571, 0.2098, 0.0821, 0.0936, 0.0589, 0.0906], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0161, 0.0167, 0.0152, 0.0145, 0.0129, 0.0143, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 22:23:02,717 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189125.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:21,436 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189137.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:25,606 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189140.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:23:43,284 INFO [train.py:904] (0/8) Epoch 19, batch 6450, loss[loss=0.1835, simple_loss=0.2798, pruned_loss=0.04363, over 16692.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.2915, pruned_loss=0.06067, over 3076784.72 frames. ], batch size: 89, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:23:58,880 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-04-30 22:24:01,005 INFO [optim.py:368] (0/8) 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 22:24:34,448 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:25:00,522 INFO [train.py:904] (0/8) Epoch 19, batch 6500, loss[loss=0.2172, simple_loss=0.287, pruned_loss=0.07371, over 11769.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2895, pruned_loss=0.0599, over 3081242.43 frames. ], batch size: 247, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:26:02,951 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6499, 2.6150, 2.4736, 3.8242, 2.4376, 3.8863, 1.3628, 2.7419], device='cuda:0'), covar=tensor([0.1477, 0.0781, 0.1194, 0.0171, 0.0200, 0.0429, 0.1869, 0.0872], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0171, 0.0193, 0.0182, 0.0205, 0.0214, 0.0196, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 22:26:14,628 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189248.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:26:19,688 INFO [train.py:904] (0/8) Epoch 19, batch 6550, loss[loss=0.2087, simple_loss=0.3018, pruned_loss=0.05782, over 16757.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2917, pruned_loss=0.06044, over 3082810.16 frames. ], batch size: 124, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:26:37,018 INFO [optim.py:368] (0/8) 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 22:27:09,608 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189285.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:12,019 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189287.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:27:34,969 INFO [train.py:904] (0/8) Epoch 19, batch 6600, loss[loss=0.1952, simple_loss=0.2876, pruned_loss=0.0514, over 16622.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.294, pruned_loss=0.06065, over 3085440.52 frames. ], batch size: 75, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:27:45,656 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189309.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:44,343 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189347.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:28:50,956 INFO [train.py:904] (0/8) Epoch 19, batch 6650, loss[loss=0.246, simple_loss=0.313, pruned_loss=0.08956, over 11233.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2937, pruned_loss=0.06101, over 3090424.00 frames. ], batch size: 246, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:29:08,220 INFO [optim.py:368] (0/8) 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 22:29:42,429 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189386.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:29:48,101 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5880, 1.7863, 2.1734, 2.4082, 2.5032, 2.7441, 1.8926, 2.6947], device='cuda:0'), covar=tensor([0.0200, 0.0493, 0.0321, 0.0355, 0.0297, 0.0207, 0.0503, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0188, 0.0174, 0.0177, 0.0188, 0.0148, 0.0190, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:30:06,174 INFO [train.py:904] (0/8) Epoch 19, batch 6700, loss[loss=0.2176, simple_loss=0.3056, pruned_loss=0.06478, over 16441.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.293, pruned_loss=0.06114, over 3104945.69 frames. ], batch size: 146, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:30:15,381 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:33,783 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189420.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:30:56,649 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189435.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:31:21,926 INFO [train.py:904] (0/8) Epoch 19, batch 6750, loss[loss=0.2033, simple_loss=0.3001, pruned_loss=0.0533, over 16798.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2921, pruned_loss=0.06131, over 3105255.16 frames. ], batch size: 83, lr: 3.55e-03, grad_scale: 4.0 2023-04-30 22:31:42,236 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.911e+02 3.292e+02 3.890e+02 7.012e+02, threshold=6.585e+02, percent-clipped=0.0 2023-04-30 22:32:06,444 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8188, 1.4018, 1.6894, 1.7001, 1.8340, 1.9180, 1.6001, 1.8373], device='cuda:0'), covar=tensor([0.0218, 0.0369, 0.0190, 0.0252, 0.0245, 0.0171, 0.0384, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0187, 0.0174, 0.0177, 0.0188, 0.0147, 0.0190, 0.0141], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:32:38,351 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189501.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:32:39,013 INFO [train.py:904] (0/8) Epoch 19, batch 6800, loss[loss=0.2214, simple_loss=0.303, pruned_loss=0.06989, over 16624.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2919, pruned_loss=0.06128, over 3101503.56 frames. ], batch size: 134, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:33:29,827 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-04-30 22:33:50,685 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8390, 3.8831, 4.4020, 1.9000, 4.5176, 4.6328, 3.4422, 3.1583], device='cuda:0'), covar=tensor([0.0730, 0.0220, 0.0140, 0.1255, 0.0052, 0.0106, 0.0276, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0105, 0.0096, 0.0137, 0.0077, 0.0120, 0.0124, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 22:33:58,211 INFO [train.py:904] (0/8) Epoch 19, batch 6850, loss[loss=0.2334, simple_loss=0.3037, pruned_loss=0.08156, over 11992.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2931, pruned_loss=0.0615, over 3105663.22 frames. ], batch size: 248, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:34:13,600 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189562.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:17,028 INFO [optim.py:368] (0/8) 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 22:34:30,861 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7226, 2.5042, 2.3808, 3.7212, 2.3888, 3.8894, 1.4593, 2.9298], device='cuda:0'), covar=tensor([0.1465, 0.0860, 0.1304, 0.0171, 0.0208, 0.0390, 0.1819, 0.0795], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0172, 0.0194, 0.0183, 0.0206, 0.0214, 0.0197, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 22:34:47,104 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189585.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:34:50,373 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189587.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:35:13,001 INFO [train.py:904] (0/8) Epoch 19, batch 6900, loss[loss=0.242, simple_loss=0.3071, pruned_loss=0.08845, over 11831.00 frames. ], tot_loss[loss=0.209, simple_loss=0.2953, pruned_loss=0.06138, over 3100976.91 frames. ], batch size: 247, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:35:16,725 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189604.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:35:25,056 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3193, 2.3649, 2.2761, 4.1389, 2.2386, 2.6274, 2.3734, 2.4380], device='cuda:0'), covar=tensor([0.1267, 0.3290, 0.2836, 0.0472, 0.3879, 0.2473, 0.3335, 0.3076], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0430, 0.0355, 0.0317, 0.0429, 0.0496, 0.0403, 0.0503], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:36:00,959 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189633.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:01,262 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3741, 2.3400, 2.3582, 4.1136, 2.2453, 2.6690, 2.4048, 2.4526], device='cuda:0'), covar=tensor([0.1147, 0.3266, 0.2701, 0.0442, 0.3923, 0.2283, 0.3297, 0.3158], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0430, 0.0355, 0.0317, 0.0429, 0.0495, 0.0403, 0.0503], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:36:04,231 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189635.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:12,231 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189640.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:36:27,607 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-04-30 22:36:29,246 INFO [train.py:904] (0/8) Epoch 19, batch 6950, loss[loss=0.1955, simple_loss=0.2811, pruned_loss=0.05498, over 16240.00 frames. ], tot_loss[loss=0.2111, simple_loss=0.2966, pruned_loss=0.06277, over 3096246.53 frames. ], batch size: 165, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:36:48,883 INFO [optim.py:368] (0/8) 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 22:36:53,009 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-04-30 22:37:21,298 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189686.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:43,840 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189701.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:44,548 INFO [train.py:904] (0/8) Epoch 19, batch 7000, loss[loss=0.2031, simple_loss=0.2995, pruned_loss=0.05335, over 16595.00 frames. ], tot_loss[loss=0.2104, simple_loss=0.2968, pruned_loss=0.06196, over 3094155.91 frames. ], batch size: 62, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:37:47,787 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189703.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:37:49,111 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189704.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:13,467 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189720.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:25,568 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=189728.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:34,807 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189734.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:38:35,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5772, 2.7105, 2.6353, 4.5803, 3.2197, 4.0877, 1.6730, 2.9762], device='cuda:0'), covar=tensor([0.1661, 0.0914, 0.1310, 0.0226, 0.0458, 0.0526, 0.1840, 0.0950], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0173, 0.0195, 0.0185, 0.0207, 0.0215, 0.0198, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 22:38:36,134 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189735.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:01,658 INFO [train.py:904] (0/8) Epoch 19, batch 7050, loss[loss=0.1974, simple_loss=0.2832, pruned_loss=0.05583, over 15356.00 frames. ], tot_loss[loss=0.2108, simple_loss=0.2978, pruned_loss=0.06187, over 3107276.17 frames. ], batch size: 190, lr: 3.55e-03, grad_scale: 8.0 2023-04-30 22:39:22,080 INFO [optim.py:368] (0/8) 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 22:39:22,591 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189765.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:27,398 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189768.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:39:50,320 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189783.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:00,658 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=189789.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:40:19,989 INFO [train.py:904] (0/8) Epoch 19, batch 7100, loss[loss=0.2018, simple_loss=0.286, pruned_loss=0.05875, over 16928.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2963, pruned_loss=0.06195, over 3086758.84 frames. ], batch size: 109, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:40:32,620 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0331, 5.7309, 5.8553, 5.5660, 5.6520, 6.2171, 5.6570, 5.4663], device='cuda:0'), covar=tensor([0.0873, 0.1683, 0.1911, 0.1790, 0.2194, 0.0894, 0.1540, 0.2251], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0573, 0.0632, 0.0479, 0.0637, 0.0662, 0.0495, 0.0643], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 22:41:38,327 INFO [train.py:904] (0/8) Epoch 19, batch 7150, loss[loss=0.2526, simple_loss=0.3117, pruned_loss=0.09679, over 11293.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2946, pruned_loss=0.06225, over 3061731.48 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 4.0 2023-04-30 22:41:45,505 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 22:42:51,652 INFO [train.py:904] (0/8) Epoch 19, batch 7200, loss[loss=0.1722, simple_loss=0.2568, pruned_loss=0.04374, over 16851.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2919, pruned_loss=0.06019, over 3070926.44 frames. ], batch size: 42, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:42:55,799 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=189904.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:44:12,506 INFO [train.py:904] (0/8) Epoch 19, batch 7250, loss[loss=0.1743, simple_loss=0.2628, pruned_loss=0.04284, over 16782.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2902, pruned_loss=0.05965, over 3056242.47 frames. ], batch size: 83, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:44:12,831 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=189952.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:44:34,220 INFO [optim.py:368] (0/8) 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 22:44:49,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0308, 2.4481, 2.6138, 1.9465, 2.6980, 2.8219, 2.3971, 2.4026], device='cuda:0'), covar=tensor([0.0720, 0.0279, 0.0232, 0.0966, 0.0120, 0.0264, 0.0441, 0.0439], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0107, 0.0097, 0.0139, 0.0078, 0.0122, 0.0127, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 22:45:19,350 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=189996.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:45:25,869 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-190000.pt 2023-04-30 22:45:32,093 INFO [train.py:904] (0/8) Epoch 19, batch 7300, loss[loss=0.2083, simple_loss=0.3104, pruned_loss=0.05312, over 16926.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2901, pruned_loss=0.06025, over 3025304.78 frames. ], batch size: 96, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:45:33,990 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190003.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:45:57,982 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-04-30 22:46:48,532 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190051.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:46:49,314 INFO [train.py:904] (0/8) Epoch 19, batch 7350, loss[loss=0.2188, simple_loss=0.3018, pruned_loss=0.06788, over 16428.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.2915, pruned_loss=0.06137, over 3008280.63 frames. ], batch size: 146, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:47:01,682 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190060.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:47:10,693 INFO [optim.py:368] (0/8) 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 22:47:40,428 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190084.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:48:08,865 INFO [train.py:904] (0/8) Epoch 19, batch 7400, loss[loss=0.2357, simple_loss=0.3092, pruned_loss=0.08108, over 11403.00 frames. ], tot_loss[loss=0.207, simple_loss=0.292, pruned_loss=0.061, over 3044236.06 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:04,908 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-04-30 22:49:09,837 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190139.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 22:49:29,563 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190151.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:49:30,841 INFO [train.py:904] (0/8) Epoch 19, batch 7450, loss[loss=0.2367, simple_loss=0.301, pruned_loss=0.08624, over 11658.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2928, pruned_loss=0.06137, over 3062226.12 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:49:41,005 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190157.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:49:55,760 INFO [optim.py:368] (0/8) 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 22:50:20,469 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1205, 2.9014, 3.1142, 1.8476, 3.2429, 3.2882, 2.6879, 2.5987], device='cuda:0'), covar=tensor([0.0816, 0.0266, 0.0228, 0.1136, 0.0104, 0.0249, 0.0432, 0.0463], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0105, 0.0095, 0.0137, 0.0077, 0.0121, 0.0125, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 22:50:51,720 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190200.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 22:50:54,227 INFO [train.py:904] (0/8) Epoch 19, batch 7500, loss[loss=0.208, simple_loss=0.2917, pruned_loss=0.06217, over 15188.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2925, pruned_loss=0.06054, over 3046939.52 frames. ], batch size: 190, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:50:59,307 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190205.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:51:10,240 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190212.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:52:13,621 INFO [train.py:904] (0/8) Epoch 19, batch 7550, loss[loss=0.2013, simple_loss=0.2868, pruned_loss=0.05796, over 15547.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2918, pruned_loss=0.06084, over 3055851.69 frames. ], batch size: 190, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:52:34,941 INFO [optim.py:368] (0/8) 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 22:53:22,542 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190296.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:53:31,302 INFO [train.py:904] (0/8) Epoch 19, batch 7600, loss[loss=0.2071, simple_loss=0.2841, pruned_loss=0.06511, over 16684.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2906, pruned_loss=0.06027, over 3079694.01 frames. ], batch size: 134, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:53:52,925 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190316.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:02,100 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-04-30 22:54:36,983 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190344.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:54:48,960 INFO [train.py:904] (0/8) Epoch 19, batch 7650, loss[loss=0.2767, simple_loss=0.33, pruned_loss=0.1117, over 11189.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.292, pruned_loss=0.06146, over 3062442.99 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:55:02,296 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190360.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:55:10,293 INFO [optim.py:368] (0/8) 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 22:55:27,574 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190384.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:02,438 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:06,254 INFO [train.py:904] (0/8) Epoch 19, batch 7700, loss[loss=0.2116, simple_loss=0.2924, pruned_loss=0.0654, over 15424.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2919, pruned_loss=0.06163, over 3077699.92 frames. ], batch size: 191, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:56:15,129 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190408.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:56:50,436 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190432.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:12,485 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 22:57:21,413 INFO [train.py:904] (0/8) Epoch 19, batch 7750, loss[loss=0.2163, simple_loss=0.2909, pruned_loss=0.07079, over 11398.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2921, pruned_loss=0.06191, over 3048705.37 frames. ], batch size: 248, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:57:35,056 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190460.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:57:36,358 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4396, 3.4585, 3.8814, 1.7229, 4.0136, 4.0668, 2.9109, 2.8692], device='cuda:0'), covar=tensor([0.0861, 0.0255, 0.0199, 0.1352, 0.0073, 0.0178, 0.0455, 0.0503], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0106, 0.0096, 0.0138, 0.0077, 0.0121, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 22:57:43,810 INFO [optim.py:368] (0/8) 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 22:58:30,078 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190495.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 22:58:33,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5819, 3.7359, 2.8178, 2.1668, 2.4954, 2.3462, 4.0582, 3.3484], device='cuda:0'), covar=tensor([0.3133, 0.0712, 0.1868, 0.2752, 0.2780, 0.2138, 0.0454, 0.1308], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0265, 0.0299, 0.0304, 0.0293, 0.0251, 0.0288, 0.0327], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 22:58:39,269 INFO [train.py:904] (0/8) Epoch 19, batch 7800, loss[loss=0.213, simple_loss=0.3008, pruned_loss=0.06265, over 16503.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2936, pruned_loss=0.06285, over 3046747.43 frames. ], batch size: 146, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 22:58:47,783 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190507.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 22:59:02,312 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1770, 1.5368, 1.8926, 2.1230, 2.1812, 2.3455, 1.7125, 2.2954], device='cuda:0'), covar=tensor([0.0216, 0.0459, 0.0270, 0.0292, 0.0276, 0.0177, 0.0431, 0.0135], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0187, 0.0173, 0.0177, 0.0188, 0.0147, 0.0189, 0.0140], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:59:02,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2493, 2.0694, 1.6538, 1.7482, 2.3307, 2.0056, 2.0403, 2.4288], device='cuda:0'), covar=tensor([0.0187, 0.0360, 0.0492, 0.0437, 0.0236, 0.0331, 0.0203, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0227, 0.0220, 0.0220, 0.0229, 0.0226, 0.0229, 0.0222], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 22:59:57,142 INFO [train.py:904] (0/8) Epoch 19, batch 7850, loss[loss=0.1967, simple_loss=0.2941, pruned_loss=0.04966, over 16868.00 frames. ], tot_loss[loss=0.2094, simple_loss=0.2939, pruned_loss=0.06241, over 3052195.44 frames. ], batch size: 96, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:00:17,837 INFO [optim.py:368] (0/8) 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 23:00:56,328 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190591.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:01:12,381 INFO [train.py:904] (0/8) Epoch 19, batch 7900, loss[loss=0.2095, simple_loss=0.2919, pruned_loss=0.06352, over 16633.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2924, pruned_loss=0.06162, over 3050900.46 frames. ], batch size: 57, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:01:23,366 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190609.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:02:28,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1440, 3.4185, 3.4223, 2.0129, 3.0342, 2.1371, 3.5649, 3.6596], device='cuda:0'), covar=tensor([0.0208, 0.0700, 0.0566, 0.2084, 0.0795, 0.1070, 0.0575, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0160, 0.0165, 0.0150, 0.0143, 0.0128, 0.0142, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-04-30 23:02:32,491 INFO [train.py:904] (0/8) Epoch 19, batch 7950, loss[loss=0.2293, simple_loss=0.3016, pruned_loss=0.07851, over 11998.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2928, pruned_loss=0.06199, over 3052655.69 frames. ], batch size: 246, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:02:32,945 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190652.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:02:54,094 INFO [optim.py:368] (0/8) 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 23:03:00,804 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190670.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:03:04,355 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190672.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:08,658 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=190675.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:03:22,588 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9787, 4.9993, 4.8592, 4.4913, 4.4964, 4.9195, 4.8022, 4.5660], device='cuda:0'), covar=tensor([0.0651, 0.0481, 0.0283, 0.0308, 0.1020, 0.0471, 0.0439, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0404, 0.0326, 0.0319, 0.0338, 0.0372, 0.0227, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:03:49,548 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-04-30 23:03:49,826 INFO [train.py:904] (0/8) Epoch 19, batch 8000, loss[loss=0.2082, simple_loss=0.2969, pruned_loss=0.05977, over 16710.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2938, pruned_loss=0.06268, over 3052797.33 frames. ], batch size: 89, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:04:43,887 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=190736.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:05,193 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6828, 2.4939, 2.3249, 3.6274, 2.5468, 3.8350, 1.4139, 2.7852], device='cuda:0'), covar=tensor([0.1350, 0.0813, 0.1296, 0.0221, 0.0206, 0.0389, 0.1689, 0.0843], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0184, 0.0207, 0.0214, 0.0198, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 23:05:07,653 INFO [train.py:904] (0/8) Epoch 19, batch 8050, loss[loss=0.2155, simple_loss=0.2994, pruned_loss=0.0658, over 15451.00 frames. ], tot_loss[loss=0.2085, simple_loss=0.2929, pruned_loss=0.06204, over 3055737.87 frames. ], batch size: 190, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:05:12,108 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190755.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:05:28,128 INFO [optim.py:368] (0/8) 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 23:06:12,399 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190795.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:06:22,320 INFO [train.py:904] (0/8) Epoch 19, batch 8100, loss[loss=0.2122, simple_loss=0.2996, pruned_loss=0.06235, over 16671.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2925, pruned_loss=0.0617, over 3034452.21 frames. ], batch size: 124, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:06:30,756 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190807.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:07:25,608 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190843.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:07:39,356 INFO [train.py:904] (0/8) Epoch 19, batch 8150, loss[loss=0.176, simple_loss=0.2635, pruned_loss=0.04422, over 16205.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2896, pruned_loss=0.06005, over 3060419.73 frames. ], batch size: 165, lr: 3.54e-03, grad_scale: 8.0 2023-04-30 23:07:44,594 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=190855.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:07:57,765 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2439, 2.3611, 2.3918, 4.1799, 2.2576, 2.6247, 2.4167, 2.4483], device='cuda:0'), covar=tensor([0.1246, 0.3263, 0.2700, 0.0435, 0.3848, 0.2408, 0.3389, 0.3192], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0433, 0.0355, 0.0320, 0.0432, 0.0498, 0.0404, 0.0506], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:08:01,322 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.839e+02 2.760e+02 3.332e+02 4.142e+02 6.995e+02, threshold=6.664e+02, percent-clipped=2.0 2023-04-30 23:08:57,394 INFO [train.py:904] (0/8) Epoch 19, batch 8200, loss[loss=0.1829, simple_loss=0.2726, pruned_loss=0.04656, over 16584.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2871, pruned_loss=0.05909, over 3061976.07 frames. ], batch size: 62, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:08:59,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1828, 4.2549, 4.5883, 4.5542, 4.5664, 4.2817, 4.2921, 4.2231], device='cuda:0'), covar=tensor([0.0350, 0.0646, 0.0450, 0.0464, 0.0451, 0.0452, 0.0950, 0.0545], device='cuda:0'), in_proj_covar=tensor([0.0396, 0.0434, 0.0421, 0.0397, 0.0468, 0.0443, 0.0539, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 23:10:11,513 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190947.0, num_to_drop=1, layers_to_drop={3} 2023-04-30 23:10:19,457 INFO [train.py:904] (0/8) Epoch 19, batch 8250, loss[loss=0.2031, simple_loss=0.294, pruned_loss=0.05608, over 15341.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2859, pruned_loss=0.05703, over 3033492.43 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:10:39,961 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=190965.0, num_to_drop=1, layers_to_drop={1} 2023-04-30 23:10:41,156 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.632e+02 2.606e+02 3.069e+02 3.589e+02 7.796e+02, threshold=6.137e+02, percent-clipped=2.0 2023-04-30 23:10:51,837 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=190972.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:11:19,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9564, 2.1215, 2.3476, 3.2040, 2.2026, 2.3176, 2.3108, 2.2071], device='cuda:0'), covar=tensor([0.1117, 0.3591, 0.2599, 0.0634, 0.4559, 0.2770, 0.3489, 0.3744], device='cuda:0'), in_proj_covar=tensor([0.0385, 0.0429, 0.0352, 0.0316, 0.0426, 0.0492, 0.0400, 0.0499], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:11:34,297 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.64 vs. limit=5.0 2023-04-30 23:11:39,011 INFO [train.py:904] (0/8) Epoch 19, batch 8300, loss[loss=0.1722, simple_loss=0.2722, pruned_loss=0.03608, over 16716.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2833, pruned_loss=0.05398, over 3032777.75 frames. ], batch size: 89, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:12:10,266 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191020.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:12:27,707 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191031.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:00,825 INFO [train.py:904] (0/8) Epoch 19, batch 8350, loss[loss=0.1924, simple_loss=0.2746, pruned_loss=0.05508, over 11919.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.283, pruned_loss=0.05222, over 3038020.91 frames. ], batch size: 246, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:13:07,420 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191055.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:13:24,350 INFO [optim.py:368] (0/8) 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 23:13:28,660 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191068.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:14:22,694 INFO [train.py:904] (0/8) Epoch 19, batch 8400, loss[loss=0.1654, simple_loss=0.2508, pruned_loss=0.04007, over 12369.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.281, pruned_loss=0.05032, over 3036999.21 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:14:24,468 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191103.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:06,895 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191129.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:11,029 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191132.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:15:20,270 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6878, 2.6337, 2.3825, 3.7268, 2.1478, 3.7814, 1.6007, 2.7114], device='cuda:0'), covar=tensor([0.1330, 0.0727, 0.1144, 0.0190, 0.0115, 0.0451, 0.1514, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0171, 0.0191, 0.0181, 0.0204, 0.0211, 0.0196, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-04-30 23:15:42,698 INFO [train.py:904] (0/8) Epoch 19, batch 8450, loss[loss=0.1765, simple_loss=0.2575, pruned_loss=0.0478, over 12617.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2788, pruned_loss=0.04835, over 3039452.69 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:16:06,335 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.605e+02 2.247e+02 2.599e+02 3.228e+02 6.628e+02, threshold=5.197e+02, percent-clipped=1.0 2023-04-30 23:16:36,998 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191185.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:16:49,436 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191193.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:17:04,716 INFO [train.py:904] (0/8) Epoch 19, batch 8500, loss[loss=0.1525, simple_loss=0.2509, pruned_loss=0.02705, over 16503.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2753, pruned_loss=0.04604, over 3052096.58 frames. ], batch size: 75, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:17:10,441 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-04-30 23:17:24,152 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8798, 2.6659, 2.4230, 1.8553, 2.5966, 2.7002, 2.5890, 1.7691], device='cuda:0'), covar=tensor([0.0404, 0.0087, 0.0083, 0.0348, 0.0117, 0.0119, 0.0119, 0.0545], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0078, 0.0079, 0.0132, 0.0093, 0.0105, 0.0091, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 23:17:43,159 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9297, 2.7347, 2.4873, 1.9795, 2.5278, 2.6806, 2.6074, 1.9871], device='cuda:0'), covar=tensor([0.0352, 0.0072, 0.0062, 0.0293, 0.0108, 0.0100, 0.0094, 0.0357], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0078, 0.0079, 0.0131, 0.0093, 0.0105, 0.0090, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 23:18:22,286 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191246.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:18:24,104 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191247.0, num_to_drop=1, layers_to_drop={2} 2023-04-30 23:18:32,470 INFO [train.py:904] (0/8) Epoch 19, batch 8550, loss[loss=0.1836, simple_loss=0.2815, pruned_loss=0.04284, over 16811.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2727, pruned_loss=0.04509, over 3046871.48 frames. ], batch size: 124, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:18:57,228 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.506e+02 2.304e+02 2.724e+02 3.153e+02 7.683e+02, threshold=5.448e+02, percent-clipped=1.0 2023-04-30 23:19:04,573 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3496, 4.3587, 4.2081, 3.9528, 3.9316, 4.3041, 4.0904, 4.0156], device='cuda:0'), covar=tensor([0.0557, 0.0587, 0.0307, 0.0278, 0.0774, 0.0500, 0.0637, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0275, 0.0399, 0.0323, 0.0314, 0.0331, 0.0367, 0.0224, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:19:56,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2192, 4.2763, 4.1318, 3.8654, 3.8165, 4.2293, 3.9129, 3.9499], device='cuda:0'), covar=tensor([0.0547, 0.0494, 0.0266, 0.0254, 0.0677, 0.0460, 0.0752, 0.0610], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0397, 0.0322, 0.0312, 0.0330, 0.0365, 0.0223, 0.0383], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:19:59,597 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191295.0, num_to_drop=1, layers_to_drop={0} 2023-04-30 23:20:13,439 INFO [train.py:904] (0/8) Epoch 19, batch 8600, loss[loss=0.1944, simple_loss=0.2931, pruned_loss=0.04789, over 15390.00 frames. ], tot_loss[loss=0.1809, simple_loss=0.2731, pruned_loss=0.04429, over 3042463.96 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:20:37,123 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191313.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:11,857 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191331.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:21:52,940 INFO [train.py:904] (0/8) Epoch 19, batch 8650, loss[loss=0.1646, simple_loss=0.2639, pruned_loss=0.03266, over 15369.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.271, pruned_loss=0.04231, over 3058988.68 frames. ], batch size: 190, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:22:06,122 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2361, 3.0390, 3.0972, 1.7925, 3.2837, 3.3508, 2.7893, 2.7918], device='cuda:0'), covar=tensor([0.0734, 0.0235, 0.0223, 0.1168, 0.0086, 0.0170, 0.0425, 0.0399], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0103, 0.0093, 0.0134, 0.0075, 0.0117, 0.0122, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-04-30 23:22:25,677 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=191365.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:22:29,964 INFO [optim.py:368] (0/8) 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 23:22:55,706 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191379.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:23:07,621 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8378, 1.3297, 1.7380, 1.7626, 1.8794, 1.8606, 1.6894, 1.8481], device='cuda:0'), covar=tensor([0.0223, 0.0407, 0.0202, 0.0260, 0.0252, 0.0177, 0.0386, 0.0124], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0186, 0.0171, 0.0175, 0.0186, 0.0145, 0.0187, 0.0139], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:23:39,032 INFO [train.py:904] (0/8) Epoch 19, batch 8700, loss[loss=0.1723, simple_loss=0.2606, pruned_loss=0.04201, over 12415.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2688, pruned_loss=0.04136, over 3064835.67 frames. ], batch size: 247, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:24:20,708 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191424.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:24:23,571 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=191426.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:25:14,448 INFO [train.py:904] (0/8) Epoch 19, batch 8750, loss[loss=0.1873, simple_loss=0.2842, pruned_loss=0.04521, over 16678.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2681, pruned_loss=0.04074, over 3056908.48 frames. ], batch size: 134, lr: 3.53e-03, grad_scale: 2.0 2023-04-30 23:25:35,359 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8025, 3.7277, 3.9318, 3.6992, 3.9025, 4.2967, 3.9300, 3.5943], device='cuda:0'), covar=tensor([0.2061, 0.2500, 0.2257, 0.2538, 0.2632, 0.1493, 0.1644, 0.2509], device='cuda:0'), in_proj_covar=tensor([0.0383, 0.0551, 0.0611, 0.0460, 0.0610, 0.0639, 0.0479, 0.0617], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 23:25:58,341 INFO [optim.py:368] (0/8) 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 23:26:27,299 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-04-30 23:26:39,718 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191488.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:27:07,640 INFO [train.py:904] (0/8) Epoch 19, batch 8800, loss[loss=0.1773, simple_loss=0.2639, pruned_loss=0.04535, over 12493.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.266, pruned_loss=0.03949, over 3057141.55 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:27:42,551 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4506, 3.0468, 2.6443, 2.2176, 2.1944, 2.2736, 3.0596, 2.8714], device='cuda:0'), covar=tensor([0.2443, 0.0773, 0.1710, 0.2710, 0.2453, 0.2155, 0.0510, 0.1369], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0257, 0.0292, 0.0296, 0.0283, 0.0244, 0.0279, 0.0318], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-04-30 23:27:48,791 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5907, 3.5523, 3.5347, 2.7912, 3.4999, 1.9660, 3.2592, 2.9182], device='cuda:0'), covar=tensor([0.0143, 0.0110, 0.0184, 0.0243, 0.0114, 0.2692, 0.0133, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0141, 0.0184, 0.0166, 0.0161, 0.0196, 0.0174, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:28:28,989 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191541.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:28:45,604 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-04-30 23:28:51,763 INFO [train.py:904] (0/8) Epoch 19, batch 8850, loss[loss=0.1917, simple_loss=0.2924, pruned_loss=0.04544, over 16871.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2683, pruned_loss=0.03907, over 3042370.31 frames. ], batch size: 116, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:29:27,905 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.97 vs. limit=5.0 2023-04-30 23:29:28,631 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.662e+02 2.105e+02 2.539e+02 3.345e+02 5.950e+02, threshold=5.078e+02, percent-clipped=2.0 2023-04-30 23:30:39,788 INFO [train.py:904] (0/8) Epoch 19, batch 8900, loss[loss=0.1704, simple_loss=0.2688, pruned_loss=0.03601, over 16282.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2685, pruned_loss=0.03856, over 3025906.26 frames. ], batch size: 166, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:32:41,894 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-04-30 23:32:45,280 INFO [train.py:904] (0/8) Epoch 19, batch 8950, loss[loss=0.1635, simple_loss=0.2612, pruned_loss=0.03288, over 16267.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2691, pruned_loss=0.03924, over 3050457.57 frames. ], batch size: 165, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:33:01,365 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-04-30 23:33:21,164 INFO [optim.py:368] (0/8) 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 23:34:18,622 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8175, 3.8152, 4.1297, 4.1176, 4.1175, 3.8988, 3.9201, 3.9788], device='cuda:0'), covar=tensor([0.0373, 0.0886, 0.0510, 0.0498, 0.0518, 0.0549, 0.0905, 0.0540], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0414, 0.0406, 0.0381, 0.0447, 0.0425, 0.0514, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 23:34:34,405 INFO [train.py:904] (0/8) Epoch 19, batch 9000, loss[loss=0.1844, simple_loss=0.2722, pruned_loss=0.04833, over 12337.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2656, pruned_loss=0.03801, over 3049566.43 frames. ], batch size: 248, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:34:34,406 INFO [train.py:929] (0/8) Computing validation loss 2023-04-30 23:34:44,209 INFO [train.py:938] (0/8) Epoch 19, validation: loss=0.1462, simple_loss=0.2506, pruned_loss=0.02087, over 944034.00 frames. 2023-04-30 23:34:44,210 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-04-30 23:35:17,758 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-04-30 23:35:25,532 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=191721.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:35:25,953 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-04-30 23:35:31,023 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191724.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:36:27,799 INFO [train.py:904] (0/8) Epoch 19, batch 9050, loss[loss=0.1632, simple_loss=0.2478, pruned_loss=0.03932, over 16645.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2654, pruned_loss=0.03789, over 3073122.17 frames. ], batch size: 134, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:37:04,258 INFO [optim.py:368] (0/8) 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 23:37:09,325 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191772.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:37:40,004 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191788.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:38:10,481 INFO [train.py:904] (0/8) Epoch 19, batch 9100, loss[loss=0.1783, simple_loss=0.2774, pruned_loss=0.03961, over 16821.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2652, pruned_loss=0.03828, over 3063332.89 frames. ], batch size: 124, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:39:32,425 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191836.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:39:43,612 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5545, 3.5433, 3.4990, 2.8417, 3.4280, 1.9785, 3.2238, 2.9304], device='cuda:0'), covar=tensor([0.0131, 0.0123, 0.0173, 0.0194, 0.0106, 0.2316, 0.0126, 0.0242], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0142, 0.0185, 0.0166, 0.0162, 0.0197, 0.0175, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:39:43,617 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=191841.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:40:08,206 INFO [train.py:904] (0/8) Epoch 19, batch 9150, loss[loss=0.1747, simple_loss=0.2648, pruned_loss=0.04223, over 16838.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2663, pruned_loss=0.03843, over 3070855.84 frames. ], batch size: 124, lr: 3.53e-03, grad_scale: 4.0 2023-04-30 23:40:23,607 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-04-30 23:40:46,345 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.243e+02 2.682e+02 3.704e+02 5.734e+02, threshold=5.364e+02, percent-clipped=7.0 2023-04-30 23:41:25,866 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3333, 4.3446, 4.7155, 4.7071, 4.7028, 4.4347, 4.3995, 4.3673], device='cuda:0'), covar=tensor([0.0334, 0.0725, 0.0515, 0.0431, 0.0451, 0.0487, 0.0897, 0.0476], device='cuda:0'), in_proj_covar=tensor([0.0382, 0.0415, 0.0408, 0.0382, 0.0449, 0.0426, 0.0517, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 23:41:30,262 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=191889.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:41:52,743 INFO [train.py:904] (0/8) Epoch 19, batch 9200, loss[loss=0.1684, simple_loss=0.2586, pruned_loss=0.03911, over 16862.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2619, pruned_loss=0.03761, over 3066105.04 frames. ], batch size: 124, lr: 3.53e-03, grad_scale: 8.0 2023-04-30 23:43:05,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3646, 1.7080, 2.0272, 2.3169, 2.3997, 2.5199, 1.8393, 2.5582], device='cuda:0'), covar=tensor([0.0226, 0.0509, 0.0323, 0.0323, 0.0334, 0.0221, 0.0515, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0184, 0.0169, 0.0172, 0.0184, 0.0142, 0.0187, 0.0137], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:43:29,330 INFO [train.py:904] (0/8) Epoch 19, batch 9250, loss[loss=0.1705, simple_loss=0.265, pruned_loss=0.03796, over 16765.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2621, pruned_loss=0.03791, over 3055960.41 frames. ], batch size: 124, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:43:34,487 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-04-30 23:44:05,913 INFO [optim.py:368] (0/8) 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 23:45:16,326 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-192000.pt 2023-04-30 23:45:23,378 INFO [train.py:904] (0/8) Epoch 19, batch 9300, loss[loss=0.155, simple_loss=0.2403, pruned_loss=0.03486, over 12202.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2611, pruned_loss=0.03776, over 3058455.57 frames. ], batch size: 247, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:46:09,815 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192021.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:47:09,794 INFO [train.py:904] (0/8) Epoch 19, batch 9350, loss[loss=0.1833, simple_loss=0.2744, pruned_loss=0.04605, over 15285.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.26, pruned_loss=0.03727, over 3056036.45 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:47:46,319 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 23:48:49,315 INFO [train.py:904] (0/8) Epoch 19, batch 9400, loss[loss=0.1713, simple_loss=0.2755, pruned_loss=0.0336, over 15392.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2599, pruned_loss=0.03691, over 3056239.81 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:48:50,426 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192102.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:29,972 INFO [train.py:904] (0/8) Epoch 19, batch 9450, loss[loss=0.1605, simple_loss=0.2521, pruned_loss=0.03441, over 16923.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2615, pruned_loss=0.03724, over 3038683.65 frames. ], batch size: 116, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:50:35,893 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192155.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:50:51,063 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.168e+02 2.516e+02 3.148e+02 5.649e+02, threshold=5.031e+02, percent-clipped=1.0 2023-04-30 23:51:28,323 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.76 vs. limit=2.0 2023-04-30 23:51:32,014 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7964, 2.6158, 2.6016, 4.4536, 3.0089, 4.2940, 1.6283, 3.1542], device='cuda:0'), covar=tensor([0.1303, 0.0784, 0.1099, 0.0183, 0.0103, 0.0289, 0.1534, 0.0602], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0167, 0.0188, 0.0175, 0.0195, 0.0206, 0.0193, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 23:51:33,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4721, 3.8708, 3.8920, 2.8151, 3.4277, 3.8551, 3.6598, 2.2583], device='cuda:0'), covar=tensor([0.0492, 0.0037, 0.0040, 0.0358, 0.0107, 0.0090, 0.0071, 0.0477], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0077, 0.0078, 0.0131, 0.0093, 0.0103, 0.0089, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 23:51:36,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8499, 2.0613, 2.2767, 3.1826, 2.1125, 2.2295, 2.2434, 2.1672], device='cuda:0'), covar=tensor([0.1278, 0.3751, 0.2786, 0.0711, 0.4678, 0.2680, 0.3659, 0.3789], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0422, 0.0350, 0.0310, 0.0421, 0.0484, 0.0394, 0.0491], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:51:47,885 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8831, 4.9227, 5.3349, 5.2965, 5.3027, 5.0099, 4.9456, 4.8029], device='cuda:0'), covar=tensor([0.0372, 0.0797, 0.0366, 0.0420, 0.0513, 0.0380, 0.0909, 0.0423], device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0408, 0.0402, 0.0377, 0.0444, 0.0420, 0.0508, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-04-30 23:52:10,464 INFO [train.py:904] (0/8) Epoch 19, batch 9500, loss[loss=0.147, simple_loss=0.2361, pruned_loss=0.02894, over 12843.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2605, pruned_loss=0.03656, over 3045101.48 frames. ], batch size: 247, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:52:39,653 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192216.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:45,077 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192219.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:49,405 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192221.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:52:55,364 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9581, 2.2086, 2.3130, 3.0350, 1.9124, 3.3175, 1.6727, 2.7845], device='cuda:0'), covar=tensor([0.1250, 0.0742, 0.1089, 0.0150, 0.0116, 0.0405, 0.1506, 0.0708], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0167, 0.0188, 0.0175, 0.0195, 0.0207, 0.0193, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-04-30 23:53:55,158 INFO [train.py:904] (0/8) Epoch 19, batch 9550, loss[loss=0.1905, simple_loss=0.2872, pruned_loss=0.04693, over 15185.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2607, pruned_loss=0.03692, over 3042037.81 frames. ], batch size: 190, lr: 3.52e-03, grad_scale: 4.0 2023-04-30 23:54:10,472 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4898, 2.0280, 1.7657, 1.6827, 2.2760, 1.9456, 1.9792, 2.3815], device='cuda:0'), covar=tensor([0.0161, 0.0394, 0.0497, 0.0476, 0.0282, 0.0355, 0.0214, 0.0250], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0223, 0.0215, 0.0216, 0.0225, 0.0222, 0.0221, 0.0214], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-04-30 23:54:34,516 INFO [optim.py:368] (0/8) 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 23:54:55,574 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192280.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:00,673 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192282.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:55:38,449 INFO [train.py:904] (0/8) Epoch 19, batch 9600, loss[loss=0.1776, simple_loss=0.2639, pruned_loss=0.04564, over 12417.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2617, pruned_loss=0.03737, over 3030128.48 frames. ], batch size: 247, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:55:54,718 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4120, 4.5046, 4.6705, 4.4514, 4.5416, 5.0356, 4.5690, 4.2547], device='cuda:0'), covar=tensor([0.1383, 0.1992, 0.2107, 0.2178, 0.2565, 0.1093, 0.1683, 0.2407], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0547, 0.0608, 0.0456, 0.0608, 0.0637, 0.0476, 0.0610], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-04-30 23:57:27,648 INFO [train.py:904] (0/8) Epoch 19, batch 9650, loss[loss=0.1559, simple_loss=0.2497, pruned_loss=0.03108, over 16442.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2637, pruned_loss=0.03742, over 3046002.77 frames. ], batch size: 68, lr: 3.52e-03, grad_scale: 8.0 2023-04-30 23:58:09,767 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.165e+02 2.448e+02 2.941e+02 5.888e+02, threshold=4.896e+02, percent-clipped=1.0 2023-04-30 23:58:54,021 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192392.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:09,401 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192399.0, num_to_drop=0, layers_to_drop=set() 2023-04-30 23:59:15,182 INFO [train.py:904] (0/8) Epoch 19, batch 9700, loss[loss=0.1732, simple_loss=0.2711, pruned_loss=0.03758, over 16677.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2631, pruned_loss=0.03714, over 3083714.81 frames. ], batch size: 134, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:19,350 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192433.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:00:57,150 INFO [train.py:904] (0/8) Epoch 19, batch 9750, loss[loss=0.172, simple_loss=0.2678, pruned_loss=0.03809, over 16262.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2624, pruned_loss=0.0374, over 3062352.94 frames. ], batch size: 165, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:00:58,995 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:09,806 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:14,132 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:01:31,657 INFO [optim.py:368] (0/8) 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 00:02:23,698 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:02:37,077 INFO [train.py:904] (0/8) Epoch 19, batch 9800, loss[loss=0.1629, simple_loss=0.249, pruned_loss=0.03839, over 12409.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2627, pruned_loss=0.03663, over 3077131.58 frames. ], batch size: 246, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:02:55,416 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192511.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:03:10,326 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.12 vs. limit=2.0 2023-05-01 00:03:17,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2221, 4.2968, 4.4344, 4.2282, 4.3387, 4.8164, 4.3585, 4.0592], device='cuda:0'), covar=tensor([0.1596, 0.2223, 0.2315, 0.2249, 0.2570, 0.1206, 0.1704, 0.2594], device='cuda:0'), in_proj_covar=tensor([0.0374, 0.0544, 0.0601, 0.0452, 0.0604, 0.0631, 0.0471, 0.0604], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 00:03:51,960 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 00:04:09,001 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.56 vs. limit=5.0 2023-05-01 00:04:23,527 INFO [train.py:904] (0/8) Epoch 19, batch 9850, loss[loss=0.1454, simple_loss=0.2546, pruned_loss=0.01811, over 17194.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2642, pruned_loss=0.03665, over 3073253.94 frames. ], batch size: 46, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:05:00,569 INFO [optim.py:368] (0/8) 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 00:05:11,441 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192575.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:05:15,473 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:06:14,640 INFO [train.py:904] (0/8) Epoch 19, batch 9900, loss[loss=0.1895, simple_loss=0.2737, pruned_loss=0.05267, over 12269.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2644, pruned_loss=0.03676, over 3070562.38 frames. ], batch size: 246, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:06:29,231 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9778, 5.2859, 5.0833, 5.0998, 4.8128, 4.8291, 4.6917, 5.3404], device='cuda:0'), covar=tensor([0.1267, 0.0898, 0.0985, 0.0791, 0.0819, 0.0825, 0.1110, 0.0903], device='cuda:0'), in_proj_covar=tensor([0.0617, 0.0758, 0.0614, 0.0562, 0.0475, 0.0484, 0.0629, 0.0579], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:06:46,552 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:07:56,786 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-01 00:08:04,205 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2995, 5.2944, 5.1422, 4.6536, 4.7441, 5.1758, 5.1570, 4.7712], device='cuda:0'), covar=tensor([0.0574, 0.0660, 0.0290, 0.0312, 0.1086, 0.0656, 0.0237, 0.0676], device='cuda:0'), in_proj_covar=tensor([0.0268, 0.0381, 0.0312, 0.0302, 0.0318, 0.0355, 0.0215, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-01 00:08:13,292 INFO [train.py:904] (0/8) Epoch 19, batch 9950, loss[loss=0.1786, simple_loss=0.2758, pruned_loss=0.04074, over 16385.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2662, pruned_loss=0.03713, over 3058706.45 frames. ], batch size: 146, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:08:54,789 INFO [optim.py:368] (0/8) 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 00:09:04,344 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:14,966 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:09:24,419 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3378, 4.3275, 4.1923, 3.6627, 4.2828, 1.7453, 4.0298, 3.9813], device='cuda:0'), covar=tensor([0.0085, 0.0098, 0.0183, 0.0248, 0.0099, 0.2639, 0.0129, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0139, 0.0182, 0.0162, 0.0159, 0.0196, 0.0172, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:10:14,424 INFO [train.py:904] (0/8) Epoch 19, batch 10000, loss[loss=0.1568, simple_loss=0.2628, pruned_loss=0.02542, over 17125.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2654, pruned_loss=0.03708, over 3064241.67 frames. ], batch size: 49, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:11:08,507 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3344, 3.0456, 2.7242, 2.2194, 2.1072, 2.2588, 3.0266, 2.7828], device='cuda:0'), covar=tensor([0.2653, 0.0609, 0.1580, 0.2678, 0.2527, 0.2119, 0.0432, 0.1349], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0254, 0.0288, 0.0293, 0.0275, 0.0241, 0.0276, 0.0314], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:11:12,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9025, 2.1493, 2.3099, 3.2301, 2.1200, 2.3153, 2.2904, 2.2347], device='cuda:0'), covar=tensor([0.1179, 0.3598, 0.2664, 0.0642, 0.4198, 0.2500, 0.3496, 0.3366], device='cuda:0'), in_proj_covar=tensor([0.0379, 0.0420, 0.0348, 0.0308, 0.0419, 0.0480, 0.0391, 0.0488], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:11:22,411 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192734.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:25,352 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192735.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:50,882 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:11:56,966 INFO [train.py:904] (0/8) Epoch 19, batch 10050, loss[loss=0.1898, simple_loss=0.2875, pruned_loss=0.04611, over 15370.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2654, pruned_loss=0.03706, over 3054412.93 frames. ], batch size: 191, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:12:02,964 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192755.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:09,639 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:12:32,889 INFO [optim.py:368] (0/8) 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 00:13:08,106 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:21,323 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192796.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:24,701 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2377, 3.4534, 3.5084, 2.4771, 3.1687, 3.5176, 3.3205, 1.9425], device='cuda:0'), covar=tensor([0.0463, 0.0061, 0.0048, 0.0353, 0.0109, 0.0096, 0.0094, 0.0507], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0077, 0.0077, 0.0130, 0.0092, 0.0102, 0.0088, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 00:13:30,585 INFO [train.py:904] (0/8) Epoch 19, batch 10100, loss[loss=0.1655, simple_loss=0.2568, pruned_loss=0.03713, over 16472.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2657, pruned_loss=0.03715, over 3049395.68 frames. ], batch size: 75, lr: 3.52e-03, grad_scale: 8.0 2023-05-01 00:13:38,858 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:13:47,908 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192811.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:14:28,791 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=192831.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:14:42,122 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 00:14:52,033 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-19.pt 2023-05-01 00:15:13,692 INFO [train.py:904] (0/8) Epoch 20, batch 0, loss[loss=0.1878, simple_loss=0.2779, pruned_loss=0.04883, over 17130.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2779, pruned_loss=0.04883, over 17130.00 frames. ], batch size: 48, lr: 3.43e-03, grad_scale: 8.0 2023-05-01 00:15:13,692 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 00:15:21,162 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 00:15:31,954 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:45,147 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6077, 2.3807, 2.2920, 4.5141, 2.3127, 2.7384, 2.4019, 2.4825], device='cuda:0'), covar=tensor([0.1173, 0.3669, 0.3131, 0.0413, 0.4173, 0.2593, 0.3488, 0.3829], device='cuda:0'), in_proj_covar=tensor([0.0383, 0.0423, 0.0352, 0.0310, 0.0423, 0.0484, 0.0395, 0.0492], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:15:49,198 INFO [optim.py:368] (0/8) 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 00:15:53,724 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192875.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:15:54,064 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 00:15:56,037 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=192877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:16:17,628 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=192892.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:16:31,061 INFO [train.py:904] (0/8) Epoch 20, batch 50, loss[loss=0.1781, simple_loss=0.2739, pruned_loss=0.04114, over 17039.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2717, pruned_loss=0.0489, over 754471.16 frames. ], batch size: 50, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:16:31,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6091, 4.9208, 4.6889, 4.6802, 4.4519, 4.4343, 4.4166, 4.9938], device='cuda:0'), covar=tensor([0.1295, 0.1029, 0.1278, 0.0888, 0.0908, 0.1315, 0.1252, 0.0926], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0766, 0.0623, 0.0567, 0.0480, 0.0489, 0.0639, 0.0584], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:16:59,575 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:02,459 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=192925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:17:36,902 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9249, 4.8994, 5.3613, 5.3516, 5.3840, 5.0180, 4.9550, 4.8079], device='cuda:0'), covar=tensor([0.0366, 0.0495, 0.0427, 0.0478, 0.0501, 0.0447, 0.1032, 0.0454], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0413, 0.0405, 0.0379, 0.0449, 0.0425, 0.0514, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:17:37,187 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 00:17:38,946 INFO [train.py:904] (0/8) Epoch 20, batch 100, loss[loss=0.2, simple_loss=0.2717, pruned_loss=0.06419, over 16940.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2677, pruned_loss=0.04683, over 1330514.63 frames. ], batch size: 109, lr: 3.43e-03, grad_scale: 2.0 2023-05-01 00:18:07,334 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=192972.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:18:15,950 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6107, 4.5738, 4.9757, 4.9826, 5.0410, 4.6498, 4.6336, 4.5155], device='cuda:0'), covar=tensor([0.0376, 0.0598, 0.0382, 0.0412, 0.0501, 0.0490, 0.1045, 0.0543], device='cuda:0'), in_proj_covar=tensor([0.0382, 0.0416, 0.0407, 0.0381, 0.0452, 0.0427, 0.0517, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:18:48,439 INFO [train.py:904] (0/8) Epoch 20, batch 150, loss[loss=0.1809, simple_loss=0.2594, pruned_loss=0.05118, over 16799.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2655, pruned_loss=0.04652, over 1776811.14 frames. ], batch size: 83, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:19:27,076 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193029.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:43,961 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:52,907 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:19:54,462 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 00:19:58,179 INFO [train.py:904] (0/8) Epoch 20, batch 200, loss[loss=0.1755, simple_loss=0.2578, pruned_loss=0.04657, over 16679.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2663, pruned_loss=0.04714, over 2114628.11 frames. ], batch size: 89, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:20:03,603 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193055.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:27,502 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.769e+02 2.323e+02 2.688e+02 3.539e+02 1.444e+03, threshold=5.377e+02, percent-clipped=5.0 2023-05-01 00:20:39,532 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1188, 5.8204, 5.9387, 5.6705, 5.8040, 6.3069, 5.8091, 5.5110], device='cuda:0'), covar=tensor([0.0897, 0.1839, 0.2093, 0.1842, 0.2277, 0.0915, 0.1522, 0.2341], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0562, 0.0622, 0.0469, 0.0627, 0.0651, 0.0489, 0.0624], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 00:20:42,438 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.65 vs. limit=5.0 2023-05-01 00:20:50,745 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:53,818 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193091.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:20:57,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8144, 5.1413, 4.8984, 4.9040, 4.6545, 4.6174, 4.6170, 5.2242], device='cuda:0'), covar=tensor([0.1272, 0.1057, 0.1251, 0.0933, 0.0962, 0.1147, 0.1170, 0.1079], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0787, 0.0638, 0.0583, 0.0492, 0.0501, 0.0655, 0.0601], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:21:00,959 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:08,812 INFO [train.py:904] (0/8) Epoch 20, batch 250, loss[loss=0.1595, simple_loss=0.2566, pruned_loss=0.03123, over 17043.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2647, pruned_loss=0.04683, over 2377795.69 frames. ], batch size: 50, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:21:09,314 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193102.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:10,239 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:21:57,543 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193137.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:22:07,356 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1959, 2.1735, 2.3325, 3.9082, 2.1974, 2.4978, 2.2274, 2.3523], device='cuda:0'), covar=tensor([0.1392, 0.3764, 0.3075, 0.0644, 0.3917, 0.2575, 0.3702, 0.3188], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0431, 0.0358, 0.0318, 0.0429, 0.0494, 0.0402, 0.0502], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:22:17,131 INFO [train.py:904] (0/8) Epoch 20, batch 300, loss[loss=0.1487, simple_loss=0.2341, pruned_loss=0.03163, over 15982.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2623, pruned_loss=0.04532, over 2589821.44 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:22:46,398 INFO [optim.py:368] (0/8) 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 00:23:05,594 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193187.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:23:10,401 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0664, 4.0821, 2.7753, 4.7355, 3.3304, 4.6575, 2.7977, 3.4368], device='cuda:0'), covar=tensor([0.0290, 0.0413, 0.1483, 0.0279, 0.0790, 0.0589, 0.1523, 0.0689], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0176, 0.0193, 0.0157, 0.0176, 0.0212, 0.0202, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:23:28,161 INFO [train.py:904] (0/8) Epoch 20, batch 350, loss[loss=0.1685, simple_loss=0.2457, pruned_loss=0.04572, over 16540.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.259, pruned_loss=0.04438, over 2753434.96 frames. ], batch size: 146, lr: 3.42e-03, grad_scale: 1.0 2023-05-01 00:23:39,626 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:24:37,895 INFO [train.py:904] (0/8) Epoch 20, batch 400, loss[loss=0.1653, simple_loss=0.2454, pruned_loss=0.04263, over 12615.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2576, pruned_loss=0.0437, over 2878266.39 frames. ], batch size: 246, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:25:03,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9077, 4.3368, 2.9932, 2.3684, 2.7589, 2.4928, 4.6917, 3.6215], device='cuda:0'), covar=tensor([0.2785, 0.0583, 0.1865, 0.2739, 0.2731, 0.2209, 0.0348, 0.1370], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0264, 0.0298, 0.0303, 0.0287, 0.0251, 0.0286, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 00:25:05,919 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193272.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:06,022 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193272.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:25:06,673 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.612e+02 2.085e+02 2.558e+02 2.961e+02 5.266e+02, threshold=5.116e+02, percent-clipped=1.0 2023-05-01 00:25:23,421 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193284.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:25:37,148 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1069, 2.0715, 2.2043, 3.7216, 2.1590, 2.4230, 2.2070, 2.2408], device='cuda:0'), covar=tensor([0.1464, 0.3769, 0.2995, 0.0645, 0.3993, 0.2458, 0.3688, 0.3385], device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0437, 0.0362, 0.0321, 0.0433, 0.0500, 0.0407, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:25:38,112 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8168, 4.5672, 4.8822, 5.0570, 5.2471, 4.5870, 5.2607, 5.1947], device='cuda:0'), covar=tensor([0.1897, 0.1385, 0.1676, 0.0729, 0.0523, 0.1055, 0.0610, 0.0672], device='cuda:0'), in_proj_covar=tensor([0.0615, 0.0764, 0.0892, 0.0780, 0.0582, 0.0611, 0.0632, 0.0731], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:25:41,681 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 00:25:46,663 INFO [train.py:904] (0/8) Epoch 20, batch 450, loss[loss=0.1734, simple_loss=0.2484, pruned_loss=0.0492, over 16888.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2559, pruned_loss=0.04328, over 2965869.42 frames. ], batch size: 96, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:25:53,152 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1704, 3.3085, 3.4079, 2.2437, 2.9565, 2.4476, 3.5280, 3.5383], device='cuda:0'), covar=tensor([0.0317, 0.0959, 0.0621, 0.1817, 0.0826, 0.0932, 0.0665, 0.0918], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0157, 0.0164, 0.0150, 0.0143, 0.0127, 0.0142, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:26:13,648 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:26,134 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:48,819 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:26:57,300 INFO [train.py:904] (0/8) Epoch 20, batch 500, loss[loss=0.1714, simple_loss=0.2421, pruned_loss=0.05032, over 16337.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2539, pruned_loss=0.04248, over 3045880.07 frames. ], batch size: 165, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:27:26,057 INFO [optim.py:368] (0/8) 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 00:27:31,765 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:51,491 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193391.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:27:59,430 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:07,109 INFO [train.py:904] (0/8) Epoch 20, batch 550, loss[loss=0.1662, simple_loss=0.2435, pruned_loss=0.04445, over 16779.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2534, pruned_loss=0.04234, over 3102281.84 frames. ], batch size: 102, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:28:10,525 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2619, 5.2208, 5.1312, 4.5947, 4.6775, 5.1580, 5.1788, 4.7587], device='cuda:0'), covar=tensor([0.0609, 0.0508, 0.0304, 0.0366, 0.1271, 0.0461, 0.0291, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0409, 0.0331, 0.0324, 0.0342, 0.0379, 0.0230, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:28:32,245 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:28:43,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9367, 2.6190, 2.6118, 4.8151, 3.7527, 4.2914, 1.5499, 3.0798], device='cuda:0'), covar=tensor([0.1346, 0.0903, 0.1231, 0.0225, 0.0251, 0.0431, 0.1772, 0.0789], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0171, 0.0191, 0.0182, 0.0200, 0.0212, 0.0198, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:28:44,851 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8284, 5.1551, 4.9044, 4.9116, 4.6825, 4.6693, 4.6416, 5.2221], device='cuda:0'), covar=tensor([0.1195, 0.0970, 0.1156, 0.0956, 0.0921, 0.1093, 0.1153, 0.1091], device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0793, 0.0646, 0.0590, 0.0498, 0.0504, 0.0664, 0.0608], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:28:56,791 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:29:00,579 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7553, 2.7475, 2.6220, 4.9158, 4.0500, 4.4300, 1.5117, 3.1022], device='cuda:0'), covar=tensor([0.1406, 0.0813, 0.1226, 0.0217, 0.0251, 0.0378, 0.1732, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0171, 0.0191, 0.0181, 0.0200, 0.0212, 0.0198, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:29:07,414 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9784, 3.8180, 4.2850, 2.0325, 4.5412, 4.6449, 3.3079, 3.4636], device='cuda:0'), covar=tensor([0.0690, 0.0237, 0.0216, 0.1193, 0.0070, 0.0147, 0.0390, 0.0405], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0107, 0.0095, 0.0139, 0.0078, 0.0121, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:29:14,809 INFO [train.py:904] (0/8) Epoch 20, batch 600, loss[loss=0.1678, simple_loss=0.2454, pruned_loss=0.04509, over 12466.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2526, pruned_loss=0.04197, over 3146211.57 frames. ], batch size: 247, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:29:18,192 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193454.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:29:43,213 INFO [optim.py:368] (0/8) 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 00:29:55,242 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193482.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:30:01,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6558, 2.6503, 2.1972, 2.5681, 3.0472, 2.8145, 3.2164, 3.2691], device='cuda:0'), covar=tensor([0.0171, 0.0460, 0.0542, 0.0452, 0.0311, 0.0402, 0.0341, 0.0265], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0235, 0.0224, 0.0227, 0.0236, 0.0234, 0.0234, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:30:03,407 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193487.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:30:11,731 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6016, 3.6474, 2.3139, 3.9811, 2.9218, 3.8787, 2.4797, 2.9912], device='cuda:0'), covar=tensor([0.0290, 0.0446, 0.1489, 0.0331, 0.0737, 0.0704, 0.1256, 0.0666], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0195, 0.0159, 0.0177, 0.0215, 0.0203, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:30:21,035 INFO [train.py:904] (0/8) Epoch 20, batch 650, loss[loss=0.1665, simple_loss=0.2653, pruned_loss=0.0338, over 17140.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2517, pruned_loss=0.04197, over 3186632.97 frames. ], batch size: 49, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:30:29,577 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 00:30:39,913 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193515.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:31:07,553 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:31:29,640 INFO [train.py:904] (0/8) Epoch 20, batch 700, loss[loss=0.209, simple_loss=0.2801, pruned_loss=0.06894, over 16255.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2515, pruned_loss=0.04139, over 3221333.64 frames. ], batch size: 164, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:31:42,815 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.70 vs. limit=5.0 2023-05-01 00:31:48,976 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193567.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 00:31:57,449 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.109e+02 2.476e+02 3.086e+02 5.855e+02, threshold=4.951e+02, percent-clipped=1.0 2023-05-01 00:32:08,558 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9656, 1.7727, 2.4887, 2.8453, 2.8678, 2.9769, 1.7971, 3.0851], device='cuda:0'), covar=tensor([0.0150, 0.0547, 0.0284, 0.0230, 0.0229, 0.0190, 0.0648, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0190, 0.0175, 0.0178, 0.0191, 0.0149, 0.0191, 0.0142], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:32:35,684 INFO [train.py:904] (0/8) Epoch 20, batch 750, loss[loss=0.1672, simple_loss=0.2623, pruned_loss=0.03609, over 17089.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.252, pruned_loss=0.0418, over 3252772.61 frames. ], batch size: 53, lr: 3.42e-03, grad_scale: 2.0 2023-05-01 00:32:54,621 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:15,603 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2634, 5.1897, 5.1003, 4.5694, 4.7127, 5.1569, 5.1568, 4.7540], device='cuda:0'), covar=tensor([0.0588, 0.0518, 0.0297, 0.0365, 0.1120, 0.0461, 0.0300, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0416, 0.0337, 0.0330, 0.0347, 0.0386, 0.0233, 0.0403], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:33:24,993 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5272, 4.4788, 4.4294, 3.8897, 4.4949, 1.8094, 4.2196, 4.0737], device='cuda:0'), covar=tensor([0.0125, 0.0100, 0.0193, 0.0315, 0.0100, 0.2638, 0.0145, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0148, 0.0192, 0.0171, 0.0168, 0.0204, 0.0181, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:33:28,443 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:33:42,948 INFO [train.py:904] (0/8) Epoch 20, batch 800, loss[loss=0.1993, simple_loss=0.2727, pruned_loss=0.06296, over 12195.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.252, pruned_loss=0.04157, over 3266465.50 frames. ], batch size: 246, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:34:10,662 INFO [optim.py:368] (0/8) 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 00:34:16,593 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:45,094 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:34:52,173 INFO [train.py:904] (0/8) Epoch 20, batch 850, loss[loss=0.1575, simple_loss=0.2358, pruned_loss=0.03966, over 16885.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2514, pruned_loss=0.04139, over 3278928.29 frames. ], batch size: 109, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:35:02,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1452, 4.1486, 4.3967, 2.4396, 4.7580, 4.8045, 3.4897, 3.7599], device='cuda:0'), covar=tensor([0.0703, 0.0237, 0.0286, 0.1066, 0.0061, 0.0162, 0.0397, 0.0362], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0107, 0.0095, 0.0138, 0.0078, 0.0121, 0.0126, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:35:20,280 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8862, 1.8548, 2.4754, 2.7624, 2.7121, 3.0879, 2.2948, 3.0581], device='cuda:0'), covar=tensor([0.0216, 0.0526, 0.0346, 0.0325, 0.0334, 0.0236, 0.0440, 0.0170], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0191, 0.0176, 0.0178, 0.0192, 0.0149, 0.0192, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:35:50,835 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:35:59,342 INFO [train.py:904] (0/8) Epoch 20, batch 900, loss[loss=0.1734, simple_loss=0.2769, pruned_loss=0.03495, over 17061.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.251, pruned_loss=0.04078, over 3289468.11 frames. ], batch size: 55, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:35:59,917 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 00:36:16,400 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3904, 2.9451, 2.6558, 2.2705, 2.2680, 2.3258, 2.9753, 2.7935], device='cuda:0'), covar=tensor([0.2541, 0.0926, 0.1715, 0.2434, 0.2321, 0.2101, 0.0550, 0.1444], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0265, 0.0299, 0.0303, 0.0290, 0.0252, 0.0287, 0.0330], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 00:36:28,231 INFO [optim.py:368] (0/8) 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 00:36:33,736 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193777.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:36:35,630 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-05-01 00:36:56,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8582, 4.5379, 4.7092, 5.0625, 5.2438, 4.6908, 5.3507, 5.2073], device='cuda:0'), covar=tensor([0.1887, 0.1620, 0.2468, 0.1068, 0.0827, 0.0895, 0.0701, 0.0912], device='cuda:0'), in_proj_covar=tensor([0.0636, 0.0789, 0.0922, 0.0805, 0.0601, 0.0627, 0.0651, 0.0753], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:37:09,304 INFO [train.py:904] (0/8) Epoch 20, batch 950, loss[loss=0.1664, simple_loss=0.2468, pruned_loss=0.04305, over 16857.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2511, pruned_loss=0.04073, over 3297832.66 frames. ], batch size: 90, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:37:20,483 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193810.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:37:36,588 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-05-01 00:38:17,868 INFO [train.py:904] (0/8) Epoch 20, batch 1000, loss[loss=0.1923, simple_loss=0.2626, pruned_loss=0.06102, over 11819.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2498, pruned_loss=0.04065, over 3300870.32 frames. ], batch size: 248, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:38:39,205 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193867.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:38:45,943 INFO [optim.py:368] (0/8) 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 00:39:00,662 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=193884.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:39:24,786 INFO [train.py:904] (0/8) Epoch 20, batch 1050, loss[loss=0.1693, simple_loss=0.2628, pruned_loss=0.03788, over 17039.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2494, pruned_loss=0.04048, over 3307240.59 frames. ], batch size: 55, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:39:43,674 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:18,218 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=193940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:25,145 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=193945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:40:35,475 INFO [train.py:904] (0/8) Epoch 20, batch 1100, loss[loss=0.1592, simple_loss=0.2367, pruned_loss=0.04088, over 16402.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2489, pruned_loss=0.04024, over 3304747.15 frames. ], batch size: 146, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:40:54,788 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0072, 3.0873, 3.3998, 2.1326, 2.8822, 2.2530, 3.4872, 3.4441], device='cuda:0'), covar=tensor([0.0231, 0.0937, 0.0538, 0.1922, 0.0871, 0.0988, 0.0541, 0.0870], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0160, 0.0166, 0.0152, 0.0144, 0.0128, 0.0144, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:41:01,700 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=193971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:04,220 INFO [optim.py:368] (0/8) 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 00:41:13,819 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9565, 2.6575, 2.2131, 2.4880, 3.0493, 2.8220, 3.0389, 3.1459], device='cuda:0'), covar=tensor([0.0215, 0.0330, 0.0437, 0.0380, 0.0189, 0.0310, 0.0214, 0.0222], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0237, 0.0225, 0.0229, 0.0239, 0.0237, 0.0237, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:41:24,332 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=193988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:41:30,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3909, 4.4476, 4.7931, 4.8131, 4.8183, 4.5191, 4.5280, 4.4118], device='cuda:0'), covar=tensor([0.0392, 0.0740, 0.0461, 0.0393, 0.0460, 0.0467, 0.0849, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0444, 0.0429, 0.0401, 0.0480, 0.0451, 0.0544, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:41:40,694 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-194000.pt 2023-05-01 00:41:46,958 INFO [train.py:904] (0/8) Epoch 20, batch 1150, loss[loss=0.158, simple_loss=0.2433, pruned_loss=0.03636, over 15917.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.249, pruned_loss=0.04022, over 3313535.96 frames. ], batch size: 35, lr: 3.42e-03, grad_scale: 4.0 2023-05-01 00:41:57,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9721, 5.0643, 5.4707, 5.5009, 5.4612, 5.1654, 5.0666, 4.8813], device='cuda:0'), covar=tensor([0.0338, 0.0588, 0.0451, 0.0415, 0.0528, 0.0393, 0.0941, 0.0461], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0444, 0.0429, 0.0402, 0.0481, 0.0451, 0.0545, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:42:13,015 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3185, 4.3830, 4.6913, 4.6989, 4.7277, 4.4647, 4.4533, 4.3356], device='cuda:0'), covar=tensor([0.0380, 0.0759, 0.0502, 0.0450, 0.0470, 0.0446, 0.0799, 0.0574], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0445, 0.0430, 0.0402, 0.0482, 0.0452, 0.0546, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:42:46,070 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8329, 2.8846, 2.9024, 4.8473, 3.8860, 4.2799, 1.7424, 3.0745], device='cuda:0'), covar=tensor([0.1376, 0.0814, 0.1142, 0.0209, 0.0248, 0.0419, 0.1612, 0.0830], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0184, 0.0202, 0.0215, 0.0199, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:42:56,056 INFO [train.py:904] (0/8) Epoch 20, batch 1200, loss[loss=0.146, simple_loss=0.24, pruned_loss=0.02602, over 17102.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2491, pruned_loss=0.03986, over 3321726.92 frames. ], batch size: 47, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:43:09,410 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4723, 4.5173, 4.7876, 4.8033, 4.8526, 4.5707, 4.5665, 4.4025], device='cuda:0'), covar=tensor([0.0446, 0.1038, 0.0711, 0.0557, 0.0491, 0.0571, 0.0989, 0.0670], device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0447, 0.0432, 0.0405, 0.0484, 0.0454, 0.0549, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 00:43:20,751 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 00:43:25,095 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:43:25,795 INFO [optim.py:368] (0/8) 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 00:43:33,043 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194077.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 00:44:06,936 INFO [train.py:904] (0/8) Epoch 20, batch 1250, loss[loss=0.1485, simple_loss=0.2257, pruned_loss=0.0356, over 16441.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2488, pruned_loss=0.03963, over 3320699.74 frames. ], batch size: 146, lr: 3.42e-03, grad_scale: 8.0 2023-05-01 00:44:16,938 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194110.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:44:38,151 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:44:38,885 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-05-01 00:44:49,868 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:45:15,933 INFO [train.py:904] (0/8) Epoch 20, batch 1300, loss[loss=0.173, simple_loss=0.2472, pruned_loss=0.0494, over 16949.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2492, pruned_loss=0.04019, over 3324025.05 frames. ], batch size: 109, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:45:26,572 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194158.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:45:46,647 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.297e+02 2.207e+02 2.541e+02 3.000e+02 4.873e+02, threshold=5.083e+02, percent-clipped=0.0 2023-05-01 00:46:27,325 INFO [train.py:904] (0/8) Epoch 20, batch 1350, loss[loss=0.1582, simple_loss=0.2456, pruned_loss=0.03545, over 17109.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2485, pruned_loss=0.03958, over 3325865.28 frames. ], batch size: 48, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:46:59,951 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0774, 1.9989, 1.6117, 1.7833, 2.2666, 1.9787, 2.0011, 2.3798], device='cuda:0'), covar=tensor([0.0350, 0.0476, 0.0602, 0.0510, 0.0278, 0.0411, 0.0263, 0.0292], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0237, 0.0225, 0.0229, 0.0239, 0.0237, 0.0237, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:47:15,495 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3036, 5.2062, 5.1280, 4.6115, 4.7035, 5.2014, 5.1275, 4.8134], device='cuda:0'), covar=tensor([0.0616, 0.0526, 0.0319, 0.0360, 0.1282, 0.0465, 0.0377, 0.0753], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0424, 0.0344, 0.0337, 0.0355, 0.0395, 0.0238, 0.0411], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:47:20,520 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194240.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:47:36,953 INFO [train.py:904] (0/8) Epoch 20, batch 1400, loss[loss=0.1544, simple_loss=0.2513, pruned_loss=0.02873, over 17128.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2482, pruned_loss=0.03938, over 3325646.84 frames. ], batch size: 47, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:47:40,302 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1247, 3.9580, 4.2847, 2.2456, 4.5657, 4.6613, 3.2553, 3.6310], device='cuda:0'), covar=tensor([0.0690, 0.0302, 0.0275, 0.1222, 0.0089, 0.0177, 0.0452, 0.0408], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0109, 0.0097, 0.0140, 0.0079, 0.0124, 0.0128, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:48:02,766 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194271.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:48:06,434 INFO [optim.py:368] (0/8) 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 00:48:44,308 INFO [train.py:904] (0/8) Epoch 20, batch 1450, loss[loss=0.1607, simple_loss=0.2616, pruned_loss=0.02991, over 17119.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.248, pruned_loss=0.03954, over 3333304.45 frames. ], batch size: 48, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:06,294 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194318.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:07,986 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:49:23,149 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 00:49:54,002 INFO [train.py:904] (0/8) Epoch 20, batch 1500, loss[loss=0.1573, simple_loss=0.2397, pruned_loss=0.03745, over 12064.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2484, pruned_loss=0.04009, over 3330952.93 frames. ], batch size: 246, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:49:55,472 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:50:07,408 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 00:50:31,259 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:50:36,370 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-01 00:51:03,663 INFO [train.py:904] (0/8) Epoch 20, batch 1550, loss[loss=0.1619, simple_loss=0.2572, pruned_loss=0.03328, over 17128.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2492, pruned_loss=0.0405, over 3323507.24 frames. ], batch size: 49, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:51:20,893 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:51:34,183 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194423.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 00:51:38,880 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7752, 3.8526, 2.4632, 4.4691, 2.7946, 4.3942, 2.6273, 3.1617], device='cuda:0'), covar=tensor([0.0303, 0.0362, 0.1572, 0.0263, 0.0904, 0.0449, 0.1389, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0180, 0.0198, 0.0165, 0.0180, 0.0220, 0.0205, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:51:40,593 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:52:13,199 INFO [train.py:904] (0/8) Epoch 20, batch 1600, loss[loss=0.1403, simple_loss=0.2295, pruned_loss=0.02555, over 16830.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2504, pruned_loss=0.04086, over 3333687.01 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:52:43,591 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 00:52:43,862 INFO [optim.py:368] (0/8) 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 00:53:22,708 INFO [train.py:904] (0/8) Epoch 20, batch 1650, loss[loss=0.1428, simple_loss=0.2267, pruned_loss=0.02945, over 17003.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2516, pruned_loss=0.04132, over 3337484.91 frames. ], batch size: 41, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 00:53:29,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2578, 3.2343, 1.8476, 3.4400, 2.5057, 3.4314, 1.8598, 2.6039], device='cuda:0'), covar=tensor([0.0283, 0.0431, 0.1778, 0.0318, 0.0781, 0.0612, 0.1770, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0180, 0.0197, 0.0164, 0.0179, 0.0220, 0.0205, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 00:53:47,347 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7958, 4.1245, 4.2655, 2.9364, 3.6775, 4.2183, 3.7501, 2.5353], device='cuda:0'), covar=tensor([0.0463, 0.0078, 0.0049, 0.0367, 0.0125, 0.0098, 0.0100, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0082, 0.0082, 0.0134, 0.0096, 0.0108, 0.0093, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:0') 2023-05-01 00:53:57,789 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.96 vs. limit=5.0 2023-05-01 00:54:16,606 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194540.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:54:33,474 INFO [train.py:904] (0/8) Epoch 20, batch 1700, loss[loss=0.1583, simple_loss=0.2508, pruned_loss=0.03288, over 17197.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2537, pruned_loss=0.04201, over 3334918.39 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:55:05,702 INFO [optim.py:368] (0/8) 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 00:55:25,136 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194588.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:55:43,641 INFO [train.py:904] (0/8) Epoch 20, batch 1750, loss[loss=0.1699, simple_loss=0.2507, pruned_loss=0.04452, over 16874.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2544, pruned_loss=0.04172, over 3338765.79 frames. ], batch size: 116, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:56:23,128 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-01 00:56:27,223 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194634.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:56:51,748 INFO [train.py:904] (0/8) Epoch 20, batch 1800, loss[loss=0.1699, simple_loss=0.2538, pruned_loss=0.04303, over 16722.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2561, pruned_loss=0.04216, over 3341311.13 frames. ], batch size: 89, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:57:22,483 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 00:57:50,506 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194695.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:57:58,020 INFO [train.py:904] (0/8) Epoch 20, batch 1850, loss[loss=0.1855, simple_loss=0.2684, pruned_loss=0.05134, over 16304.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2564, pruned_loss=0.04203, over 3345413.68 frames. ], batch size: 165, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:58:08,656 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:21,460 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194718.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 00:58:33,800 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 00:58:39,913 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 00:59:06,820 INFO [train.py:904] (0/8) Epoch 20, batch 1900, loss[loss=0.1775, simple_loss=0.2529, pruned_loss=0.051, over 16732.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2559, pruned_loss=0.04162, over 3342069.87 frames. ], batch size: 134, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 00:59:15,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3085, 5.6692, 5.4246, 5.5417, 5.1273, 5.1138, 5.1244, 5.8166], device='cuda:0'), covar=tensor([0.1380, 0.0922, 0.1088, 0.0787, 0.0893, 0.0774, 0.1164, 0.0952], device='cuda:0'), in_proj_covar=tensor([0.0671, 0.0828, 0.0675, 0.0616, 0.0518, 0.0525, 0.0693, 0.0635], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:59:37,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0293, 4.9992, 4.7689, 4.2447, 4.8950, 1.8348, 4.6081, 4.6406], device='cuda:0'), covar=tensor([0.0098, 0.0092, 0.0218, 0.0388, 0.0105, 0.2904, 0.0153, 0.0218], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0152, 0.0197, 0.0177, 0.0174, 0.0208, 0.0187, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 00:59:38,388 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.030e+02 2.433e+02 3.010e+02 6.888e+02, threshold=4.866e+02, percent-clipped=1.0 2023-05-01 00:59:41,139 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=194776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:00:16,334 INFO [train.py:904] (0/8) Epoch 20, batch 1950, loss[loss=0.1659, simple_loss=0.2578, pruned_loss=0.03701, over 16592.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.256, pruned_loss=0.04138, over 3339167.12 frames. ], batch size: 68, lr: 3.41e-03, grad_scale: 4.0 2023-05-01 01:01:17,150 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=194847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:01:23,591 INFO [train.py:904] (0/8) Epoch 20, batch 2000, loss[loss=0.1411, simple_loss=0.2187, pruned_loss=0.03181, over 16822.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2558, pruned_loss=0.0416, over 3331933.14 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:01:54,986 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.555e+02 2.120e+02 2.565e+02 3.052e+02 4.628e+02, threshold=5.130e+02, percent-clipped=0.0 2023-05-01 01:02:32,415 INFO [train.py:904] (0/8) Epoch 20, batch 2050, loss[loss=0.1511, simple_loss=0.2516, pruned_loss=0.02529, over 17147.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2565, pruned_loss=0.0421, over 3319334.97 frames. ], batch size: 48, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:02:38,205 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3923, 4.2782, 4.2675, 4.0232, 4.0673, 4.3302, 4.0059, 4.1381], device='cuda:0'), covar=tensor([0.0627, 0.0880, 0.0306, 0.0255, 0.0655, 0.0497, 0.0840, 0.0564], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0430, 0.0348, 0.0342, 0.0359, 0.0401, 0.0240, 0.0418], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:02:41,414 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=194908.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:03:25,482 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.55 vs. limit=5.0 2023-05-01 01:03:36,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4134, 3.5077, 3.9148, 1.9851, 3.0754, 2.4822, 3.7540, 3.7569], device='cuda:0'), covar=tensor([0.0267, 0.0969, 0.0520, 0.2108, 0.0839, 0.0984, 0.0631, 0.1085], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0162, 0.0166, 0.0152, 0.0144, 0.0128, 0.0144, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:03:41,681 INFO [train.py:904] (0/8) Epoch 20, batch 2100, loss[loss=0.1669, simple_loss=0.2553, pruned_loss=0.03931, over 17201.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2579, pruned_loss=0.04263, over 3317173.83 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:04:12,901 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=194974.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:13,685 INFO [optim.py:368] (0/8) 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 01:04:23,264 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0195, 4.8003, 5.0333, 5.2610, 5.4617, 4.7479, 5.4049, 5.4326], device='cuda:0'), covar=tensor([0.2105, 0.1425, 0.2035, 0.0922, 0.0710, 0.0945, 0.0714, 0.0704], device='cuda:0'), in_proj_covar=tensor([0.0655, 0.0810, 0.0948, 0.0832, 0.0620, 0.0651, 0.0667, 0.0774], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:04:34,704 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=194990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:04:50,815 INFO [train.py:904] (0/8) Epoch 20, batch 2150, loss[loss=0.1453, simple_loss=0.2405, pruned_loss=0.02511, over 17184.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2581, pruned_loss=0.04236, over 3318312.72 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:05:01,390 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:03,785 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:13,284 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195018.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:05:18,552 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:05:38,274 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 01:05:49,687 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195044.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:06:01,915 INFO [train.py:904] (0/8) Epoch 20, batch 2200, loss[loss=0.1468, simple_loss=0.2267, pruned_loss=0.03345, over 16774.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2594, pruned_loss=0.04351, over 3301522.82 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:06:09,189 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:20,444 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195066.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:06:28,285 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:06:30,117 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195072.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 01:06:33,767 INFO [optim.py:368] (0/8) 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 01:07:10,873 INFO [train.py:904] (0/8) Epoch 20, batch 2250, loss[loss=0.1931, simple_loss=0.2687, pruned_loss=0.0587, over 16829.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2602, pruned_loss=0.0445, over 3296325.09 frames. ], batch size: 116, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:07:15,436 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195105.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 01:07:54,109 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:08:21,153 INFO [train.py:904] (0/8) Epoch 20, batch 2300, loss[loss=0.1468, simple_loss=0.2315, pruned_loss=0.03106, over 16802.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2605, pruned_loss=0.04415, over 3303541.62 frames. ], batch size: 39, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:08:51,385 INFO [optim.py:368] (0/8) 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 01:09:29,611 INFO [train.py:904] (0/8) Epoch 20, batch 2350, loss[loss=0.2002, simple_loss=0.273, pruned_loss=0.06373, over 16760.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2599, pruned_loss=0.04394, over 3318401.19 frames. ], batch size: 124, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:09:31,140 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:10:36,702 INFO [train.py:904] (0/8) Epoch 20, batch 2400, loss[loss=0.1547, simple_loss=0.2418, pruned_loss=0.03379, over 17187.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2602, pruned_loss=0.04387, over 3312721.81 frames. ], batch size: 46, lr: 3.41e-03, grad_scale: 8.0 2023-05-01 01:11:07,015 INFO [optim.py:368] (0/8) 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 01:11:29,101 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:11:45,817 INFO [train.py:904] (0/8) Epoch 20, batch 2450, loss[loss=0.1857, simple_loss=0.2612, pruned_loss=0.05514, over 16749.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2606, pruned_loss=0.04372, over 3313281.64 frames. ], batch size: 124, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:12:35,688 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:12:54,088 INFO [train.py:904] (0/8) Epoch 20, batch 2500, loss[loss=0.1825, simple_loss=0.2573, pruned_loss=0.05387, over 16777.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2611, pruned_loss=0.04428, over 3310452.64 frames. ], batch size: 124, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:13:09,195 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 01:13:15,895 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195367.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:13:26,928 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.442e+02 2.185e+02 2.562e+02 3.023e+02 6.708e+02, threshold=5.124e+02, percent-clipped=4.0 2023-05-01 01:13:33,684 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0035, 3.0014, 2.6809, 2.8989, 3.2865, 3.0030, 3.5981, 3.4679], device='cuda:0'), covar=tensor([0.0133, 0.0384, 0.0480, 0.0370, 0.0265, 0.0346, 0.0273, 0.0246], device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0241, 0.0227, 0.0230, 0.0241, 0.0240, 0.0241, 0.0234], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:13:50,925 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=195392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:14:01,274 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195400.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 01:14:04,078 INFO [train.py:904] (0/8) Epoch 20, batch 2550, loss[loss=0.1815, simple_loss=0.2691, pruned_loss=0.04696, over 11930.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2608, pruned_loss=0.04419, over 3312510.14 frames. ], batch size: 246, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:14:38,093 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:14:47,513 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0041, 5.0124, 5.4201, 5.4155, 5.4292, 5.0978, 5.0529, 4.7728], device='cuda:0'), covar=tensor([0.0298, 0.0470, 0.0357, 0.0392, 0.0445, 0.0393, 0.0892, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0444, 0.0430, 0.0404, 0.0478, 0.0455, 0.0544, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 01:14:48,000 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-05-01 01:14:57,678 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0045, 4.0779, 2.6856, 4.7826, 3.2681, 4.6880, 2.7726, 3.4127], device='cuda:0'), covar=tensor([0.0275, 0.0386, 0.1444, 0.0231, 0.0727, 0.0476, 0.1410, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0181, 0.0197, 0.0166, 0.0179, 0.0222, 0.0205, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:15:11,783 INFO [train.py:904] (0/8) Epoch 20, batch 2600, loss[loss=0.1846, simple_loss=0.2734, pruned_loss=0.0479, over 16759.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2609, pruned_loss=0.04384, over 3321039.34 frames. ], batch size: 124, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:15:13,326 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=195453.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:15:27,762 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5851, 2.4381, 1.9918, 2.3264, 2.8393, 2.6313, 3.1411, 3.1487], device='cuda:0'), covar=tensor([0.0179, 0.0551, 0.0705, 0.0554, 0.0345, 0.0489, 0.0340, 0.0303], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0242, 0.0228, 0.0231, 0.0242, 0.0241, 0.0242, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:15:42,991 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.224e+02 2.625e+02 3.381e+02 7.141e+02, threshold=5.251e+02, percent-clipped=4.0 2023-05-01 01:16:20,747 INFO [train.py:904] (0/8) Epoch 20, batch 2650, loss[loss=0.1628, simple_loss=0.2562, pruned_loss=0.03469, over 17219.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.261, pruned_loss=0.04314, over 3318114.21 frames. ], batch size: 45, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:16:22,314 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:28,732 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:17:29,631 INFO [train.py:904] (0/8) Epoch 20, batch 2700, loss[loss=0.1847, simple_loss=0.2635, pruned_loss=0.05291, over 16723.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.2607, pruned_loss=0.04229, over 3322200.31 frames. ], batch size: 124, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:18:00,664 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.486e+02 2.029e+02 2.571e+02 3.039e+02 8.808e+02, threshold=5.142e+02, percent-clipped=4.0 2023-05-01 01:18:02,905 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 01:18:25,811 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3373, 5.2761, 5.1750, 4.7039, 4.8160, 5.2314, 5.1569, 4.8393], device='cuda:0'), covar=tensor([0.0612, 0.0548, 0.0275, 0.0353, 0.1056, 0.0455, 0.0326, 0.0791], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0433, 0.0350, 0.0345, 0.0362, 0.0402, 0.0242, 0.0419], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 01:18:39,452 INFO [train.py:904] (0/8) Epoch 20, batch 2750, loss[loss=0.1708, simple_loss=0.2618, pruned_loss=0.03996, over 16680.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2606, pruned_loss=0.04182, over 3330960.49 frames. ], batch size: 76, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:19:47,496 INFO [train.py:904] (0/8) Epoch 20, batch 2800, loss[loss=0.1754, simple_loss=0.2701, pruned_loss=0.04042, over 17078.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2607, pruned_loss=0.04152, over 3334477.25 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:19:56,955 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 01:19:57,676 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6592, 3.4992, 2.8485, 2.2360, 2.3336, 2.3637, 3.6707, 3.1799], device='cuda:0'), covar=tensor([0.2467, 0.0757, 0.1600, 0.2791, 0.2594, 0.1964, 0.0518, 0.1516], device='cuda:0'), in_proj_covar=tensor([0.0322, 0.0268, 0.0301, 0.0305, 0.0295, 0.0254, 0.0291, 0.0335], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 01:20:06,728 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195667.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 01:20:18,616 INFO [optim.py:368] (0/8) 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 01:20:52,394 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195700.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:20:54,307 INFO [train.py:904] (0/8) Epoch 20, batch 2850, loss[loss=0.1778, simple_loss=0.2752, pruned_loss=0.04023, over 17238.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2599, pruned_loss=0.04185, over 3330610.05 frames. ], batch size: 52, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:21:13,264 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195715.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:29,069 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=195727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:56,487 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195748.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 01:21:56,497 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=195748.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:21:58,487 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 01:22:02,489 INFO [train.py:904] (0/8) Epoch 20, batch 2900, loss[loss=0.1709, simple_loss=0.2699, pruned_loss=0.03598, over 17232.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.259, pruned_loss=0.0422, over 3332891.88 frames. ], batch size: 44, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:22:33,093 INFO [optim.py:368] (0/8) 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 01:22:33,334 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=195775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:23:10,991 INFO [train.py:904] (0/8) Epoch 20, batch 2950, loss[loss=0.1742, simple_loss=0.2702, pruned_loss=0.03906, over 17069.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2587, pruned_loss=0.04296, over 3336799.13 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,961 INFO [train.py:904] (0/8) Epoch 20, batch 3000, loss[loss=0.1712, simple_loss=0.2573, pruned_loss=0.04255, over 17062.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2585, pruned_loss=0.04312, over 3345716.23 frames. ], batch size: 53, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:24:18,962 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 01:24:27,132 INFO [train.py:938] (0/8) Epoch 20, validation: loss=0.1354, simple_loss=0.2409, pruned_loss=0.01492, over 944034.00 frames. 2023-05-01 01:24:27,132 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 01:24:58,716 INFO [optim.py:368] (0/8) 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 01:25:24,680 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8910, 4.9955, 5.3923, 5.3900, 5.3522, 5.0812, 5.0212, 4.8252], device='cuda:0'), covar=tensor([0.0307, 0.0427, 0.0342, 0.0349, 0.0505, 0.0353, 0.0855, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0455, 0.0440, 0.0411, 0.0486, 0.0464, 0.0556, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 01:25:38,247 INFO [train.py:904] (0/8) Epoch 20, batch 3050, loss[loss=0.1891, simple_loss=0.2659, pruned_loss=0.05616, over 16654.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2583, pruned_loss=0.04318, over 3345666.25 frames. ], batch size: 134, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:26:46,792 INFO [train.py:904] (0/8) Epoch 20, batch 3100, loss[loss=0.1605, simple_loss=0.253, pruned_loss=0.03404, over 16599.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2582, pruned_loss=0.04297, over 3341186.21 frames. ], batch size: 62, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:27:16,967 INFO [optim.py:368] (0/8) 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 01:27:43,675 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5273, 2.3007, 2.3325, 4.3494, 2.2824, 2.7152, 2.3813, 2.5001], device='cuda:0'), covar=tensor([0.1272, 0.3675, 0.3140, 0.0513, 0.4307, 0.2679, 0.3576, 0.3744], device='cuda:0'), in_proj_covar=tensor([0.0400, 0.0442, 0.0365, 0.0328, 0.0435, 0.0510, 0.0413, 0.0518], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:27:49,989 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-196000.pt 2023-05-01 01:27:54,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8365, 3.9151, 2.7488, 2.4104, 2.5368, 2.3236, 3.9666, 3.3327], device='cuda:0'), covar=tensor([0.2524, 0.0638, 0.1959, 0.2661, 0.2793, 0.2293, 0.0587, 0.1522], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0271, 0.0304, 0.0308, 0.0299, 0.0256, 0.0294, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 01:27:55,520 INFO [train.py:904] (0/8) Epoch 20, batch 3150, loss[loss=0.1645, simple_loss=0.2453, pruned_loss=0.04183, over 16827.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2572, pruned_loss=0.0428, over 3338713.96 frames. ], batch size: 90, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:28:08,786 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6940, 2.5282, 1.9467, 2.2594, 2.8704, 2.5901, 3.3983, 3.1663], device='cuda:0'), covar=tensor([0.0186, 0.0536, 0.0749, 0.0621, 0.0396, 0.0509, 0.0241, 0.0356], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0240, 0.0228, 0.0231, 0.0241, 0.0239, 0.0243, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:28:57,143 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:29:03,233 INFO [train.py:904] (0/8) Epoch 20, batch 3200, loss[loss=0.1853, simple_loss=0.2641, pruned_loss=0.05329, over 16753.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2564, pruned_loss=0.04244, over 3340482.92 frames. ], batch size: 102, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:29:33,001 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2255, 5.1806, 4.9518, 4.4907, 5.0163, 1.9603, 4.7465, 4.9115], device='cuda:0'), covar=tensor([0.0077, 0.0068, 0.0210, 0.0392, 0.0114, 0.2731, 0.0145, 0.0195], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0155, 0.0201, 0.0181, 0.0177, 0.0210, 0.0190, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:29:35,522 INFO [optim.py:368] (0/8) 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 01:30:04,480 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=196096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:30:11,579 INFO [train.py:904] (0/8) Epoch 20, batch 3250, loss[loss=0.1825, simple_loss=0.2716, pruned_loss=0.0467, over 17252.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2558, pruned_loss=0.04241, over 3342942.14 frames. ], batch size: 52, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:19,987 INFO [train.py:904] (0/8) Epoch 20, batch 3300, loss[loss=0.1699, simple_loss=0.2667, pruned_loss=0.03653, over 17101.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2572, pruned_loss=0.04252, over 3335246.78 frames. ], batch size: 49, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:31:20,657 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.12 vs. limit=5.0 2023-05-01 01:31:27,266 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-01 01:31:52,355 INFO [optim.py:368] (0/8) 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 01:32:16,640 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6494, 4.6066, 4.5607, 4.0033, 4.5991, 1.7758, 4.3557, 4.2611], device='cuda:0'), covar=tensor([0.0107, 0.0088, 0.0175, 0.0331, 0.0099, 0.2732, 0.0140, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0153, 0.0199, 0.0179, 0.0176, 0.0208, 0.0189, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:32:28,250 INFO [train.py:904] (0/8) Epoch 20, batch 3350, loss[loss=0.1628, simple_loss=0.2654, pruned_loss=0.0301, over 17282.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2567, pruned_loss=0.042, over 3337156.85 frames. ], batch size: 52, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:33:35,779 INFO [train.py:904] (0/8) Epoch 20, batch 3400, loss[loss=0.1549, simple_loss=0.2527, pruned_loss=0.0286, over 17183.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2567, pruned_loss=0.04209, over 3326311.98 frames. ], batch size: 46, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:33:51,896 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7694, 2.4761, 2.5130, 4.6486, 2.4354, 2.9963, 2.5981, 2.7596], device='cuda:0'), covar=tensor([0.1165, 0.3704, 0.2953, 0.0487, 0.4243, 0.2559, 0.3403, 0.3634], device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0443, 0.0366, 0.0330, 0.0437, 0.0512, 0.0413, 0.0520], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:33:57,231 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 01:34:06,839 INFO [optim.py:368] (0/8) 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 01:34:44,350 INFO [train.py:904] (0/8) Epoch 20, batch 3450, loss[loss=0.18, simple_loss=0.2515, pruned_loss=0.05423, over 16724.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2553, pruned_loss=0.04186, over 3330871.21 frames. ], batch size: 134, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:35:10,303 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:35:50,470 INFO [train.py:904] (0/8) Epoch 20, batch 3500, loss[loss=0.1644, simple_loss=0.2606, pruned_loss=0.03407, over 17122.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2543, pruned_loss=0.04187, over 3329509.46 frames. ], batch size: 49, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:36:03,314 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 01:36:25,524 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7569, 1.8840, 2.3295, 2.6264, 2.6548, 2.6862, 1.8589, 2.8756], device='cuda:0'), covar=tensor([0.0167, 0.0475, 0.0334, 0.0274, 0.0309, 0.0284, 0.0544, 0.0164], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0193, 0.0178, 0.0183, 0.0196, 0.0154, 0.0195, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:36:35,853 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:37:01,680 INFO [train.py:904] (0/8) Epoch 20, batch 3550, loss[loss=0.1539, simple_loss=0.23, pruned_loss=0.03889, over 16831.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2531, pruned_loss=0.04115, over 3331105.49 frames. ], batch size: 96, lr: 3.40e-03, grad_scale: 8.0 2023-05-01 01:37:15,827 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 01:37:27,886 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196421.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:38:06,505 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 01:38:10,214 INFO [train.py:904] (0/8) Epoch 20, batch 3600, loss[loss=0.1591, simple_loss=0.2561, pruned_loss=0.03103, over 17278.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2526, pruned_loss=0.04073, over 3332087.17 frames. ], batch size: 52, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:38:41,983 INFO [optim.py:368] (0/8) 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 01:39:20,698 INFO [train.py:904] (0/8) Epoch 20, batch 3650, loss[loss=0.1559, simple_loss=0.2339, pruned_loss=0.03896, over 16724.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2514, pruned_loss=0.04156, over 3322986.14 frames. ], batch size: 134, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 01:40:08,893 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 01:40:32,648 INFO [train.py:904] (0/8) Epoch 20, batch 3700, loss[loss=0.1605, simple_loss=0.2379, pruned_loss=0.04152, over 16834.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2505, pruned_loss=0.04316, over 3295439.75 frames. ], batch size: 116, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:41:07,137 INFO [optim.py:368] (0/8) 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 01:41:11,812 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:41:47,094 INFO [train.py:904] (0/8) Epoch 20, batch 3750, loss[loss=0.1672, simple_loss=0.2374, pruned_loss=0.04855, over 16818.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2508, pruned_loss=0.04465, over 3284731.46 frames. ], batch size: 96, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:41:54,768 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3443, 3.6319, 3.8177, 2.5761, 3.4711, 3.9250, 3.6174, 2.1778], device='cuda:0'), covar=tensor([0.0486, 0.0103, 0.0047, 0.0351, 0.0095, 0.0074, 0.0081, 0.0417], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0082, 0.0081, 0.0133, 0.0096, 0.0107, 0.0093, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 01:42:38,988 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:42:58,332 INFO [train.py:904] (0/8) Epoch 20, batch 3800, loss[loss=0.1918, simple_loss=0.2612, pruned_loss=0.06119, over 16868.00 frames. ], tot_loss[loss=0.1723, simple_loss=0.2525, pruned_loss=0.04601, over 3280595.79 frames. ], batch size: 109, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:43:31,148 INFO [optim.py:368] (0/8) 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 01:43:31,724 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3876, 3.4631, 3.6022, 2.2480, 3.0705, 2.4952, 3.8664, 3.8772], device='cuda:0'), covar=tensor([0.0200, 0.0777, 0.0561, 0.1874, 0.0795, 0.0907, 0.0415, 0.0696], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:43:35,325 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:44:10,775 INFO [train.py:904] (0/8) Epoch 20, batch 3850, loss[loss=0.1839, simple_loss=0.2579, pruned_loss=0.05495, over 16746.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2529, pruned_loss=0.04662, over 3278171.85 frames. ], batch size: 124, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:44:32,372 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:45:24,177 INFO [train.py:904] (0/8) Epoch 20, batch 3900, loss[loss=0.1679, simple_loss=0.2487, pruned_loss=0.0435, over 16568.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2529, pruned_loss=0.04693, over 3280279.61 frames. ], batch size: 75, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:45:30,793 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7031, 2.6137, 2.6255, 4.1029, 3.3662, 4.1012, 1.6130, 2.9017], device='cuda:0'), covar=tensor([0.1358, 0.0716, 0.1041, 0.0196, 0.0159, 0.0323, 0.1510, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0173, 0.0192, 0.0187, 0.0205, 0.0215, 0.0198, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:45:52,254 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 01:45:57,454 INFO [optim.py:368] (0/8) 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 01:46:04,909 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196779.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:46:36,470 INFO [train.py:904] (0/8) Epoch 20, batch 3950, loss[loss=0.1739, simple_loss=0.2508, pruned_loss=0.04852, over 16848.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2529, pruned_loss=0.04737, over 3282677.24 frames. ], batch size: 102, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:22,796 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7235, 4.8776, 5.0519, 4.8458, 4.9232, 5.5146, 5.0513, 4.7812], device='cuda:0'), covar=tensor([0.1444, 0.2049, 0.2075, 0.2229, 0.2535, 0.1002, 0.1561, 0.2457], device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0602, 0.0661, 0.0504, 0.0666, 0.0696, 0.0514, 0.0671], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 01:47:31,592 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196840.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:48,967 INFO [train.py:904] (0/8) Epoch 20, batch 4000, loss[loss=0.1656, simple_loss=0.2465, pruned_loss=0.04237, over 16894.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2532, pruned_loss=0.04759, over 3279677.58 frames. ], batch size: 90, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:47:52,380 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:47:58,195 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 01:48:21,946 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.289e+02 2.579e+02 3.040e+02 5.791e+02, threshold=5.158e+02, percent-clipped=1.0 2023-05-01 01:48:31,799 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=196881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:01,024 INFO [train.py:904] (0/8) Epoch 20, batch 4050, loss[loss=0.1744, simple_loss=0.2631, pruned_loss=0.04283, over 16834.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2537, pruned_loss=0.04656, over 3280256.17 frames. ], batch size: 116, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:49:19,925 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:47,745 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=196934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:49:47,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0569, 5.0382, 4.8693, 4.2057, 4.9871, 1.8564, 4.7208, 4.4726], device='cuda:0'), covar=tensor([0.0066, 0.0057, 0.0159, 0.0337, 0.0072, 0.2774, 0.0116, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0153, 0.0199, 0.0180, 0.0176, 0.0208, 0.0189, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:49:57,988 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=196942.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:50:12,379 INFO [train.py:904] (0/8) Epoch 20, batch 4100, loss[loss=0.168, simple_loss=0.2548, pruned_loss=0.04058, over 17220.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2553, pruned_loss=0.04607, over 3274795.34 frames. ], batch size: 44, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:50:39,760 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 01:50:44,594 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 1.746e+02 2.102e+02 2.519e+02 4.672e+02, threshold=4.204e+02, percent-clipped=0.0 2023-05-01 01:50:47,891 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=196977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:50:55,446 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-01 01:51:16,096 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9541, 5.3032, 5.5034, 5.2073, 5.3436, 5.8895, 5.3705, 5.1141], device='cuda:0'), covar=tensor([0.0998, 0.1698, 0.1890, 0.2065, 0.2377, 0.0879, 0.1379, 0.2240], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0601, 0.0658, 0.0503, 0.0664, 0.0694, 0.0513, 0.0670], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 01:51:23,823 INFO [train.py:904] (0/8) Epoch 20, batch 4150, loss[loss=0.194, simple_loss=0.2791, pruned_loss=0.05443, over 17199.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2622, pruned_loss=0.04842, over 3236211.86 frames. ], batch size: 46, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:51:29,343 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9168, 5.2034, 5.3956, 5.2042, 5.2065, 5.7919, 5.2665, 5.0074], device='cuda:0'), covar=tensor([0.0936, 0.1780, 0.1863, 0.1953, 0.2433, 0.0833, 0.1361, 0.2345], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0600, 0.0657, 0.0502, 0.0663, 0.0693, 0.0511, 0.0669], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 01:51:45,596 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197016.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:00,051 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:52:40,722 INFO [train.py:904] (0/8) Epoch 20, batch 4200, loss[loss=0.2107, simple_loss=0.3011, pruned_loss=0.06021, over 16901.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2696, pruned_loss=0.05045, over 3202956.03 frames. ], batch size: 109, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:52:58,768 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:53:12,973 INFO [optim.py:368] (0/8) 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 01:53:26,008 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5498, 3.6797, 2.2193, 4.0142, 2.7062, 4.0900, 2.4986, 2.9967], device='cuda:0'), covar=tensor([0.0287, 0.0378, 0.1698, 0.0407, 0.0798, 0.0529, 0.1402, 0.0718], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0176, 0.0193, 0.0161, 0.0176, 0.0216, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:53:39,401 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3682, 3.4391, 2.0624, 3.7503, 2.5388, 3.7922, 2.3263, 2.8474], device='cuda:0'), covar=tensor([0.0275, 0.0354, 0.1697, 0.0288, 0.0814, 0.0548, 0.1456, 0.0700], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0176, 0.0193, 0.0161, 0.0176, 0.0216, 0.0200, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:53:51,294 INFO [train.py:904] (0/8) Epoch 20, batch 4250, loss[loss=0.1595, simple_loss=0.2588, pruned_loss=0.03011, over 16785.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.273, pruned_loss=0.05064, over 3176704.74 frames. ], batch size: 83, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:54:12,868 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 01:54:28,668 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 01:54:39,857 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:54:59,315 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5368, 3.5997, 3.3596, 2.9972, 3.2022, 3.4947, 3.3077, 3.3199], device='cuda:0'), covar=tensor([0.0540, 0.0627, 0.0296, 0.0272, 0.0523, 0.0419, 0.1455, 0.0458], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0423, 0.0342, 0.0338, 0.0353, 0.0391, 0.0235, 0.0409], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 01:55:04,298 INFO [train.py:904] (0/8) Epoch 20, batch 4300, loss[loss=0.2017, simple_loss=0.2936, pruned_loss=0.05496, over 16691.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2737, pruned_loss=0.04953, over 3174112.05 frames. ], batch size: 134, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:55:37,987 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.176e+02 2.515e+02 3.039e+02 5.119e+02, threshold=5.030e+02, percent-clipped=0.0 2023-05-01 01:56:18,921 INFO [train.py:904] (0/8) Epoch 20, batch 4350, loss[loss=0.1823, simple_loss=0.2775, pruned_loss=0.04352, over 17030.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2766, pruned_loss=0.05033, over 3169264.62 frames. ], batch size: 50, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:56:29,918 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:56:33,579 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1750, 2.9690, 3.2337, 1.6245, 3.4457, 3.4926, 2.6576, 2.6230], device='cuda:0'), covar=tensor([0.0902, 0.0308, 0.0238, 0.1274, 0.0081, 0.0143, 0.0503, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0139, 0.0080, 0.0124, 0.0127, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:57:05,373 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:09,115 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:57:31,893 INFO [train.py:904] (0/8) Epoch 20, batch 4400, loss[loss=0.1776, simple_loss=0.27, pruned_loss=0.04267, over 17265.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2786, pruned_loss=0.0516, over 3154897.20 frames. ], batch size: 52, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:58:04,308 INFO [optim.py:368] (0/8) 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 01:58:12,914 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2815, 3.5220, 3.6829, 2.1473, 3.0371, 2.2685, 3.6345, 3.6611], device='cuda:0'), covar=tensor([0.0200, 0.0724, 0.0472, 0.1937, 0.0766, 0.0922, 0.0558, 0.0923], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0150, 0.0143, 0.0128, 0.0143, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 01:58:13,876 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 01:58:42,804 INFO [train.py:904] (0/8) Epoch 20, batch 4450, loss[loss=0.1971, simple_loss=0.2822, pruned_loss=0.05603, over 16774.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2814, pruned_loss=0.05251, over 3160704.74 frames. ], batch size: 39, lr: 3.39e-03, grad_scale: 16.0 2023-05-01 01:59:55,983 INFO [train.py:904] (0/8) Epoch 20, batch 4500, loss[loss=0.1914, simple_loss=0.2794, pruned_loss=0.05172, over 16632.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2821, pruned_loss=0.05334, over 3177093.56 frames. ], batch size: 62, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:00:30,612 INFO [optim.py:368] (0/8) 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 02:01:08,139 INFO [train.py:904] (0/8) Epoch 20, batch 4550, loss[loss=0.2181, simple_loss=0.2927, pruned_loss=0.0718, over 11790.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.283, pruned_loss=0.05423, over 3186620.23 frames. ], batch size: 247, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:01:20,800 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6895, 2.4151, 2.3419, 3.2075, 2.2687, 3.5719, 1.5143, 2.7394], device='cuda:0'), covar=tensor([0.1414, 0.0819, 0.1274, 0.0166, 0.0233, 0.0356, 0.1784, 0.0875], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0173, 0.0192, 0.0186, 0.0206, 0.0214, 0.0199, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:01:27,162 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197415.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:01:56,102 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197435.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:02:18,312 INFO [train.py:904] (0/8) Epoch 20, batch 4600, loss[loss=0.1995, simple_loss=0.294, pruned_loss=0.05251, over 17221.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2842, pruned_loss=0.05434, over 3210179.62 frames. ], batch size: 45, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:02:47,977 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8196, 5.0983, 5.2840, 4.9853, 5.0715, 5.6985, 5.1071, 4.8375], device='cuda:0'), covar=tensor([0.0954, 0.1657, 0.1837, 0.1905, 0.2339, 0.0844, 0.1453, 0.2259], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0578, 0.0632, 0.0481, 0.0639, 0.0672, 0.0495, 0.0645], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:02:52,208 INFO [optim.py:368] (0/8) 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 02:02:52,795 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197476.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 02:02:54,106 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 02:03:02,411 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:30,132 INFO [train.py:904] (0/8) Epoch 20, batch 4650, loss[loss=0.1762, simple_loss=0.2607, pruned_loss=0.04587, over 16493.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2831, pruned_loss=0.05443, over 3217204.79 frames. ], batch size: 68, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:03:41,195 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:03:49,635 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197516.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:05,604 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4770, 3.3767, 2.6389, 2.1807, 2.2427, 2.1353, 3.5977, 3.0176], device='cuda:0'), covar=tensor([0.2914, 0.0715, 0.1779, 0.2577, 0.2579, 0.2337, 0.0494, 0.1270], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0269, 0.0303, 0.0307, 0.0297, 0.0254, 0.0293, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:04:20,839 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=197537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:32,448 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197545.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:04:42,156 INFO [train.py:904] (0/8) Epoch 20, batch 4700, loss[loss=0.1675, simple_loss=0.2532, pruned_loss=0.04086, over 16567.00 frames. ], tot_loss[loss=0.1947, simple_loss=0.2817, pruned_loss=0.05385, over 3210089.58 frames. ], batch size: 57, lr: 3.39e-03, grad_scale: 8.0 2023-05-01 02:04:52,430 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:18,238 INFO [optim.py:368] (0/8) 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 02:05:19,885 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197577.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:05:30,681 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=197585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:05:41,916 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7028, 3.5688, 2.1680, 4.3113, 2.7516, 4.1899, 2.4570, 2.9843], device='cuda:0'), covar=tensor([0.0271, 0.0447, 0.1873, 0.0126, 0.0910, 0.0518, 0.1498, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0192, 0.0158, 0.0173, 0.0213, 0.0198, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:05:46,771 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6458, 3.5385, 2.0767, 4.2132, 2.7121, 4.1210, 2.3924, 2.9240], device='cuda:0'), covar=tensor([0.0277, 0.0401, 0.1889, 0.0133, 0.0926, 0.0472, 0.1515, 0.0805], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0175, 0.0192, 0.0158, 0.0173, 0.0213, 0.0198, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:05:55,408 INFO [train.py:904] (0/8) Epoch 20, batch 4750, loss[loss=0.1529, simple_loss=0.2461, pruned_loss=0.02983, over 16535.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2773, pruned_loss=0.0518, over 3201003.68 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:06:01,842 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:06:24,268 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 02:07:08,970 INFO [train.py:904] (0/8) Epoch 20, batch 4800, loss[loss=0.2005, simple_loss=0.2752, pruned_loss=0.06297, over 11991.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2734, pruned_loss=0.04929, over 3210469.84 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:07:32,692 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9976, 5.4901, 5.5914, 5.3525, 5.3510, 5.9550, 5.4078, 5.1877], device='cuda:0'), covar=tensor([0.0914, 0.1686, 0.1934, 0.1818, 0.2493, 0.0846, 0.1389, 0.2302], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0580, 0.0636, 0.0484, 0.0645, 0.0675, 0.0498, 0.0648], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:07:40,080 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 02:07:45,093 INFO [optim.py:368] (0/8) 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 02:08:24,133 INFO [train.py:904] (0/8) Epoch 20, batch 4850, loss[loss=0.1746, simple_loss=0.2686, pruned_loss=0.04025, over 16461.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2742, pruned_loss=0.04866, over 3192628.76 frames. ], batch size: 68, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:08:35,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-01 02:08:52,780 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8736, 3.8422, 2.3228, 4.6335, 2.9293, 4.4275, 2.4957, 3.0497], device='cuda:0'), covar=tensor([0.0244, 0.0388, 0.1648, 0.0118, 0.0847, 0.0455, 0.1405, 0.0772], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0174, 0.0192, 0.0157, 0.0173, 0.0212, 0.0197, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:09:38,115 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1586, 5.1491, 5.0048, 4.6202, 4.6045, 5.0710, 4.9690, 4.7283], device='cuda:0'), covar=tensor([0.0565, 0.0431, 0.0270, 0.0290, 0.1079, 0.0409, 0.0303, 0.0596], device='cuda:0'), in_proj_covar=tensor([0.0285, 0.0409, 0.0331, 0.0327, 0.0342, 0.0377, 0.0228, 0.0395], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:09:40,082 INFO [train.py:904] (0/8) Epoch 20, batch 4900, loss[loss=0.1706, simple_loss=0.2538, pruned_loss=0.04372, over 17115.00 frames. ], tot_loss[loss=0.1842, simple_loss=0.2735, pruned_loss=0.04749, over 3198249.07 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:10:08,666 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197771.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:10:15,934 INFO [optim.py:368] (0/8) 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 02:10:48,372 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7670, 3.7459, 2.2771, 4.4922, 2.7682, 4.3325, 2.4923, 3.0980], device='cuda:0'), covar=tensor([0.0261, 0.0361, 0.1679, 0.0109, 0.0858, 0.0493, 0.1471, 0.0671], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0174, 0.0192, 0.0157, 0.0174, 0.0212, 0.0198, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:10:52,836 INFO [train.py:904] (0/8) Epoch 20, batch 4950, loss[loss=0.1909, simple_loss=0.2886, pruned_loss=0.04656, over 16449.00 frames. ], tot_loss[loss=0.1832, simple_loss=0.2727, pruned_loss=0.04689, over 3199964.79 frames. ], batch size: 146, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:11:27,958 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.04 vs. limit=5.0 2023-05-01 02:11:56,671 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3103, 5.3591, 5.6919, 5.6336, 5.7196, 5.3529, 5.3000, 4.9405], device='cuda:0'), covar=tensor([0.0260, 0.0445, 0.0307, 0.0355, 0.0474, 0.0327, 0.1020, 0.0424], device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0430, 0.0414, 0.0388, 0.0462, 0.0438, 0.0527, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 02:12:04,532 INFO [train.py:904] (0/8) Epoch 20, batch 5000, loss[loss=0.1731, simple_loss=0.2668, pruned_loss=0.0397, over 17097.00 frames. ], tot_loss[loss=0.1845, simple_loss=0.2745, pruned_loss=0.04723, over 3189286.84 frames. ], batch size: 49, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:12:32,800 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197872.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:12:38,860 INFO [optim.py:368] (0/8) 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 02:13:01,265 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=197892.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:13,658 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=197901.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:13:14,331 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 02:13:14,486 INFO [train.py:904] (0/8) Epoch 20, batch 5050, loss[loss=0.1745, simple_loss=0.2727, pruned_loss=0.03821, over 16883.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2753, pruned_loss=0.04699, over 3206798.53 frames. ], batch size: 96, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:14:09,937 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7676, 5.0464, 5.2285, 4.9525, 5.0385, 5.6501, 5.0797, 4.7754], device='cuda:0'), covar=tensor([0.0972, 0.1755, 0.1584, 0.1693, 0.2194, 0.0764, 0.1361, 0.2074], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0571, 0.0624, 0.0475, 0.0635, 0.0664, 0.0491, 0.0640], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:14:25,053 INFO [train.py:904] (0/8) Epoch 20, batch 5100, loss[loss=0.1811, simple_loss=0.2685, pruned_loss=0.0469, over 16245.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.273, pruned_loss=0.04612, over 3214968.68 frames. ], batch size: 165, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:14:27,241 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=197953.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:15:00,614 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.528e+02 1.957e+02 2.274e+02 2.606e+02 4.454e+02, threshold=4.549e+02, percent-clipped=0.0 2023-05-01 02:15:05,268 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 02:15:15,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6878, 3.4754, 4.1385, 1.8880, 4.2397, 4.2980, 3.1130, 2.9894], device='cuda:0'), covar=tensor([0.0745, 0.0289, 0.0121, 0.1250, 0.0057, 0.0115, 0.0341, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0108, 0.0096, 0.0138, 0.0079, 0.0123, 0.0127, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:15:35,521 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-198000.pt 2023-05-01 02:15:41,438 INFO [train.py:904] (0/8) Epoch 20, batch 5150, loss[loss=0.1713, simple_loss=0.2684, pruned_loss=0.03709, over 16718.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2731, pruned_loss=0.04542, over 3206557.19 frames. ], batch size: 83, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:16:22,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2481, 3.9937, 3.9686, 2.6304, 3.5381, 3.9839, 3.5020, 2.1342], device='cuda:0'), covar=tensor([0.0566, 0.0045, 0.0041, 0.0380, 0.0089, 0.0092, 0.0095, 0.0495], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0081, 0.0080, 0.0131, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:16:50,971 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 02:16:52,362 INFO [train.py:904] (0/8) Epoch 20, batch 5200, loss[loss=0.1737, simple_loss=0.263, pruned_loss=0.04218, over 12007.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2717, pruned_loss=0.04487, over 3206638.88 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:17:06,808 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 02:17:19,630 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198071.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:17:27,748 INFO [optim.py:368] (0/8) 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 02:17:34,771 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198081.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:18:03,852 INFO [train.py:904] (0/8) Epoch 20, batch 5250, loss[loss=0.1631, simple_loss=0.2578, pruned_loss=0.03426, over 16885.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2698, pruned_loss=0.04511, over 3201700.18 frames. ], batch size: 96, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:18:26,373 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 02:18:28,865 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198119.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:19:01,572 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:19:15,057 INFO [train.py:904] (0/8) Epoch 20, batch 5300, loss[loss=0.144, simple_loss=0.2274, pruned_loss=0.03035, over 16595.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.267, pruned_loss=0.04441, over 3204579.14 frames. ], batch size: 57, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:19:44,958 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198172.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:19:49,807 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 1.825e+02 2.153e+02 2.534e+02 4.591e+02, threshold=4.307e+02, percent-clipped=1.0 2023-05-01 02:20:26,997 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198201.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:28,374 INFO [train.py:904] (0/8) Epoch 20, batch 5350, loss[loss=0.1848, simple_loss=0.2624, pruned_loss=0.05361, over 11945.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2651, pruned_loss=0.04365, over 3196704.96 frames. ], batch size: 246, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:20:28,965 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3335, 4.1671, 4.1183, 2.6729, 3.6087, 4.1416, 3.6186, 2.3025], device='cuda:0'), covar=tensor([0.0547, 0.0037, 0.0039, 0.0392, 0.0096, 0.0080, 0.0101, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0081, 0.0080, 0.0132, 0.0095, 0.0106, 0.0092, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:20:30,909 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4999, 2.8999, 3.1019, 2.0522, 2.7973, 2.1488, 3.0963, 3.1697], device='cuda:0'), covar=tensor([0.0297, 0.0772, 0.0593, 0.1859, 0.0794, 0.0996, 0.0613, 0.0772], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0162, 0.0166, 0.0151, 0.0143, 0.0128, 0.0144, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:20:54,788 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198220.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:20:55,400 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 02:21:13,360 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9699, 5.0047, 4.8604, 4.4794, 4.4877, 4.9352, 4.8338, 4.6342], device='cuda:0'), covar=tensor([0.0671, 0.0606, 0.0275, 0.0296, 0.1028, 0.0534, 0.0354, 0.0568], device='cuda:0'), in_proj_covar=tensor([0.0289, 0.0417, 0.0336, 0.0332, 0.0348, 0.0387, 0.0232, 0.0403], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:21:35,329 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198248.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:36,564 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198249.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:21:40,984 INFO [train.py:904] (0/8) Epoch 20, batch 5400, loss[loss=0.1917, simple_loss=0.2932, pruned_loss=0.04513, over 16507.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2668, pruned_loss=0.04374, over 3205343.52 frames. ], batch size: 75, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:22:15,985 INFO [optim.py:368] (0/8) 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 02:22:57,924 INFO [train.py:904] (0/8) Epoch 20, batch 5450, loss[loss=0.207, simple_loss=0.2942, pruned_loss=0.0599, over 16695.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2701, pruned_loss=0.04545, over 3185892.06 frames. ], batch size: 134, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:23:04,297 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0682, 2.4867, 2.6021, 1.9129, 2.7486, 2.8387, 2.4628, 2.3959], device='cuda:0'), covar=tensor([0.0709, 0.0262, 0.0204, 0.0931, 0.0106, 0.0231, 0.0413, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0110, 0.0098, 0.0141, 0.0081, 0.0125, 0.0129, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:24:08,127 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 02:24:14,768 INFO [train.py:904] (0/8) Epoch 20, batch 5500, loss[loss=0.2025, simple_loss=0.2929, pruned_loss=0.056, over 16659.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2767, pruned_loss=0.04893, over 3182689.91 frames. ], batch size: 134, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:24:51,696 INFO [optim.py:368] (0/8) 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 02:25:21,789 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4374, 1.6666, 2.1607, 2.3417, 2.4276, 2.7084, 1.7974, 2.6586], device='cuda:0'), covar=tensor([0.0213, 0.0493, 0.0273, 0.0328, 0.0290, 0.0178, 0.0509, 0.0138], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0191, 0.0176, 0.0181, 0.0192, 0.0149, 0.0192, 0.0145], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:25:34,176 INFO [train.py:904] (0/8) Epoch 20, batch 5550, loss[loss=0.3023, simple_loss=0.3559, pruned_loss=0.1244, over 11266.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.284, pruned_loss=0.05377, over 3158274.67 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:26:30,798 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198437.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:26:35,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0804, 2.9950, 3.2887, 1.7096, 3.4507, 3.5148, 2.7702, 2.5766], device='cuda:0'), covar=tensor([0.0890, 0.0279, 0.0220, 0.1252, 0.0079, 0.0186, 0.0458, 0.0512], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0109, 0.0098, 0.0140, 0.0080, 0.0124, 0.0128, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:26:52,167 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.44 vs. limit=5.0 2023-05-01 02:26:53,782 INFO [train.py:904] (0/8) Epoch 20, batch 5600, loss[loss=0.296, simple_loss=0.3572, pruned_loss=0.1174, over 11323.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2896, pruned_loss=0.05829, over 3127585.56 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:27:34,788 INFO [optim.py:368] (0/8) 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 02:28:14,145 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198499.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:28:18,613 INFO [train.py:904] (0/8) Epoch 20, batch 5650, loss[loss=0.2153, simple_loss=0.302, pruned_loss=0.06426, over 16382.00 frames. ], tot_loss[loss=0.2093, simple_loss=0.2941, pruned_loss=0.0622, over 3093114.19 frames. ], batch size: 146, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:28:29,247 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4722, 3.4707, 2.0211, 4.0087, 2.6808, 3.9309, 2.2882, 2.7722], device='cuda:0'), covar=tensor([0.0315, 0.0429, 0.1796, 0.0159, 0.0831, 0.0539, 0.1551, 0.0798], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0157, 0.0175, 0.0213, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:29:31,879 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198548.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:29:36,990 INFO [train.py:904] (0/8) Epoch 20, batch 5700, loss[loss=0.2238, simple_loss=0.316, pruned_loss=0.06581, over 16929.00 frames. ], tot_loss[loss=0.2123, simple_loss=0.2961, pruned_loss=0.06428, over 3066248.04 frames. ], batch size: 116, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:29:50,276 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:14,612 INFO [optim.py:368] (0/8) 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 02:30:27,745 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-01 02:30:36,246 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 02:30:46,649 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:30:46,850 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6128, 3.6196, 2.2561, 4.2978, 2.8427, 4.1874, 2.4235, 2.9191], device='cuda:0'), covar=tensor([0.0310, 0.0406, 0.1742, 0.0167, 0.0830, 0.0538, 0.1503, 0.0809], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0193, 0.0157, 0.0175, 0.0213, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:30:55,565 INFO [train.py:904] (0/8) Epoch 20, batch 5750, loss[loss=0.2439, simple_loss=0.3059, pruned_loss=0.09101, over 11322.00 frames. ], tot_loss[loss=0.2152, simple_loss=0.299, pruned_loss=0.06566, over 3057024.93 frames. ], batch size: 248, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:31:17,290 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:32:16,958 INFO [train.py:904] (0/8) Epoch 20, batch 5800, loss[loss=0.2007, simple_loss=0.2962, pruned_loss=0.05263, over 16252.00 frames. ], tot_loss[loss=0.2132, simple_loss=0.2984, pruned_loss=0.06401, over 3068560.54 frames. ], batch size: 165, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:32:53,799 INFO [optim.py:368] (0/8) 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 02:32:54,372 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198676.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:33:34,967 INFO [train.py:904] (0/8) Epoch 20, batch 5850, loss[loss=0.189, simple_loss=0.284, pruned_loss=0.04695, over 16749.00 frames. ], tot_loss[loss=0.2115, simple_loss=0.2969, pruned_loss=0.06309, over 3052517.27 frames. ], batch size: 83, lr: 3.38e-03, grad_scale: 8.0 2023-05-01 02:34:30,874 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=198737.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:34:53,519 INFO [train.py:904] (0/8) Epoch 20, batch 5900, loss[loss=0.2089, simple_loss=0.292, pruned_loss=0.06293, over 16430.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2969, pruned_loss=0.06317, over 3063041.44 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:35:34,253 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=198785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:35:58,676 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7657, 2.5899, 2.5954, 1.9102, 2.5317, 2.6202, 2.5758, 1.8651], device='cuda:0'), covar=tensor([0.0457, 0.0120, 0.0088, 0.0376, 0.0133, 0.0151, 0.0109, 0.0415], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0082, 0.0081, 0.0133, 0.0096, 0.0108, 0.0093, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:36:12,991 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6203, 4.6850, 4.5192, 4.1935, 4.1853, 4.6106, 4.3946, 4.3159], device='cuda:0'), covar=tensor([0.0747, 0.0697, 0.0323, 0.0357, 0.0889, 0.0500, 0.0536, 0.0661], device='cuda:0'), in_proj_covar=tensor([0.0284, 0.0411, 0.0331, 0.0325, 0.0341, 0.0380, 0.0228, 0.0396], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:36:14,572 INFO [train.py:904] (0/8) Epoch 20, batch 5950, loss[loss=0.1812, simple_loss=0.2762, pruned_loss=0.04313, over 16841.00 frames. ], tot_loss[loss=0.2097, simple_loss=0.2965, pruned_loss=0.06146, over 3069518.65 frames. ], batch size: 102, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:01,091 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5319, 3.4960, 2.1233, 4.1250, 2.7418, 4.0209, 2.2326, 2.8446], device='cuda:0'), covar=tensor([0.0309, 0.0480, 0.1798, 0.0216, 0.0855, 0.0637, 0.1626, 0.0794], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0192, 0.0157, 0.0175, 0.0213, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:37:31,022 INFO [train.py:904] (0/8) Epoch 20, batch 6000, loss[loss=0.1928, simple_loss=0.2807, pruned_loss=0.05242, over 16395.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2952, pruned_loss=0.06042, over 3085088.42 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:37:31,023 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 02:37:41,833 INFO [train.py:938] (0/8) Epoch 20, validation: loss=0.1516, simple_loss=0.2644, pruned_loss=0.01942, over 944034.00 frames. 2023-05-01 02:37:41,834 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 02:37:42,431 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:37:46,502 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198855.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:38:10,671 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7398, 2.7799, 2.6128, 4.6965, 3.4467, 4.1734, 1.8245, 3.0369], device='cuda:0'), covar=tensor([0.1350, 0.0843, 0.1252, 0.0172, 0.0320, 0.0423, 0.1602, 0.0831], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0173, 0.0193, 0.0185, 0.0205, 0.0212, 0.0199, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:38:17,163 INFO [optim.py:368] (0/8) 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 02:38:28,015 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:38:56,338 INFO [train.py:904] (0/8) Epoch 20, batch 6050, loss[loss=0.1964, simple_loss=0.2935, pruned_loss=0.04961, over 16587.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2933, pruned_loss=0.05967, over 3088551.93 frames. ], batch size: 62, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:39:14,687 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198913.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:39:36,259 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=198927.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:05,135 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198944.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:13,070 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9469, 3.8450, 4.0067, 4.1345, 4.2478, 3.8135, 4.1959, 4.2593], device='cuda:0'), covar=tensor([0.1838, 0.1301, 0.1556, 0.0730, 0.0600, 0.1659, 0.0799, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0623, 0.0770, 0.0895, 0.0782, 0.0589, 0.0615, 0.0630, 0.0729], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:40:17,609 INFO [train.py:904] (0/8) Epoch 20, batch 6100, loss[loss=0.2039, simple_loss=0.2916, pruned_loss=0.05808, over 17070.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2929, pruned_loss=0.05868, over 3107107.90 frames. ], batch size: 53, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:40:46,241 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=198971.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:40:53,273 INFO [optim.py:368] (0/8) 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 02:41:12,522 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=198988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:41:15,632 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 02:41:32,341 INFO [train.py:904] (0/8) Epoch 20, batch 6150, loss[loss=0.2227, simple_loss=0.2993, pruned_loss=0.07306, over 11653.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2915, pruned_loss=0.0587, over 3085775.37 frames. ], batch size: 247, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:41:40,511 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6468, 2.2797, 1.8182, 2.0362, 2.6588, 2.2618, 2.4846, 2.7468], device='cuda:0'), covar=tensor([0.0195, 0.0380, 0.0550, 0.0461, 0.0240, 0.0378, 0.0196, 0.0259], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0228, 0.0220, 0.0221, 0.0231, 0.0228, 0.0229, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:42:49,588 INFO [train.py:904] (0/8) Epoch 20, batch 6200, loss[loss=0.182, simple_loss=0.2747, pruned_loss=0.04465, over 16699.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2891, pruned_loss=0.05784, over 3095806.39 frames. ], batch size: 76, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:43:28,026 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.886e+02 2.934e+02 3.552e+02 4.446e+02 1.093e+03, threshold=7.104e+02, percent-clipped=2.0 2023-05-01 02:43:38,904 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4437, 3.3547, 2.6294, 2.1067, 2.1977, 2.2702, 3.4959, 3.0786], device='cuda:0'), covar=tensor([0.2896, 0.0682, 0.1793, 0.2612, 0.2637, 0.2202, 0.0531, 0.1292], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0267, 0.0303, 0.0306, 0.0295, 0.0253, 0.0293, 0.0331], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:44:06,538 INFO [train.py:904] (0/8) Epoch 20, batch 6250, loss[loss=0.1928, simple_loss=0.2828, pruned_loss=0.05142, over 15424.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2888, pruned_loss=0.05796, over 3090617.44 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:21,393 INFO [train.py:904] (0/8) Epoch 20, batch 6300, loss[loss=0.1748, simple_loss=0.2781, pruned_loss=0.03573, over 16846.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2883, pruned_loss=0.0571, over 3085812.29 frames. ], batch size: 96, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:45:25,932 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199155.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:45:37,258 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 02:45:59,363 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:00,085 INFO [optim.py:368] (0/8) 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 02:46:01,782 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9202, 2.4608, 2.0094, 2.2049, 2.8309, 2.4459, 2.7062, 2.9520], device='cuda:0'), covar=tensor([0.0185, 0.0388, 0.0518, 0.0411, 0.0217, 0.0318, 0.0204, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0229, 0.0222, 0.0222, 0.0232, 0.0230, 0.0231, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:46:38,976 INFO [train.py:904] (0/8) Epoch 20, batch 6350, loss[loss=0.1877, simple_loss=0.2786, pruned_loss=0.04834, over 16875.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2896, pruned_loss=0.05864, over 3077199.68 frames. ], batch size: 96, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 02:46:40,593 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:46:47,430 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199208.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:30,800 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199237.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 02:47:33,741 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199239.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:47:38,859 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 02:47:52,854 INFO [train.py:904] (0/8) Epoch 20, batch 6400, loss[loss=0.1784, simple_loss=0.2638, pruned_loss=0.0465, over 16705.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2903, pruned_loss=0.05975, over 3077316.29 frames. ], batch size: 76, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:47:53,823 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-05-01 02:47:54,560 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5951, 2.3239, 2.1352, 3.2501, 2.0301, 3.5253, 1.5038, 2.5253], device='cuda:0'), covar=tensor([0.1712, 0.1003, 0.1508, 0.0258, 0.0234, 0.0414, 0.2108, 0.1034], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0186, 0.0207, 0.0214, 0.0201, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 02:48:21,607 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199271.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:48:23,523 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 02:48:25,701 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 02:48:29,164 INFO [optim.py:368] (0/8) 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 02:48:39,448 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199283.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:49:07,114 INFO [train.py:904] (0/8) Epoch 20, batch 6450, loss[loss=0.2105, simple_loss=0.3058, pruned_loss=0.05765, over 15433.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2901, pruned_loss=0.05892, over 3081484.92 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:49:17,294 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.17 vs. limit=5.0 2023-05-01 02:49:33,334 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:50:24,670 INFO [train.py:904] (0/8) Epoch 20, batch 6500, loss[loss=0.2003, simple_loss=0.272, pruned_loss=0.06429, over 12049.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2878, pruned_loss=0.05829, over 3076953.82 frames. ], batch size: 248, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:50:43,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1849, 4.1076, 4.2943, 4.4332, 4.5344, 4.1308, 4.4799, 4.5601], device='cuda:0'), covar=tensor([0.2012, 0.1267, 0.1379, 0.0644, 0.0572, 0.1276, 0.0766, 0.0661], device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0761, 0.0890, 0.0773, 0.0585, 0.0608, 0.0624, 0.0724], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 02:51:02,010 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.751e+02 2.867e+02 3.442e+02 4.204e+02 8.359e+02, threshold=6.884e+02, percent-clipped=2.0 2023-05-01 02:51:39,114 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:51:41,757 INFO [train.py:904] (0/8) Epoch 20, batch 6550, loss[loss=0.1926, simple_loss=0.2983, pruned_loss=0.04343, over 16384.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2906, pruned_loss=0.05914, over 3068042.42 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:51:58,394 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.57 vs. limit=5.0 2023-05-01 02:52:36,201 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 02:52:56,170 INFO [train.py:904] (0/8) Epoch 20, batch 6600, loss[loss=0.2177, simple_loss=0.3008, pruned_loss=0.06727, over 15283.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2918, pruned_loss=0.05872, over 3100575.16 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:53:05,825 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199458.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:53:10,052 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199461.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:53:33,378 INFO [optim.py:368] (0/8) 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 02:53:59,284 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199494.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:54:11,574 INFO [train.py:904] (0/8) Epoch 20, batch 6650, loss[loss=0.1933, simple_loss=0.2803, pruned_loss=0.05317, over 16929.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2922, pruned_loss=0.05948, over 3107970.66 frames. ], batch size: 109, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:54:20,063 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199508.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:54:37,242 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199519.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:54:56,464 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199532.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:00,544 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:06,204 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199539.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:55:25,191 INFO [train.py:904] (0/8) Epoch 20, batch 6700, loss[loss=0.2109, simple_loss=0.2927, pruned_loss=0.06454, over 15230.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.292, pruned_loss=0.06058, over 3090671.10 frames. ], batch size: 190, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:55:30,726 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199555.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 02:55:32,563 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199556.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:02,760 INFO [optim.py:368] (0/8) 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 02:56:13,387 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:18,212 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:56:29,432 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-01 02:56:31,519 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199596.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 02:56:38,616 INFO [train.py:904] (0/8) Epoch 20, batch 6750, loss[loss=0.2649, simple_loss=0.3296, pruned_loss=0.1001, over 11485.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2909, pruned_loss=0.06069, over 3088495.43 frames. ], batch size: 247, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:56:57,521 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-01 02:56:59,551 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4671, 3.4860, 2.5791, 2.1571, 2.3910, 2.2646, 3.5899, 3.1759], device='cuda:0'), covar=tensor([0.3224, 0.0731, 0.2094, 0.2849, 0.2655, 0.2244, 0.0614, 0.1336], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0268, 0.0302, 0.0307, 0.0295, 0.0253, 0.0292, 0.0332], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 02:57:20,692 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 02:57:53,104 INFO [train.py:904] (0/8) Epoch 20, batch 6800, loss[loss=0.2419, simple_loss=0.3122, pruned_loss=0.08576, over 11396.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2912, pruned_loss=0.06093, over 3072363.65 frames. ], batch size: 247, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 02:58:29,264 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.005e+02 2.711e+02 3.238e+02 3.835e+02 6.988e+02, threshold=6.476e+02, percent-clipped=0.0 2023-05-01 02:59:05,443 INFO [train.py:904] (0/8) Epoch 20, batch 6850, loss[loss=0.1987, simple_loss=0.2956, pruned_loss=0.05092, over 16432.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.292, pruned_loss=0.06081, over 3068421.04 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 8.0 2023-05-01 03:00:20,629 INFO [train.py:904] (0/8) Epoch 20, batch 6900, loss[loss=0.2145, simple_loss=0.3026, pruned_loss=0.06324, over 16139.00 frames. ], tot_loss[loss=0.2069, simple_loss=0.2941, pruned_loss=0.0598, over 3090786.36 frames. ], batch size: 165, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:00:26,876 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199756.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:00:33,711 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4108, 4.6747, 4.4629, 4.4935, 4.2050, 4.2343, 4.2400, 4.7110], device='cuda:0'), covar=tensor([0.1159, 0.0856, 0.1058, 0.0839, 0.0786, 0.1348, 0.1074, 0.0920], device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0786, 0.0650, 0.0591, 0.0497, 0.0506, 0.0660, 0.0608], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:00:59,800 INFO [optim.py:368] (0/8) 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 03:01:00,652 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 03:01:07,551 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.00 vs. limit=5.0 2023-05-01 03:01:33,822 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 03:01:36,165 INFO [train.py:904] (0/8) Epoch 20, batch 6950, loss[loss=0.1967, simple_loss=0.2982, pruned_loss=0.04762, over 16834.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2963, pruned_loss=0.06192, over 3071598.11 frames. ], batch size: 102, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:01:54,393 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199814.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:02:22,401 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=199832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:02:47,656 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199850.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:02:49,636 INFO [train.py:904] (0/8) Epoch 20, batch 7000, loss[loss=0.2007, simple_loss=0.2984, pruned_loss=0.0515, over 16362.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2963, pruned_loss=0.06104, over 3084924.04 frames. ], batch size: 146, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:03:29,996 INFO [optim.py:368] (0/8) 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 03:03:33,416 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=199880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:03:49,870 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=199891.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:04:07,202 INFO [train.py:904] (0/8) Epoch 20, batch 7050, loss[loss=0.2012, simple_loss=0.3067, pruned_loss=0.04781, over 17040.00 frames. ], tot_loss[loss=0.2092, simple_loss=0.2969, pruned_loss=0.06081, over 3091831.07 frames. ], batch size: 50, lr: 3.37e-03, grad_scale: 4.0 2023-05-01 03:04:17,914 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=199909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:05:25,781 INFO [train.py:904] (0/8) Epoch 20, batch 7100, loss[loss=0.1848, simple_loss=0.2745, pruned_loss=0.04756, over 16999.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2955, pruned_loss=0.06118, over 3071728.14 frames. ], batch size: 41, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:05:54,857 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=199970.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:06:05,960 INFO [optim.py:368] (0/8) 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 03:06:39,360 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-200000.pt 2023-05-01 03:06:45,332 INFO [train.py:904] (0/8) Epoch 20, batch 7150, loss[loss=0.2058, simple_loss=0.2941, pruned_loss=0.05872, over 16383.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2931, pruned_loss=0.06069, over 3068931.81 frames. ], batch size: 146, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:07:59,663 INFO [train.py:904] (0/8) Epoch 20, batch 7200, loss[loss=0.1754, simple_loss=0.2693, pruned_loss=0.04078, over 16528.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2908, pruned_loss=0.05888, over 3074876.24 frames. ], batch size: 68, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:08:05,985 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200056.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:08:16,591 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5064, 3.5070, 1.8963, 4.0282, 2.6410, 3.9551, 2.0189, 2.6579], device='cuda:0'), covar=tensor([0.0302, 0.0395, 0.2040, 0.0204, 0.0891, 0.0579, 0.1979, 0.0926], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0174, 0.0193, 0.0156, 0.0174, 0.0213, 0.0200, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 03:08:40,490 INFO [optim.py:368] (0/8) 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 03:09:02,753 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 03:09:19,963 INFO [train.py:904] (0/8) Epoch 20, batch 7250, loss[loss=0.1999, simple_loss=0.2818, pruned_loss=0.05896, over 16827.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2876, pruned_loss=0.05711, over 3072924.35 frames. ], batch size: 124, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:09:23,331 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200104.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:09:37,483 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:10:01,277 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4225, 5.7305, 5.4625, 5.5099, 5.1975, 5.1438, 5.1481, 5.8519], device='cuda:0'), covar=tensor([0.1141, 0.0775, 0.1015, 0.0832, 0.0784, 0.0751, 0.1212, 0.0824], device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0789, 0.0652, 0.0594, 0.0497, 0.0508, 0.0662, 0.0610], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:10:31,968 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200150.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:10:35,114 INFO [train.py:904] (0/8) Epoch 20, batch 7300, loss[loss=0.1794, simple_loss=0.2767, pruned_loss=0.0411, over 16762.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2875, pruned_loss=0.0574, over 3052391.05 frames. ], batch size: 76, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:10:50,985 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200162.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:11:14,612 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200177.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:11:15,883 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.776e+02 2.904e+02 3.413e+02 4.614e+02 7.836e+02, threshold=6.825e+02, percent-clipped=9.0 2023-05-01 03:11:37,015 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200191.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:11:47,184 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200198.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 03:11:52,599 INFO [train.py:904] (0/8) Epoch 20, batch 7350, loss[loss=0.2014, simple_loss=0.2888, pruned_loss=0.05699, over 16688.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2897, pruned_loss=0.05931, over 3038467.02 frames. ], batch size: 134, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:11:53,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6360, 4.8806, 4.6458, 4.6668, 4.4141, 4.3587, 4.3623, 4.9123], device='cuda:0'), covar=tensor([0.1016, 0.0739, 0.1058, 0.0821, 0.0748, 0.1236, 0.1105, 0.0806], device='cuda:0'), in_proj_covar=tensor([0.0639, 0.0785, 0.0648, 0.0589, 0.0494, 0.0505, 0.0659, 0.0607], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:12:51,348 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200238.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:12:52,208 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200239.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:13:12,421 INFO [train.py:904] (0/8) Epoch 20, batch 7400, loss[loss=0.2199, simple_loss=0.3044, pruned_loss=0.06766, over 16603.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2907, pruned_loss=0.05984, over 3047172.70 frames. ], batch size: 62, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:13:13,029 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7224, 2.7754, 2.2838, 2.6693, 3.1853, 2.8062, 3.2783, 3.3282], device='cuda:0'), covar=tensor([0.0099, 0.0404, 0.0535, 0.0383, 0.0242, 0.0347, 0.0242, 0.0235], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0227, 0.0220, 0.0220, 0.0229, 0.0228, 0.0227, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:13:33,061 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:13:52,751 INFO [optim.py:368] (0/8) 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 03:13:55,476 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:13:57,045 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9626, 4.9738, 4.7652, 4.0796, 4.8735, 1.9582, 4.6131, 4.4342], device='cuda:0'), covar=tensor([0.0084, 0.0070, 0.0175, 0.0360, 0.0085, 0.2530, 0.0123, 0.0222], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0148, 0.0192, 0.0175, 0.0169, 0.0203, 0.0182, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:14:32,179 INFO [train.py:904] (0/8) Epoch 20, batch 7450, loss[loss=0.1941, simple_loss=0.2884, pruned_loss=0.04989, over 16850.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2916, pruned_loss=0.06027, over 3062468.93 frames. ], batch size: 116, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:15:35,814 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200340.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:15:49,125 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7874, 1.7728, 1.6071, 1.4455, 1.9331, 1.5617, 1.5981, 1.8762], device='cuda:0'), covar=tensor([0.0193, 0.0284, 0.0425, 0.0340, 0.0214, 0.0255, 0.0175, 0.0201], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0227, 0.0220, 0.0220, 0.0229, 0.0228, 0.0227, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:15:54,173 INFO [train.py:904] (0/8) Epoch 20, batch 7500, loss[loss=0.2666, simple_loss=0.3292, pruned_loss=0.102, over 11288.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2922, pruned_loss=0.06009, over 3046499.68 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:16:34,101 INFO [optim.py:368] (0/8) 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 03:17:11,216 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 03:17:11,468 INFO [train.py:904] (0/8) Epoch 20, batch 7550, loss[loss=0.2376, simple_loss=0.3079, pruned_loss=0.08371, over 11904.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2915, pruned_loss=0.06083, over 3037480.37 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:17:20,695 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3759, 3.3113, 3.4115, 3.4847, 3.5375, 3.2955, 3.4953, 3.5604], device='cuda:0'), covar=tensor([0.1363, 0.1097, 0.1071, 0.0662, 0.0688, 0.2250, 0.1133, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0613, 0.0754, 0.0882, 0.0771, 0.0581, 0.0603, 0.0623, 0.0719], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:18:26,305 INFO [train.py:904] (0/8) Epoch 20, batch 7600, loss[loss=0.2007, simple_loss=0.2769, pruned_loss=0.06224, over 16712.00 frames. ], tot_loss[loss=0.2055, simple_loss=0.2905, pruned_loss=0.0603, over 3064258.24 frames. ], batch size: 39, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:18:47,733 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 03:19:06,494 INFO [optim.py:368] (0/8) 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 03:19:44,064 INFO [train.py:904] (0/8) Epoch 20, batch 7650, loss[loss=0.2257, simple_loss=0.3036, pruned_loss=0.07389, over 15305.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2912, pruned_loss=0.06108, over 3053443.67 frames. ], batch size: 190, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:20:33,303 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200533.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 03:21:02,186 INFO [train.py:904] (0/8) Epoch 20, batch 7700, loss[loss=0.2053, simple_loss=0.2872, pruned_loss=0.06169, over 16756.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2912, pruned_loss=0.06156, over 3046405.82 frames. ], batch size: 62, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:21:22,811 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 03:22:21,791 INFO [train.py:904] (0/8) Epoch 20, batch 7750, loss[loss=0.2452, simple_loss=0.3077, pruned_loss=0.09131, over 11408.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2919, pruned_loss=0.06167, over 3054360.17 frames. ], batch size: 248, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:22:38,654 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200613.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:11,951 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=200635.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:23:20,275 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9270, 5.0250, 5.4125, 5.3919, 5.3989, 5.0577, 4.9861, 4.6816], device='cuda:0'), covar=tensor([0.0345, 0.0488, 0.0320, 0.0340, 0.0470, 0.0336, 0.0960, 0.0476], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0440, 0.0425, 0.0397, 0.0475, 0.0448, 0.0539, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 03:23:27,480 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=200645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:23:36,971 INFO [train.py:904] (0/8) Epoch 20, batch 7800, loss[loss=0.2108, simple_loss=0.2949, pruned_loss=0.06337, over 16925.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2932, pruned_loss=0.06226, over 3050956.74 frames. ], batch size: 109, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:23:53,484 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 03:24:18,683 INFO [optim.py:368] (0/8) 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 03:24:30,680 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2469, 4.2185, 4.1236, 3.3533, 4.1725, 1.7237, 3.9313, 3.6958], device='cuda:0'), covar=tensor([0.0119, 0.0102, 0.0172, 0.0294, 0.0087, 0.2830, 0.0136, 0.0252], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0147, 0.0190, 0.0173, 0.0167, 0.0201, 0.0180, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:24:53,420 INFO [train.py:904] (0/8) Epoch 20, batch 7850, loss[loss=0.2481, simple_loss=0.3146, pruned_loss=0.09079, over 11881.00 frames. ], tot_loss[loss=0.2082, simple_loss=0.2935, pruned_loss=0.06144, over 3050149.47 frames. ], batch size: 247, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:24:59,641 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7351, 3.7310, 2.3024, 4.3527, 2.8825, 4.2490, 2.4113, 3.0160], device='cuda:0'), covar=tensor([0.0290, 0.0381, 0.1697, 0.0209, 0.0828, 0.0570, 0.1567, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0176, 0.0195, 0.0157, 0.0176, 0.0215, 0.0201, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 03:24:59,660 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=200706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:25:41,932 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3746, 3.6021, 3.7269, 2.2043, 3.2627, 2.4688, 3.7160, 3.7884], device='cuda:0'), covar=tensor([0.0220, 0.0775, 0.0574, 0.2036, 0.0792, 0.0968, 0.0596, 0.0939], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0162, 0.0166, 0.0152, 0.0144, 0.0130, 0.0144, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 03:26:08,730 INFO [train.py:904] (0/8) Epoch 20, batch 7900, loss[loss=0.2039, simple_loss=0.286, pruned_loss=0.06093, over 16578.00 frames. ], tot_loss[loss=0.2064, simple_loss=0.292, pruned_loss=0.06044, over 3071454.43 frames. ], batch size: 35, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:26:15,113 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0760, 5.0899, 4.9103, 4.5533, 4.5920, 5.0063, 4.8999, 4.6356], device='cuda:0'), covar=tensor([0.0684, 0.0588, 0.0310, 0.0312, 0.1058, 0.0462, 0.0323, 0.0688], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0407, 0.0328, 0.0324, 0.0337, 0.0377, 0.0227, 0.0394], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:26:47,283 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.71 vs. limit=2.0 2023-05-01 03:26:49,222 INFO [optim.py:368] (0/8) 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 03:26:52,490 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-01 03:27:27,137 INFO [train.py:904] (0/8) Epoch 20, batch 7950, loss[loss=0.1944, simple_loss=0.2779, pruned_loss=0.05546, over 16535.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2916, pruned_loss=0.06047, over 3071975.05 frames. ], batch size: 75, lr: 3.36e-03, grad_scale: 4.0 2023-05-01 03:28:15,520 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200833.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:28:44,613 INFO [train.py:904] (0/8) Epoch 20, batch 8000, loss[loss=0.1997, simple_loss=0.2886, pruned_loss=0.05541, over 16717.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2918, pruned_loss=0.06091, over 3074667.41 frames. ], batch size: 134, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:29:24,828 INFO [optim.py:368] (0/8) 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 03:29:28,344 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200881.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 03:29:59,864 INFO [train.py:904] (0/8) Epoch 20, batch 8050, loss[loss=0.2309, simple_loss=0.3019, pruned_loss=0.07991, over 12032.00 frames. ], tot_loss[loss=0.2068, simple_loss=0.2917, pruned_loss=0.06096, over 3053998.68 frames. ], batch size: 247, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:30:50,031 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=200935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:31:15,307 INFO [train.py:904] (0/8) Epoch 20, batch 8100, loss[loss=0.2118, simple_loss=0.3083, pruned_loss=0.05759, over 16834.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2913, pruned_loss=0.06056, over 3053451.84 frames. ], batch size: 102, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:31:49,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8057, 4.0482, 3.7588, 3.6032, 3.3140, 3.9818, 3.6629, 3.6727], device='cuda:0'), covar=tensor([0.0958, 0.0728, 0.0472, 0.0419, 0.1252, 0.0589, 0.1309, 0.0691], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0406, 0.0326, 0.0322, 0.0336, 0.0374, 0.0226, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:31:57,147 INFO [optim.py:368] (0/8) 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 03:32:04,402 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=200983.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:31,849 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:32:33,321 INFO [train.py:904] (0/8) Epoch 20, batch 8150, loss[loss=0.1824, simple_loss=0.271, pruned_loss=0.04689, over 16858.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2888, pruned_loss=0.05957, over 3054758.63 frames. ], batch size: 42, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:33:00,944 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2879, 5.2886, 5.0996, 4.4610, 5.2039, 1.7439, 4.8823, 4.8252], device='cuda:0'), covar=tensor([0.0071, 0.0066, 0.0170, 0.0352, 0.0074, 0.2671, 0.0108, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0147, 0.0189, 0.0173, 0.0167, 0.0201, 0.0180, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:33:52,113 INFO [train.py:904] (0/8) Epoch 20, batch 8200, loss[loss=0.217, simple_loss=0.2842, pruned_loss=0.07495, over 11487.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2869, pruned_loss=0.05932, over 3048219.17 frames. ], batch size: 246, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:34:24,333 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201071.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:34:36,698 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.058e+02 2.732e+02 3.390e+02 4.593e+02 1.143e+03, threshold=6.781e+02, percent-clipped=6.0 2023-05-01 03:35:15,221 INFO [train.py:904] (0/8) Epoch 20, batch 8250, loss[loss=0.1991, simple_loss=0.288, pruned_loss=0.05513, over 16750.00 frames. ], tot_loss[loss=0.199, simple_loss=0.285, pruned_loss=0.05651, over 3047181.08 frames. ], batch size: 124, lr: 3.36e-03, grad_scale: 8.0 2023-05-01 03:36:05,982 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:36:22,927 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2186, 4.2724, 4.4584, 4.2106, 4.3081, 4.8271, 4.4068, 4.0609], device='cuda:0'), covar=tensor([0.1629, 0.2209, 0.2149, 0.2139, 0.2755, 0.1157, 0.1577, 0.2632], device='cuda:0'), in_proj_covar=tensor([0.0396, 0.0577, 0.0634, 0.0477, 0.0634, 0.0666, 0.0497, 0.0646], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 03:36:38,976 INFO [train.py:904] (0/8) Epoch 20, batch 8300, loss[loss=0.1793, simple_loss=0.2787, pruned_loss=0.03992, over 15202.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2819, pruned_loss=0.0535, over 3024076.55 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:37:03,261 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 03:37:22,450 INFO [optim.py:368] (0/8) 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 03:38:00,757 INFO [train.py:904] (0/8) Epoch 20, batch 8350, loss[loss=0.208, simple_loss=0.284, pruned_loss=0.06597, over 12178.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2813, pruned_loss=0.05145, over 3027995.70 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:23,231 INFO [train.py:904] (0/8) Epoch 20, batch 8400, loss[loss=0.1864, simple_loss=0.2774, pruned_loss=0.04768, over 16722.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2783, pruned_loss=0.04887, over 3036180.27 frames. ], batch size: 124, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:39:25,952 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:00,145 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:06,024 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4023, 3.7054, 3.7109, 2.6157, 3.3302, 3.6784, 3.4348, 2.3103], device='cuda:0'), covar=tensor([0.0456, 0.0050, 0.0046, 0.0341, 0.0107, 0.0096, 0.0085, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0080, 0.0080, 0.0131, 0.0095, 0.0107, 0.0091, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 03:40:08,502 INFO [optim.py:368] (0/8) 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 03:40:44,325 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201301.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:40:45,025 INFO [train.py:904] (0/8) Epoch 20, batch 8450, loss[loss=0.1894, simple_loss=0.281, pruned_loss=0.04894, over 16701.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2766, pruned_loss=0.04724, over 3044994.84 frames. ], batch size: 124, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:40:46,058 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 03:40:53,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0513, 1.8164, 1.6690, 1.4727, 1.9891, 1.6573, 1.5755, 1.9625], device='cuda:0'), covar=tensor([0.0189, 0.0270, 0.0412, 0.0366, 0.0202, 0.0306, 0.0158, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0223, 0.0216, 0.0216, 0.0225, 0.0224, 0.0223, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:41:05,348 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:41:15,171 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9572, 2.1951, 2.3279, 2.8833, 1.7127, 3.2035, 1.7807, 2.7025], device='cuda:0'), covar=tensor([0.1225, 0.0701, 0.0961, 0.0165, 0.0084, 0.0327, 0.1528, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0171, 0.0192, 0.0183, 0.0205, 0.0210, 0.0198, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 03:41:38,825 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:02,468 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:42:07,362 INFO [train.py:904] (0/8) Epoch 20, batch 8500, loss[loss=0.1755, simple_loss=0.2605, pruned_loss=0.04527, over 16283.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2732, pruned_loss=0.045, over 3053345.49 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:42:49,516 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0625, 3.3668, 3.7647, 2.1900, 3.1461, 2.3883, 3.5933, 3.5683], device='cuda:0'), covar=tensor([0.0249, 0.0898, 0.0453, 0.1984, 0.0746, 0.0969, 0.0622, 0.0981], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0157, 0.0161, 0.0148, 0.0140, 0.0126, 0.0141, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 03:42:50,126 INFO [optim.py:368] (0/8) 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 03:43:31,053 INFO [train.py:904] (0/8) Epoch 20, batch 8550, loss[loss=0.1695, simple_loss=0.2524, pruned_loss=0.04327, over 11750.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2701, pruned_loss=0.04386, over 3014384.95 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:43:39,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3118, 2.9299, 3.1503, 1.8309, 3.2886, 3.3292, 2.7383, 2.7082], device='cuda:0'), covar=tensor([0.0675, 0.0242, 0.0195, 0.1152, 0.0097, 0.0188, 0.0422, 0.0417], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0106, 0.0094, 0.0136, 0.0077, 0.0120, 0.0124, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 03:44:18,643 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201427.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:44:45,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1002, 3.0685, 1.6414, 3.2944, 2.2633, 3.2935, 1.8469, 2.5566], device='cuda:0'), covar=tensor([0.0270, 0.0395, 0.1823, 0.0263, 0.0833, 0.0559, 0.1829, 0.0754], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0169, 0.0187, 0.0151, 0.0170, 0.0206, 0.0195, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-01 03:45:09,646 INFO [train.py:904] (0/8) Epoch 20, batch 8600, loss[loss=0.1736, simple_loss=0.2543, pruned_loss=0.04642, over 11914.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2704, pruned_loss=0.0431, over 3005880.12 frames. ], batch size: 250, lr: 3.35e-03, grad_scale: 8.0 2023-05-01 03:46:02,944 INFO [optim.py:368] (0/8) 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 03:46:31,022 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 03:46:48,563 INFO [train.py:904] (0/8) Epoch 20, batch 8650, loss[loss=0.1699, simple_loss=0.2739, pruned_loss=0.03291, over 16422.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2691, pruned_loss=0.04194, over 3016080.57 frames. ], batch size: 146, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:48:36,205 INFO [train.py:904] (0/8) Epoch 20, batch 8700, loss[loss=0.1768, simple_loss=0.2765, pruned_loss=0.03852, over 16409.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2662, pruned_loss=0.04066, over 3036076.11 frames. ], batch size: 146, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:49:28,974 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.047e+02 2.487e+02 3.171e+02 4.995e+02, threshold=4.975e+02, percent-clipped=0.0 2023-05-01 03:50:13,606 INFO [train.py:904] (0/8) Epoch 20, batch 8750, loss[loss=0.1871, simple_loss=0.2866, pruned_loss=0.04383, over 15451.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2661, pruned_loss=0.04014, over 3049823.26 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:50:31,554 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201609.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:21,448 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=201630.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:51:51,095 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201644.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:52:07,684 INFO [train.py:904] (0/8) Epoch 20, batch 8800, loss[loss=0.1733, simple_loss=0.2572, pruned_loss=0.04473, over 12600.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2642, pruned_loss=0.03911, over 3054278.51 frames. ], batch size: 249, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:52:24,885 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201661.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:52:29,964 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5333, 2.5782, 2.4782, 3.8089, 2.2036, 3.7752, 1.4225, 2.7817], device='cuda:0'), covar=tensor([0.1505, 0.0764, 0.1147, 0.0151, 0.0113, 0.0381, 0.1837, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0180, 0.0202, 0.0208, 0.0197, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 03:53:06,519 INFO [optim.py:368] (0/8) 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 03:53:45,506 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8740, 2.0981, 2.3214, 3.1524, 2.1398, 2.2481, 2.2656, 2.1414], device='cuda:0'), covar=tensor([0.1320, 0.3873, 0.2738, 0.0682, 0.4492, 0.2758, 0.3502, 0.3955], device='cuda:0'), in_proj_covar=tensor([0.0387, 0.0433, 0.0357, 0.0316, 0.0427, 0.0497, 0.0403, 0.0506], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 03:53:51,637 INFO [train.py:904] (0/8) Epoch 20, batch 8850, loss[loss=0.1511, simple_loss=0.2412, pruned_loss=0.03047, over 12532.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2673, pruned_loss=0.03871, over 3050873.90 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 03:53:57,874 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:53:59,776 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201706.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:34,766 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:54:45,912 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:55:24,566 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 03:55:38,112 INFO [train.py:904] (0/8) Epoch 20, batch 8900, loss[loss=0.1621, simple_loss=0.259, pruned_loss=0.03264, over 15381.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2673, pruned_loss=0.03785, over 3053719.51 frames. ], batch size: 191, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:56:08,880 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=201767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:27,848 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201775.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 03:56:46,041 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7966, 2.5272, 2.3043, 3.5797, 1.9498, 3.6249, 1.5385, 2.8225], device='cuda:0'), covar=tensor([0.1232, 0.0675, 0.1137, 0.0142, 0.0080, 0.0321, 0.1610, 0.0696], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0169, 0.0190, 0.0180, 0.0201, 0.0208, 0.0197, 0.0188], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 03:56:46,608 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.216e+02 2.705e+02 3.264e+02 5.822e+02, threshold=5.410e+02, percent-clipped=1.0 2023-05-01 03:57:41,717 INFO [train.py:904] (0/8) Epoch 20, batch 8950, loss[loss=0.1769, simple_loss=0.2623, pruned_loss=0.0458, over 12864.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.2673, pruned_loss=0.0382, over 3065677.63 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:58:48,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4549, 3.6482, 3.3747, 3.1934, 3.0546, 3.5453, 3.2958, 3.3431], device='cuda:0'), covar=tensor([0.0826, 0.0574, 0.0482, 0.0375, 0.1013, 0.0507, 0.2137, 0.0640], device='cuda:0'), in_proj_covar=tensor([0.0273, 0.0393, 0.0318, 0.0314, 0.0325, 0.0363, 0.0220, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-01 03:59:30,704 INFO [train.py:904] (0/8) Epoch 20, batch 9000, loss[loss=0.1696, simple_loss=0.2556, pruned_loss=0.04177, over 12076.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2639, pruned_loss=0.03702, over 3063198.17 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 03:59:30,706 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 03:59:41,112 INFO [train.py:938] (0/8) Epoch 20, validation: loss=0.1464, simple_loss=0.2502, pruned_loss=0.02125, over 944034.00 frames. 2023-05-01 03:59:41,114 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 04:00:38,724 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 04:00:41,797 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.381e+02 2.083e+02 2.552e+02 3.280e+02 1.556e+03, threshold=5.104e+02, percent-clipped=3.0 2023-05-01 04:01:24,602 INFO [train.py:904] (0/8) Epoch 20, batch 9050, loss[loss=0.1706, simple_loss=0.258, pruned_loss=0.04163, over 16192.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2643, pruned_loss=0.03746, over 3061849.16 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:01:40,930 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:02:09,309 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-01 04:02:21,220 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=201930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:03:12,554 INFO [train.py:904] (0/8) Epoch 20, batch 9100, loss[loss=0.1784, simple_loss=0.275, pruned_loss=0.0409, over 16963.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2645, pruned_loss=0.03822, over 3062093.70 frames. ], batch size: 109, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:03:22,658 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201957.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:03:30,384 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6604, 4.7021, 4.5370, 4.1941, 4.1885, 4.6620, 4.4597, 4.3176], device='cuda:0'), covar=tensor([0.0746, 0.0780, 0.0381, 0.0351, 0.1012, 0.0678, 0.0440, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0274, 0.0394, 0.0318, 0.0314, 0.0326, 0.0362, 0.0220, 0.0378], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:04:14,334 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=201978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:23,056 INFO [optim.py:368] (0/8) 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 04:04:42,917 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=201990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:04:59,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4336, 3.3489, 3.4788, 3.5472, 3.5867, 3.3043, 3.5492, 3.6183], device='cuda:0'), covar=tensor([0.1136, 0.1000, 0.1049, 0.0618, 0.0583, 0.2051, 0.0777, 0.0749], device='cuda:0'), in_proj_covar=tensor([0.0595, 0.0733, 0.0856, 0.0751, 0.0567, 0.0588, 0.0604, 0.0699], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:05:07,508 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-202000.pt 2023-05-01 04:05:11,357 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202000.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:05:12,644 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5846, 4.8904, 4.7088, 4.7102, 4.4126, 4.4102, 4.3272, 4.9312], device='cuda:0'), covar=tensor([0.1087, 0.0813, 0.0832, 0.0692, 0.0761, 0.1191, 0.1175, 0.0843], device='cuda:0'), in_proj_covar=tensor([0.0627, 0.0766, 0.0629, 0.0577, 0.0488, 0.0498, 0.0642, 0.0594], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:05:14,320 INFO [train.py:904] (0/8) Epoch 20, batch 9150, loss[loss=0.1782, simple_loss=0.2732, pruned_loss=0.04155, over 16272.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2653, pruned_loss=0.03774, over 3071652.69 frames. ], batch size: 165, lr: 3.35e-03, grad_scale: 2.0 2023-05-01 04:05:48,144 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202017.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:07,233 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7384, 4.2859, 4.2620, 3.0780, 3.7887, 4.3057, 3.9789, 2.1631], device='cuda:0'), covar=tensor([0.0559, 0.0067, 0.0062, 0.0391, 0.0124, 0.0110, 0.0083, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0078, 0.0078, 0.0129, 0.0094, 0.0104, 0.0089, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:06:43,844 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.14 vs. limit=5.0 2023-05-01 04:06:58,558 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202051.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:06:59,885 INFO [train.py:904] (0/8) Epoch 20, batch 9200, loss[loss=0.1471, simple_loss=0.2327, pruned_loss=0.03077, over 12289.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2614, pruned_loss=0.03714, over 3072200.85 frames. ], batch size: 246, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:07:19,891 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:07:55,481 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.302e+02 2.670e+02 3.574e+02 1.158e+03, threshold=5.341e+02, percent-clipped=6.0 2023-05-01 04:08:37,616 INFO [train.py:904] (0/8) Epoch 20, batch 9250, loss[loss=0.1573, simple_loss=0.2449, pruned_loss=0.03488, over 12274.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2608, pruned_loss=0.03701, over 3047212.88 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:09:51,238 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9538, 2.1504, 2.3576, 3.1774, 2.2466, 2.3252, 2.3226, 2.2135], device='cuda:0'), covar=tensor([0.1235, 0.3755, 0.2654, 0.0675, 0.4171, 0.2769, 0.3479, 0.3761], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0432, 0.0357, 0.0315, 0.0425, 0.0495, 0.0403, 0.0504], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:10:05,796 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5708, 2.9195, 3.2468, 1.9417, 2.8130, 2.1246, 3.1550, 3.1217], device='cuda:0'), covar=tensor([0.0328, 0.0946, 0.0501, 0.2097, 0.0810, 0.1005, 0.0713, 0.1073], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0153, 0.0158, 0.0146, 0.0138, 0.0124, 0.0138, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 04:10:29,972 INFO [train.py:904] (0/8) Epoch 20, batch 9300, loss[loss=0.1659, simple_loss=0.2508, pruned_loss=0.04054, over 16726.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2597, pruned_loss=0.0367, over 3066153.11 frames. ], batch size: 134, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:10:56,344 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9733, 4.2514, 4.0550, 4.1295, 3.7884, 3.8201, 3.8399, 4.2309], device='cuda:0'), covar=tensor([0.1096, 0.0877, 0.1004, 0.0761, 0.0834, 0.1702, 0.0975, 0.0997], device='cuda:0'), in_proj_covar=tensor([0.0632, 0.0769, 0.0633, 0.0579, 0.0491, 0.0500, 0.0645, 0.0597], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:11:01,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4084, 2.8336, 3.1461, 1.9940, 2.7644, 2.1019, 3.0819, 3.0915], device='cuda:0'), covar=tensor([0.0287, 0.0852, 0.0494, 0.1906, 0.0796, 0.1004, 0.0608, 0.0788], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0153, 0.0158, 0.0146, 0.0138, 0.0124, 0.0138, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 04:11:37,349 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.443e+02 1.963e+02 2.278e+02 2.874e+02 5.860e+02, threshold=4.556e+02, percent-clipped=2.0 2023-05-01 04:12:16,149 INFO [train.py:904] (0/8) Epoch 20, batch 9350, loss[loss=0.1717, simple_loss=0.261, pruned_loss=0.04119, over 12207.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2598, pruned_loss=0.03651, over 3067314.92 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:13:59,190 INFO [train.py:904] (0/8) Epoch 20, batch 9400, loss[loss=0.1431, simple_loss=0.2306, pruned_loss=0.0278, over 12629.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2594, pruned_loss=0.03632, over 3048394.17 frames. ], batch size: 247, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:15:00,976 INFO [optim.py:368] (0/8) 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 04:15:39,467 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:15:41,534 INFO [train.py:904] (0/8) Epoch 20, batch 9450, loss[loss=0.1749, simple_loss=0.2716, pruned_loss=0.03915, over 12452.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2609, pruned_loss=0.03669, over 3024740.06 frames. ], batch size: 248, lr: 3.35e-03, grad_scale: 4.0 2023-05-01 04:16:10,381 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5593, 4.8358, 4.6048, 4.6321, 4.3665, 4.3438, 4.2581, 4.9055], device='cuda:0'), covar=tensor([0.1294, 0.1016, 0.1135, 0.0897, 0.0832, 0.1248, 0.1298, 0.0936], device='cuda:0'), in_proj_covar=tensor([0.0628, 0.0762, 0.0626, 0.0575, 0.0487, 0.0496, 0.0639, 0.0593], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:16:13,015 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202317.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:14,533 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=202346.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:17,848 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202348.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:25,753 INFO [train.py:904] (0/8) Epoch 20, batch 9500, loss[loss=0.1834, simple_loss=0.2789, pruned_loss=0.04391, over 15415.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2605, pruned_loss=0.03633, over 3040624.51 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:17:49,556 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202362.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:17:54,705 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:18:26,630 INFO [optim.py:368] (0/8) 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 04:19:13,224 INFO [train.py:904] (0/8) Epoch 20, batch 9550, loss[loss=0.1664, simple_loss=0.2584, pruned_loss=0.03714, over 12242.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2604, pruned_loss=0.03652, over 3031141.07 frames. ], batch size: 248, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:19:32,645 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202410.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:20:55,241 INFO [train.py:904] (0/8) Epoch 20, batch 9600, loss[loss=0.1467, simple_loss=0.2487, pruned_loss=0.02237, over 17145.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2617, pruned_loss=0.03675, over 3043797.77 frames. ], batch size: 49, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:20:55,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5670, 4.3949, 4.5954, 4.7497, 4.9444, 4.4564, 4.9170, 4.9328], device='cuda:0'), covar=tensor([0.1836, 0.1181, 0.1618, 0.0741, 0.0492, 0.0874, 0.0530, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0592, 0.0727, 0.0848, 0.0745, 0.0564, 0.0582, 0.0601, 0.0695], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:21:53,899 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.693e+02 2.243e+02 2.599e+02 3.368e+02 6.147e+02, threshold=5.198e+02, percent-clipped=3.0 2023-05-01 04:22:41,355 INFO [train.py:904] (0/8) Epoch 20, batch 9650, loss[loss=0.1615, simple_loss=0.2612, pruned_loss=0.03086, over 15442.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2632, pruned_loss=0.03702, over 3044432.39 frames. ], batch size: 191, lr: 3.34e-03, grad_scale: 8.0 2023-05-01 04:23:40,551 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 04:23:50,355 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2668, 3.9535, 4.5503, 2.2241, 4.7118, 4.7939, 3.4021, 3.5411], device='cuda:0'), covar=tensor([0.0642, 0.0265, 0.0173, 0.1157, 0.0068, 0.0086, 0.0376, 0.0392], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0105, 0.0092, 0.0136, 0.0076, 0.0117, 0.0124, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 04:24:31,601 INFO [train.py:904] (0/8) Epoch 20, batch 9700, loss[loss=0.1727, simple_loss=0.2698, pruned_loss=0.03784, over 16221.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2624, pruned_loss=0.03664, over 3052644.99 frames. ], batch size: 165, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:24:37,253 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8019, 5.0242, 5.1485, 4.9917, 5.0212, 5.5524, 5.0615, 4.7261], device='cuda:0'), covar=tensor([0.1039, 0.2092, 0.2455, 0.1870, 0.2703, 0.0896, 0.1635, 0.2552], device='cuda:0'), in_proj_covar=tensor([0.0376, 0.0552, 0.0607, 0.0455, 0.0605, 0.0636, 0.0476, 0.0611], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:25:37,517 INFO [optim.py:368] (0/8) 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 04:26:13,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8554, 3.8221, 3.9859, 3.8033, 3.8985, 4.3317, 3.9770, 3.6791], device='cuda:0'), covar=tensor([0.1998, 0.2604, 0.2576, 0.2350, 0.2809, 0.1533, 0.1767, 0.2703], device='cuda:0'), in_proj_covar=tensor([0.0377, 0.0552, 0.0607, 0.0456, 0.0606, 0.0637, 0.0477, 0.0612], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:26:15,978 INFO [train.py:904] (0/8) Epoch 20, batch 9750, loss[loss=0.1644, simple_loss=0.2609, pruned_loss=0.03395, over 16930.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2615, pruned_loss=0.0373, over 3039548.74 frames. ], batch size: 109, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:27:17,779 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5448, 4.3598, 4.3423, 4.7244, 4.9009, 4.4910, 4.8549, 4.8736], device='cuda:0'), covar=tensor([0.1813, 0.1399, 0.2365, 0.0951, 0.0816, 0.1143, 0.0977, 0.1068], device='cuda:0'), in_proj_covar=tensor([0.0590, 0.0725, 0.0844, 0.0745, 0.0563, 0.0582, 0.0599, 0.0694], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:27:47,508 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=202646.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:27:57,091 INFO [train.py:904] (0/8) Epoch 20, batch 9800, loss[loss=0.1838, simple_loss=0.2877, pruned_loss=0.03998, over 16669.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2622, pruned_loss=0.0365, over 3055809.50 frames. ], batch size: 134, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:28:57,878 INFO [optim.py:368] (0/8) 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 04:29:12,245 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8580, 4.8577, 4.5804, 4.0559, 4.7066, 1.8163, 4.4853, 4.4170], device='cuda:0'), covar=tensor([0.0085, 0.0070, 0.0195, 0.0286, 0.0087, 0.2566, 0.0111, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0153, 0.0144, 0.0183, 0.0164, 0.0163, 0.0197, 0.0174, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:29:25,964 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=202694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:29:42,360 INFO [train.py:904] (0/8) Epoch 20, batch 9850, loss[loss=0.1785, simple_loss=0.2694, pruned_loss=0.0438, over 16943.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2635, pruned_loss=0.03603, over 3071469.81 frames. ], batch size: 116, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:30:21,811 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9556, 2.2227, 1.8787, 1.9686, 2.5927, 2.2089, 2.4027, 2.7356], device='cuda:0'), covar=tensor([0.0146, 0.0454, 0.0547, 0.0519, 0.0285, 0.0462, 0.0181, 0.0252], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0224, 0.0216, 0.0217, 0.0224, 0.0224, 0.0219, 0.0217], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:31:34,889 INFO [train.py:904] (0/8) Epoch 20, batch 9900, loss[loss=0.1638, simple_loss=0.2479, pruned_loss=0.03979, over 12248.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2637, pruned_loss=0.03614, over 3069353.35 frames. ], batch size: 248, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:32:03,103 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4196, 2.9293, 3.1476, 1.9431, 2.7297, 2.1273, 3.0264, 3.0913], device='cuda:0'), covar=tensor([0.0292, 0.0897, 0.0480, 0.1994, 0.0821, 0.1013, 0.0712, 0.0932], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0153, 0.0159, 0.0147, 0.0138, 0.0124, 0.0139, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 04:32:49,194 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.306e+02 2.190e+02 2.483e+02 3.109e+02 6.841e+02, threshold=4.966e+02, percent-clipped=3.0 2023-05-01 04:33:30,684 INFO [train.py:904] (0/8) Epoch 20, batch 9950, loss[loss=0.1739, simple_loss=0.2727, pruned_loss=0.03755, over 16998.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2662, pruned_loss=0.03671, over 3074613.34 frames. ], batch size: 109, lr: 3.34e-03, grad_scale: 2.0 2023-05-01 04:33:54,964 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9735, 3.8400, 4.0378, 4.1417, 4.2650, 3.8297, 4.2331, 4.2686], device='cuda:0'), covar=tensor([0.1706, 0.1064, 0.1280, 0.0707, 0.0590, 0.1540, 0.0672, 0.0697], device='cuda:0'), in_proj_covar=tensor([0.0593, 0.0727, 0.0847, 0.0748, 0.0564, 0.0583, 0.0603, 0.0696], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:34:41,368 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=202830.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:35:32,310 INFO [train.py:904] (0/8) Epoch 20, batch 10000, loss[loss=0.1958, simple_loss=0.2929, pruned_loss=0.04929, over 15393.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2649, pruned_loss=0.03633, over 3084084.79 frames. ], batch size: 192, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:36:01,855 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8455, 2.6679, 2.6343, 1.9166, 2.6355, 2.8338, 2.6761, 1.7394], device='cuda:0'), covar=tensor([0.0490, 0.0086, 0.0078, 0.0409, 0.0117, 0.0108, 0.0099, 0.0568], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0078, 0.0077, 0.0129, 0.0094, 0.0103, 0.0088, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:36:36,181 INFO [optim.py:368] (0/8) 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 04:36:51,652 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=202891.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 04:37:11,284 INFO [train.py:904] (0/8) Epoch 20, batch 10050, loss[loss=0.1683, simple_loss=0.2684, pruned_loss=0.0341, over 16505.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.265, pruned_loss=0.03628, over 3080735.79 frames. ], batch size: 68, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:38:41,715 INFO [train.py:904] (0/8) Epoch 20, batch 10100, loss[loss=0.1708, simple_loss=0.2486, pruned_loss=0.04652, over 12528.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2654, pruned_loss=0.03685, over 3069083.40 frames. ], batch size: 248, lr: 3.34e-03, grad_scale: 4.0 2023-05-01 04:39:29,704 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6649, 3.0182, 3.3032, 1.9960, 2.7602, 2.0699, 3.1777, 3.1968], device='cuda:0'), covar=tensor([0.0277, 0.0896, 0.0503, 0.2015, 0.0860, 0.1000, 0.0658, 0.0981], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0152, 0.0159, 0.0147, 0.0138, 0.0124, 0.0138, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 04:39:40,384 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.271e+02 2.682e+02 3.301e+02 6.522e+02, threshold=5.364e+02, percent-clipped=2.0 2023-05-01 04:40:00,142 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-20.pt 2023-05-01 04:40:23,131 INFO [train.py:904] (0/8) Epoch 21, batch 0, loss[loss=0.1736, simple_loss=0.2543, pruned_loss=0.04643, over 16793.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2543, pruned_loss=0.04643, over 16793.00 frames. ], batch size: 39, lr: 3.26e-03, grad_scale: 8.0 2023-05-01 04:40:23,132 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 04:40:30,893 INFO [train.py:938] (0/8) Epoch 21, validation: loss=0.1451, simple_loss=0.2491, pruned_loss=0.02058, over 944034.00 frames. 2023-05-01 04:40:30,894 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 04:40:49,566 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3367, 4.4729, 4.6245, 4.3869, 4.4040, 5.0038, 4.6037, 4.2605], device='cuda:0'), covar=tensor([0.1705, 0.2132, 0.2203, 0.2247, 0.2657, 0.1173, 0.1593, 0.2551], device='cuda:0'), in_proj_covar=tensor([0.0373, 0.0549, 0.0604, 0.0456, 0.0604, 0.0637, 0.0474, 0.0610], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:41:11,107 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4981, 4.9028, 4.4399, 4.7349, 4.4655, 4.3453, 4.5060, 4.9571], device='cuda:0'), covar=tensor([0.2514, 0.1866, 0.3045, 0.1668, 0.1966, 0.2218, 0.2580, 0.2121], device='cuda:0'), in_proj_covar=tensor([0.0621, 0.0760, 0.0622, 0.0570, 0.0483, 0.0493, 0.0638, 0.0589], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:41:19,648 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:41:21,280 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 04:41:36,915 INFO [train.py:904] (0/8) Epoch 21, batch 50, loss[loss=0.1479, simple_loss=0.2365, pruned_loss=0.0296, over 16983.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2722, pruned_loss=0.05028, over 754272.46 frames. ], batch size: 41, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:41:46,964 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 04:42:24,761 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 04:42:43,815 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203099.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:42:48,528 INFO [train.py:904] (0/8) Epoch 21, batch 100, loss[loss=0.2073, simple_loss=0.2944, pruned_loss=0.06005, over 15579.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2697, pruned_loss=0.04845, over 1324225.82 frames. ], batch size: 191, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:43:49,572 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:43:57,346 INFO [train.py:904] (0/8) Epoch 21, batch 150, loss[loss=0.1783, simple_loss=0.2768, pruned_loss=0.03987, over 17025.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.266, pruned_loss=0.04614, over 1770381.59 frames. ], batch size: 50, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:44:17,140 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-05-01 04:44:32,515 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7066, 3.7348, 2.8820, 2.3038, 2.4084, 2.2991, 3.8338, 3.3006], device='cuda:0'), covar=tensor([0.2570, 0.0651, 0.1558, 0.2914, 0.2636, 0.2026, 0.0478, 0.1510], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0263, 0.0297, 0.0303, 0.0285, 0.0251, 0.0287, 0.0326], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:44:44,683 INFO [optim.py:368] (0/8) 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 04:44:44,949 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203186.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 04:44:46,755 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-01 04:45:04,861 INFO [train.py:904] (0/8) Epoch 21, batch 200, loss[loss=0.1841, simple_loss=0.2818, pruned_loss=0.04315, over 16619.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2659, pruned_loss=0.04689, over 2113988.53 frames. ], batch size: 57, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:45:25,327 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3150, 5.2696, 5.1378, 4.6745, 4.8465, 5.1710, 5.2415, 4.8262], device='cuda:0'), covar=tensor([0.0663, 0.0505, 0.0332, 0.0345, 0.0997, 0.0421, 0.0279, 0.0745], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0399, 0.0323, 0.0318, 0.0329, 0.0368, 0.0221, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 04:46:15,304 INFO [train.py:904] (0/8) Epoch 21, batch 250, loss[loss=0.1628, simple_loss=0.2595, pruned_loss=0.03305, over 17130.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2635, pruned_loss=0.0462, over 2378178.24 frames. ], batch size: 47, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:46:29,571 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 04:47:00,833 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.546e+02 2.163e+02 2.580e+02 3.064e+02 7.723e+02, threshold=5.160e+02, percent-clipped=4.0 2023-05-01 04:47:20,241 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2926, 3.4915, 3.8492, 2.1698, 2.9696, 2.3412, 3.7660, 3.6529], device='cuda:0'), covar=tensor([0.0314, 0.0976, 0.0530, 0.2058, 0.0928, 0.1031, 0.0656, 0.1200], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0157, 0.0162, 0.0150, 0.0141, 0.0127, 0.0141, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 04:47:23,340 INFO [train.py:904] (0/8) Epoch 21, batch 300, loss[loss=0.1553, simple_loss=0.246, pruned_loss=0.03233, over 17192.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2607, pruned_loss=0.0452, over 2585032.10 frames. ], batch size: 44, lr: 3.26e-03, grad_scale: 2.0 2023-05-01 04:48:32,613 INFO [train.py:904] (0/8) Epoch 21, batch 350, loss[loss=0.1713, simple_loss=0.2634, pruned_loss=0.03957, over 16526.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2584, pruned_loss=0.04417, over 2753746.39 frames. ], batch size: 62, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:49:17,945 INFO [optim.py:368] (0/8) 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 04:49:30,104 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203394.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:49:41,963 INFO [train.py:904] (0/8) Epoch 21, batch 400, loss[loss=0.1855, simple_loss=0.2659, pruned_loss=0.05254, over 16293.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2564, pruned_loss=0.04327, over 2878659.77 frames. ], batch size: 165, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:50:36,383 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203441.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:50:47,461 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1875, 5.8089, 5.9248, 5.6860, 5.7752, 6.3378, 5.8109, 5.5121], device='cuda:0'), covar=tensor([0.0934, 0.1904, 0.2362, 0.2106, 0.2603, 0.0973, 0.1627, 0.2348], device='cuda:0'), in_proj_covar=tensor([0.0395, 0.0578, 0.0636, 0.0477, 0.0637, 0.0669, 0.0500, 0.0639], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:50:50,735 INFO [train.py:904] (0/8) Epoch 21, batch 450, loss[loss=0.1888, simple_loss=0.2853, pruned_loss=0.04611, over 17045.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2556, pruned_loss=0.0432, over 2985532.62 frames. ], batch size: 50, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:50:55,497 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6853, 3.7954, 2.9032, 2.2258, 2.3980, 2.3196, 3.8540, 3.2446], device='cuda:0'), covar=tensor([0.2767, 0.0564, 0.1682, 0.3042, 0.2722, 0.2166, 0.0503, 0.1593], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0267, 0.0303, 0.0308, 0.0291, 0.0256, 0.0291, 0.0333], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 04:51:11,877 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:51:37,696 INFO [optim.py:368] (0/8) 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 04:51:38,068 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203486.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 04:51:44,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0152, 5.0577, 5.4993, 5.4621, 5.5052, 5.1631, 5.0700, 4.8648], device='cuda:0'), covar=tensor([0.0362, 0.0531, 0.0416, 0.0453, 0.0541, 0.0353, 0.0968, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0439, 0.0427, 0.0396, 0.0472, 0.0448, 0.0533, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 04:51:59,114 INFO [train.py:904] (0/8) Epoch 21, batch 500, loss[loss=0.1581, simple_loss=0.2434, pruned_loss=0.03639, over 15439.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2541, pruned_loss=0.04268, over 3055192.53 frames. ], batch size: 190, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:52:20,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9705, 3.9954, 2.5857, 4.6339, 3.2536, 4.5918, 2.6745, 3.4375], device='cuda:0'), covar=tensor([0.0286, 0.0399, 0.1469, 0.0247, 0.0754, 0.0514, 0.1504, 0.0658], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0159, 0.0176, 0.0213, 0.0204, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 04:52:36,256 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203528.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:52:43,064 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203534.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 04:53:08,047 INFO [train.py:904] (0/8) Epoch 21, batch 550, loss[loss=0.1551, simple_loss=0.2426, pruned_loss=0.03384, over 17190.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2526, pruned_loss=0.0415, over 3118999.33 frames. ], batch size: 44, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:53:56,914 INFO [optim.py:368] (0/8) 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 04:54:17,595 INFO [train.py:904] (0/8) Epoch 21, batch 600, loss[loss=0.1717, simple_loss=0.2682, pruned_loss=0.03762, over 17110.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2525, pruned_loss=0.04179, over 3162697.92 frames. ], batch size: 49, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:55:27,195 INFO [train.py:904] (0/8) Epoch 21, batch 650, loss[loss=0.1484, simple_loss=0.2315, pruned_loss=0.03267, over 17029.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2512, pruned_loss=0.0412, over 3203337.10 frames. ], batch size: 41, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:56:14,242 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.016e+02 2.395e+02 3.216e+02 7.414e+02, threshold=4.790e+02, percent-clipped=2.0 2023-05-01 04:56:24,849 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:56:34,647 INFO [train.py:904] (0/8) Epoch 21, batch 700, loss[loss=0.1801, simple_loss=0.2566, pruned_loss=0.05175, over 16524.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2513, pruned_loss=0.04115, over 3229908.30 frames. ], batch size: 146, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:57:30,279 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=203741.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:31,299 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:37,736 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203747.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:57:44,453 INFO [train.py:904] (0/8) Epoch 21, batch 750, loss[loss=0.2023, simple_loss=0.2692, pruned_loss=0.06766, over 16853.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.252, pruned_loss=0.04136, over 3257288.03 frames. ], batch size: 116, lr: 3.25e-03, grad_scale: 2.0 2023-05-01 04:58:03,772 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:33,074 INFO [optim.py:368] (0/8) 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 04:58:36,278 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=203789.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:58:54,764 INFO [train.py:904] (0/8) Epoch 21, batch 800, loss[loss=0.1538, simple_loss=0.2466, pruned_loss=0.03051, over 17120.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2514, pruned_loss=0.04116, over 3269839.04 frames. ], batch size: 48, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 04:59:03,100 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203808.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:24,742 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=203823.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:30,071 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 04:59:58,180 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:00:04,935 INFO [train.py:904] (0/8) Epoch 21, batch 850, loss[loss=0.1703, simple_loss=0.2674, pruned_loss=0.03657, over 17105.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2513, pruned_loss=0.04082, over 3284716.02 frames. ], batch size: 49, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:00:53,313 INFO [optim.py:368] (0/8) 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 05:01:05,134 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 05:01:11,193 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=203900.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:01:12,971 INFO [train.py:904] (0/8) Epoch 21, batch 900, loss[loss=0.1823, simple_loss=0.2714, pruned_loss=0.04658, over 17062.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2509, pruned_loss=0.03996, over 3290702.49 frames. ], batch size: 53, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:01:21,653 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203908.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:01:43,713 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7941, 2.7402, 2.4506, 2.7556, 3.0965, 2.9078, 3.4921, 3.2915], device='cuda:0'), covar=tensor([0.0123, 0.0409, 0.0488, 0.0393, 0.0281, 0.0395, 0.0194, 0.0277], device='cuda:0'), in_proj_covar=tensor([0.0207, 0.0238, 0.0228, 0.0228, 0.0237, 0.0236, 0.0238, 0.0233], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:02:11,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4252, 4.1668, 4.3573, 4.6270, 4.7416, 4.3553, 4.6821, 4.7170], device='cuda:0'), covar=tensor([0.1731, 0.1546, 0.1960, 0.0916, 0.0869, 0.1179, 0.1917, 0.1126], device='cuda:0'), in_proj_covar=tensor([0.0646, 0.0789, 0.0921, 0.0809, 0.0607, 0.0629, 0.0652, 0.0752], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:02:21,435 INFO [train.py:904] (0/8) Epoch 21, batch 950, loss[loss=0.1708, simple_loss=0.261, pruned_loss=0.04033, over 17136.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2511, pruned_loss=0.04004, over 3299217.58 frames. ], batch size: 47, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:02:35,118 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=203961.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:03:10,189 INFO [optim.py:368] (0/8) 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 05:03:28,280 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-204000.pt 2023-05-01 05:03:34,214 INFO [train.py:904] (0/8) Epoch 21, batch 1000, loss[loss=0.1637, simple_loss=0.2423, pruned_loss=0.04252, over 16445.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2499, pruned_loss=0.04019, over 3308274.27 frames. ], batch size: 68, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:03:57,566 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 05:04:01,498 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9172, 2.6804, 2.9168, 2.1594, 2.7128, 2.2090, 2.7456, 2.8601], device='cuda:0'), covar=tensor([0.0294, 0.0898, 0.0458, 0.1705, 0.0758, 0.0849, 0.0563, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0162, 0.0166, 0.0153, 0.0144, 0.0130, 0.0144, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:04:43,751 INFO [train.py:904] (0/8) Epoch 21, batch 1050, loss[loss=0.1929, simple_loss=0.2647, pruned_loss=0.06051, over 16884.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2497, pruned_loss=0.04043, over 3308196.26 frames. ], batch size: 116, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:04:52,378 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8165, 3.9733, 3.0142, 2.3390, 2.6472, 2.4339, 4.2765, 3.4705], device='cuda:0'), covar=tensor([0.2763, 0.0710, 0.1772, 0.2910, 0.2829, 0.2097, 0.0474, 0.1473], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0269, 0.0303, 0.0309, 0.0292, 0.0257, 0.0293, 0.0335], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 05:05:19,328 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204077.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:05:34,281 INFO [optim.py:368] (0/8) 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 05:05:42,754 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5872, 4.4431, 4.4635, 4.1520, 4.2355, 4.4748, 4.2722, 4.2621], device='cuda:0'), covar=tensor([0.0553, 0.0672, 0.0290, 0.0277, 0.0738, 0.0507, 0.0542, 0.0623], device='cuda:0'), in_proj_covar=tensor([0.0295, 0.0422, 0.0341, 0.0337, 0.0349, 0.0393, 0.0234, 0.0410], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:05:48,440 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2123, 5.7069, 5.8077, 5.4883, 5.6052, 6.1904, 5.6560, 5.3131], device='cuda:0'), covar=tensor([0.0885, 0.2079, 0.2574, 0.2026, 0.2661, 0.0926, 0.1626, 0.2253], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0593, 0.0653, 0.0490, 0.0655, 0.0683, 0.0514, 0.0658], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 05:05:54,731 INFO [train.py:904] (0/8) Epoch 21, batch 1100, loss[loss=0.1683, simple_loss=0.2421, pruned_loss=0.04727, over 16485.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2494, pruned_loss=0.04002, over 3307719.72 frames. ], batch size: 75, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:05:56,315 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:24,058 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:25,301 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204123.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:06:28,478 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7367, 4.0547, 4.2835, 2.1434, 4.5732, 4.7694, 3.3211, 3.3933], device='cuda:0'), covar=tensor([0.1037, 0.0198, 0.0225, 0.1291, 0.0086, 0.0130, 0.0445, 0.0522], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0108, 0.0096, 0.0139, 0.0079, 0.0122, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:06:45,596 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204138.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:07:05,265 INFO [train.py:904] (0/8) Epoch 21, batch 1150, loss[loss=0.1827, simple_loss=0.2771, pruned_loss=0.04413, over 17057.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2492, pruned_loss=0.03986, over 3314484.47 frames. ], batch size: 55, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:07:20,270 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=204163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:07:28,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0075, 3.9520, 2.5797, 4.7028, 3.2287, 4.6547, 2.5408, 3.4308], device='cuda:0'), covar=tensor([0.0256, 0.0410, 0.1470, 0.0228, 0.0731, 0.0457, 0.1478, 0.0622], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0177, 0.0194, 0.0162, 0.0177, 0.0214, 0.0203, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:07:31,663 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 05:08:01,367 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6911, 2.8101, 2.9284, 4.9499, 4.0901, 4.3644, 1.7529, 3.0075], device='cuda:0'), covar=tensor([0.1410, 0.0830, 0.1104, 0.0199, 0.0258, 0.0397, 0.1588, 0.0858], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0174, 0.0193, 0.0187, 0.0204, 0.0215, 0.0201, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:08:13,782 INFO [train.py:904] (0/8) Epoch 21, batch 1200, loss[loss=0.1562, simple_loss=0.2488, pruned_loss=0.03181, over 17166.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2489, pruned_loss=0.03919, over 3319463.60 frames. ], batch size: 46, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:08:15,280 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204203.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 05:08:44,949 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204224.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:09:24,353 INFO [train.py:904] (0/8) Epoch 21, batch 1250, loss[loss=0.1461, simple_loss=0.2384, pruned_loss=0.02692, over 17206.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2481, pruned_loss=0.03934, over 3323616.59 frames. ], batch size: 45, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:09:30,092 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204256.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:10:12,300 INFO [optim.py:368] (0/8) 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 05:10:30,682 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7692, 5.1246, 4.9030, 4.9038, 4.6711, 4.6515, 4.5875, 5.2024], device='cuda:0'), covar=tensor([0.1239, 0.0889, 0.0991, 0.0887, 0.0818, 0.1069, 0.1152, 0.0861], device='cuda:0'), in_proj_covar=tensor([0.0681, 0.0833, 0.0681, 0.0626, 0.0529, 0.0537, 0.0700, 0.0646], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:10:33,949 INFO [train.py:904] (0/8) Epoch 21, batch 1300, loss[loss=0.1646, simple_loss=0.2487, pruned_loss=0.04021, over 16874.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2487, pruned_loss=0.03964, over 3331734.15 frames. ], batch size: 96, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:10:54,942 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2530, 4.0797, 4.3487, 4.4532, 4.5539, 4.0869, 4.3277, 4.5269], device='cuda:0'), covar=tensor([0.1598, 0.1085, 0.1188, 0.0654, 0.0554, 0.1411, 0.2162, 0.0754], device='cuda:0'), in_proj_covar=tensor([0.0648, 0.0793, 0.0927, 0.0815, 0.0610, 0.0634, 0.0655, 0.0756], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:11:42,908 INFO [train.py:904] (0/8) Epoch 21, batch 1350, loss[loss=0.169, simple_loss=0.2603, pruned_loss=0.03884, over 17108.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.249, pruned_loss=0.03998, over 3325723.45 frames. ], batch size: 48, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:12:31,607 INFO [optim.py:368] (0/8) 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 05:12:41,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2721, 2.2691, 2.4431, 4.0631, 2.2711, 2.6176, 2.3314, 2.4348], device='cuda:0'), covar=tensor([0.1411, 0.3672, 0.2918, 0.0644, 0.3885, 0.2657, 0.3853, 0.3213], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0448, 0.0370, 0.0331, 0.0437, 0.0515, 0.0419, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:12:52,518 INFO [train.py:904] (0/8) Epoch 21, batch 1400, loss[loss=0.1716, simple_loss=0.254, pruned_loss=0.04463, over 16926.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2489, pruned_loss=0.03973, over 3329772.69 frames. ], batch size: 90, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:12:54,506 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204403.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:20,216 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204422.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:13:35,313 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=204433.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:14:00,497 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204451.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:01,409 INFO [train.py:904] (0/8) Epoch 21, batch 1450, loss[loss=0.1758, simple_loss=0.2507, pruned_loss=0.05046, over 16715.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2476, pruned_loss=0.03913, over 3332123.65 frames. ], batch size: 89, lr: 3.25e-03, grad_scale: 8.0 2023-05-01 05:14:26,858 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204470.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:14:50,909 INFO [optim.py:368] (0/8) 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 05:14:51,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4584, 3.5979, 3.9176, 2.1844, 3.1110, 2.4669, 3.8161, 3.7763], device='cuda:0'), covar=tensor([0.0256, 0.0893, 0.0501, 0.1971, 0.0808, 0.0972, 0.0603, 0.1159], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0163, 0.0167, 0.0153, 0.0145, 0.0130, 0.0145, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:15:10,561 INFO [train.py:904] (0/8) Epoch 21, batch 1500, loss[loss=0.1784, simple_loss=0.2471, pruned_loss=0.05487, over 16854.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2474, pruned_loss=0.03949, over 3320547.72 frames. ], batch size: 96, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:15:12,682 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:15:33,861 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:16:18,110 INFO [train.py:904] (0/8) Epoch 21, batch 1550, loss[loss=0.199, simple_loss=0.2673, pruned_loss=0.06536, over 16922.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2489, pruned_loss=0.0401, over 3324813.88 frames. ], batch size: 116, lr: 3.25e-03, grad_scale: 4.0 2023-05-01 05:16:22,975 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204556.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:16:28,120 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6266, 4.6575, 4.9814, 5.0120, 5.0384, 4.7123, 4.7046, 4.4596], device='cuda:0'), covar=tensor([0.0392, 0.0725, 0.0528, 0.0409, 0.0575, 0.0444, 0.0889, 0.0744], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0455, 0.0441, 0.0410, 0.0490, 0.0465, 0.0553, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 05:16:52,107 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-01 05:17:07,002 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.391e+02 2.852e+02 3.354e+02 7.624e+02, threshold=5.704e+02, percent-clipped=5.0 2023-05-01 05:17:26,283 INFO [train.py:904] (0/8) Epoch 21, batch 1600, loss[loss=0.1887, simple_loss=0.2673, pruned_loss=0.055, over 16849.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2508, pruned_loss=0.04128, over 3312374.58 frames. ], batch size: 96, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:17:28,785 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204604.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 05:17:29,369 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 05:17:30,829 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7454, 4.1580, 4.2088, 1.9288, 4.4550, 4.7657, 3.4345, 3.2677], device='cuda:0'), covar=tensor([0.1078, 0.0185, 0.0256, 0.1495, 0.0163, 0.0170, 0.0447, 0.0598], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0108, 0.0096, 0.0138, 0.0079, 0.0122, 0.0127, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:17:53,945 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7238, 3.5134, 3.8703, 2.0016, 4.0115, 4.0033, 3.2197, 2.9733], device='cuda:0'), covar=tensor([0.0720, 0.0242, 0.0169, 0.1180, 0.0103, 0.0177, 0.0370, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0108, 0.0096, 0.0138, 0.0079, 0.0122, 0.0127, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 05:18:35,828 INFO [train.py:904] (0/8) Epoch 21, batch 1650, loss[loss=0.1851, simple_loss=0.2625, pruned_loss=0.05381, over 16511.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.252, pruned_loss=0.04157, over 3314115.69 frames. ], batch size: 146, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:19:25,766 INFO [optim.py:368] (0/8) 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 05:19:45,547 INFO [train.py:904] (0/8) Epoch 21, batch 1700, loss[loss=0.1753, simple_loss=0.2675, pruned_loss=0.0415, over 17064.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2544, pruned_loss=0.04253, over 3304406.56 frames. ], batch size: 53, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:20:29,947 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204733.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:20:38,317 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2250, 5.1573, 4.9664, 4.0662, 4.9786, 1.7971, 4.6844, 4.7974], device='cuda:0'), covar=tensor([0.0117, 0.0105, 0.0247, 0.0607, 0.0147, 0.3166, 0.0184, 0.0291], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0155, 0.0197, 0.0178, 0.0175, 0.0209, 0.0188, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:20:55,690 INFO [train.py:904] (0/8) Epoch 21, batch 1750, loss[loss=0.1663, simple_loss=0.2632, pruned_loss=0.03465, over 17115.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.255, pruned_loss=0.04191, over 3311749.40 frames. ], batch size: 48, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:21:31,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7398, 2.8287, 2.5838, 4.4257, 3.5282, 4.0730, 1.6847, 3.0156], device='cuda:0'), covar=tensor([0.1414, 0.0733, 0.1204, 0.0187, 0.0201, 0.0411, 0.1602, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0175, 0.0195, 0.0190, 0.0205, 0.0217, 0.0202, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:21:37,076 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204781.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:21:46,296 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.363e+02 2.773e+02 3.345e+02 7.836e+02, threshold=5.546e+02, percent-clipped=6.0 2023-05-01 05:22:06,995 INFO [train.py:904] (0/8) Epoch 21, batch 1800, loss[loss=0.1874, simple_loss=0.2699, pruned_loss=0.05252, over 16855.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2572, pruned_loss=0.04244, over 3299365.30 frames. ], batch size: 102, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:22:30,051 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=204819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:22:32,922 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 21, batch 1850, loss[loss=0.1525, simple_loss=0.2499, pruned_loss=0.02756, over 17150.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2577, pruned_loss=0.04266, over 3307107.74 frames. ], batch size: 48, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:23:37,453 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=204867.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:23:57,931 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=204882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:24:06,223 INFO [optim.py:368] (0/8) 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 05:24:11,640 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 05:24:26,175 INFO [train.py:904] (0/8) Epoch 21, batch 1900, loss[loss=0.1549, simple_loss=0.2416, pruned_loss=0.03409, over 17224.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2562, pruned_loss=0.04153, over 3314193.08 frames. ], batch size: 45, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:24:30,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0896, 3.2043, 3.3216, 2.2948, 3.0689, 3.4105, 3.1540, 1.9332], device='cuda:0'), covar=tensor([0.0530, 0.0146, 0.0067, 0.0412, 0.0138, 0.0108, 0.0105, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0084, 0.0083, 0.0135, 0.0099, 0.0110, 0.0094, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0004], device='cuda:0') 2023-05-01 05:25:35,966 INFO [train.py:904] (0/8) Epoch 21, batch 1950, loss[loss=0.1603, simple_loss=0.252, pruned_loss=0.03427, over 17198.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2559, pruned_loss=0.04114, over 3311575.00 frames. ], batch size: 46, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:26:26,418 INFO [optim.py:368] (0/8) 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 05:26:44,690 INFO [train.py:904] (0/8) Epoch 21, batch 2000, loss[loss=0.1816, simple_loss=0.2567, pruned_loss=0.05324, over 16379.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2549, pruned_loss=0.04088, over 3320484.88 frames. ], batch size: 146, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:27:55,537 INFO [train.py:904] (0/8) Epoch 21, batch 2050, loss[loss=0.1586, simple_loss=0.248, pruned_loss=0.03463, over 17215.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2544, pruned_loss=0.04061, over 3326557.81 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:27:59,958 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5186, 5.9427, 5.6413, 5.7054, 5.2809, 5.2595, 5.2896, 6.0339], device='cuda:0'), covar=tensor([0.1390, 0.0896, 0.1093, 0.0931, 0.0984, 0.0761, 0.1244, 0.0906], device='cuda:0'), in_proj_covar=tensor([0.0686, 0.0834, 0.0687, 0.0630, 0.0531, 0.0537, 0.0705, 0.0644], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:28:39,590 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 05:28:44,316 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.164e+02 2.478e+02 3.098e+02 5.896e+02, threshold=4.956e+02, percent-clipped=1.0 2023-05-01 05:29:04,145 INFO [train.py:904] (0/8) Epoch 21, batch 2100, loss[loss=0.1519, simple_loss=0.2367, pruned_loss=0.03354, over 16747.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2561, pruned_loss=0.04175, over 3321855.12 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:29:14,663 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205109.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:29:19,836 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-01 05:29:34,124 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6581, 2.4835, 2.4977, 4.6266, 2.4347, 2.8487, 2.5060, 2.6562], device='cuda:0'), covar=tensor([0.1233, 0.3527, 0.3007, 0.0432, 0.4005, 0.2544, 0.3454, 0.3518], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0448, 0.0368, 0.0331, 0.0435, 0.0514, 0.0417, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:29:36,237 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6457, 6.0269, 5.7394, 5.8391, 5.4383, 5.3890, 5.4534, 6.1665], device='cuda:0'), covar=tensor([0.1480, 0.0911, 0.1139, 0.0892, 0.0911, 0.0722, 0.1211, 0.0893], device='cuda:0'), in_proj_covar=tensor([0.0687, 0.0838, 0.0689, 0.0631, 0.0533, 0.0537, 0.0707, 0.0648], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:29:45,619 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.63 vs. limit=5.0 2023-05-01 05:30:14,931 INFO [train.py:904] (0/8) Epoch 21, batch 2150, loss[loss=0.1551, simple_loss=0.2437, pruned_loss=0.0332, over 16797.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2563, pruned_loss=0.04168, over 3321731.87 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:30:39,351 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205170.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:30:49,805 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205177.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:04,863 INFO [optim.py:368] (0/8) 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 05:31:09,482 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 05:31:11,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1736, 2.4211, 2.9366, 3.1276, 2.9535, 3.6947, 2.5825, 3.5694], device='cuda:0'), covar=tensor([0.0247, 0.0460, 0.0299, 0.0303, 0.0317, 0.0179, 0.0469, 0.0174], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0194, 0.0180, 0.0185, 0.0197, 0.0154, 0.0197, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:31:19,699 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205198.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:31:23,199 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 05:31:24,883 INFO [train.py:904] (0/8) Epoch 21, batch 2200, loss[loss=0.2104, simple_loss=0.2875, pruned_loss=0.06665, over 11905.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2575, pruned_loss=0.04209, over 3321812.27 frames. ], batch size: 246, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:31:34,630 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9518, 3.9882, 4.2788, 4.2843, 4.3023, 4.0354, 4.0731, 3.9978], device='cuda:0'), covar=tensor([0.0387, 0.0630, 0.0442, 0.0409, 0.0489, 0.0442, 0.0721, 0.0587], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0453, 0.0442, 0.0410, 0.0486, 0.0462, 0.0551, 0.0371], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 05:31:57,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6253, 2.7976, 2.8118, 4.7626, 3.8704, 4.3482, 1.6900, 3.0349], device='cuda:0'), covar=tensor([0.1657, 0.0872, 0.1188, 0.0249, 0.0290, 0.0393, 0.1840, 0.0882], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0174, 0.0193, 0.0190, 0.0205, 0.0215, 0.0201, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:32:34,220 INFO [train.py:904] (0/8) Epoch 21, batch 2250, loss[loss=0.1437, simple_loss=0.2237, pruned_loss=0.03187, over 16841.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.258, pruned_loss=0.04242, over 3333293.97 frames. ], batch size: 42, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:32:45,048 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:33:23,479 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.631e+02 2.149e+02 2.507e+02 2.983e+02 5.754e+02, threshold=5.014e+02, percent-clipped=1.0 2023-05-01 05:33:44,292 INFO [train.py:904] (0/8) Epoch 21, batch 2300, loss[loss=0.2029, simple_loss=0.2757, pruned_loss=0.065, over 16933.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2586, pruned_loss=0.04297, over 3328095.23 frames. ], batch size: 109, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:34:02,425 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-01 05:34:18,828 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 05:34:40,580 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3381, 5.2613, 5.1579, 4.6990, 4.8420, 5.2601, 5.2016, 4.8286], device='cuda:0'), covar=tensor([0.0600, 0.0501, 0.0304, 0.0337, 0.1053, 0.0408, 0.0299, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0437, 0.0354, 0.0351, 0.0362, 0.0406, 0.0242, 0.0427], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 05:34:50,777 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7380, 4.7762, 5.2117, 5.2075, 5.2414, 4.8405, 4.8283, 4.5883], device='cuda:0'), covar=tensor([0.0365, 0.0614, 0.0402, 0.0408, 0.0558, 0.0423, 0.1031, 0.0555], device='cuda:0'), in_proj_covar=tensor([0.0409, 0.0451, 0.0440, 0.0409, 0.0485, 0.0461, 0.0548, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 05:34:53,191 INFO [train.py:904] (0/8) Epoch 21, batch 2350, loss[loss=0.2045, simple_loss=0.2891, pruned_loss=0.05999, over 16538.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2593, pruned_loss=0.0436, over 3323772.44 frames. ], batch size: 68, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:34:58,766 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5376, 5.9019, 5.6300, 5.7359, 5.2850, 5.3149, 5.2808, 6.0377], device='cuda:0'), covar=tensor([0.1449, 0.1034, 0.1238, 0.0909, 0.1009, 0.0737, 0.1299, 0.0987], device='cuda:0'), in_proj_covar=tensor([0.0680, 0.0836, 0.0685, 0.0630, 0.0531, 0.0535, 0.0702, 0.0643], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:35:12,069 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6014, 2.3819, 1.9503, 2.0117, 2.6546, 2.4045, 2.6612, 2.7759], device='cuda:0'), covar=tensor([0.0198, 0.0384, 0.0508, 0.0514, 0.0264, 0.0374, 0.0199, 0.0286], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0242, 0.0230, 0.0231, 0.0242, 0.0241, 0.0244, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:35:40,999 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8298, 2.7594, 2.4926, 2.5720, 3.0825, 2.8704, 3.4092, 3.3035], device='cuda:0'), covar=tensor([0.0127, 0.0438, 0.0514, 0.0505, 0.0320, 0.0430, 0.0267, 0.0310], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0242, 0.0230, 0.0231, 0.0242, 0.0241, 0.0243, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:35:42,805 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.163e+02 2.457e+02 2.980e+02 4.846e+02, threshold=4.914e+02, percent-clipped=0.0 2023-05-01 05:35:47,807 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 05:35:48,098 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-01 05:36:02,966 INFO [train.py:904] (0/8) Epoch 21, batch 2400, loss[loss=0.1982, simple_loss=0.2771, pruned_loss=0.05969, over 16525.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.26, pruned_loss=0.04326, over 3327276.43 frames. ], batch size: 75, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:37:10,843 INFO [train.py:904] (0/8) Epoch 21, batch 2450, loss[loss=0.1751, simple_loss=0.2537, pruned_loss=0.04823, over 16883.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.26, pruned_loss=0.04253, over 3335582.10 frames. ], batch size: 109, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:37:16,095 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7570, 4.1552, 3.0274, 2.2504, 2.6825, 2.5881, 4.5736, 3.6065], device='cuda:0'), covar=tensor([0.3010, 0.0634, 0.1836, 0.2939, 0.2823, 0.2014, 0.0402, 0.1391], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0271, 0.0305, 0.0310, 0.0297, 0.0258, 0.0296, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 05:37:29,566 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205465.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 05:37:45,904 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205477.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:37:52,176 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 05:38:00,845 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.425e+02 2.346e+02 2.648e+02 3.284e+02 5.369e+02, threshold=5.296e+02, percent-clipped=1.0 2023-05-01 05:38:21,267 INFO [train.py:904] (0/8) Epoch 21, batch 2500, loss[loss=0.1814, simple_loss=0.278, pruned_loss=0.0424, over 17049.00 frames. ], tot_loss[loss=0.1726, simple_loss=0.26, pruned_loss=0.04261, over 3321928.43 frames. ], batch size: 55, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:38:52,867 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:38:57,803 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1794, 3.9304, 4.4592, 2.2703, 4.7657, 4.7325, 3.4838, 3.6875], device='cuda:0'), covar=tensor([0.0694, 0.0259, 0.0237, 0.1141, 0.0072, 0.0168, 0.0380, 0.0376], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0140, 0.0080, 0.0125, 0.0129, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:39:30,351 INFO [train.py:904] (0/8) Epoch 21, batch 2550, loss[loss=0.1705, simple_loss=0.2649, pruned_loss=0.03803, over 17057.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2602, pruned_loss=0.04228, over 3328852.71 frames. ], batch size: 55, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:39:33,578 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=205554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:40:19,335 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.360e+02 2.328e+02 2.679e+02 3.297e+02 1.189e+03, threshold=5.358e+02, percent-clipped=3.0 2023-05-01 05:40:38,687 INFO [train.py:904] (0/8) Epoch 21, batch 2600, loss[loss=0.1598, simple_loss=0.2431, pruned_loss=0.03825, over 16809.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2596, pruned_loss=0.04202, over 3335804.28 frames. ], batch size: 39, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:40:39,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4568, 3.4766, 2.1118, 3.7587, 2.7303, 3.6837, 2.2104, 2.7871], device='cuda:0'), covar=tensor([0.0269, 0.0447, 0.1559, 0.0292, 0.0757, 0.0767, 0.1448, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0180, 0.0197, 0.0166, 0.0179, 0.0221, 0.0205, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 05:41:27,953 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 05:41:49,805 INFO [train.py:904] (0/8) Epoch 21, batch 2650, loss[loss=0.1781, simple_loss=0.2754, pruned_loss=0.04039, over 17041.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2603, pruned_loss=0.04157, over 3330167.26 frames. ], batch size: 55, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:41:54,705 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205655.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:42:40,191 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.500e+02 2.048e+02 2.349e+02 2.910e+02 5.870e+02, threshold=4.699e+02, percent-clipped=1.0 2023-05-01 05:43:00,012 INFO [train.py:904] (0/8) Epoch 21, batch 2700, loss[loss=0.1757, simple_loss=0.2494, pruned_loss=0.05107, over 16703.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.26, pruned_loss=0.04123, over 3334887.82 frames. ], batch size: 124, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:43:18,666 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205716.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:43:59,628 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:44:09,883 INFO [train.py:904] (0/8) Epoch 21, batch 2750, loss[loss=0.1593, simple_loss=0.2534, pruned_loss=0.03265, over 17223.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.26, pruned_loss=0.04111, over 3331803.04 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 4.0 2023-05-01 05:44:29,207 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205765.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:44:32,633 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7965, 6.1012, 5.8705, 6.0387, 5.5209, 5.4649, 5.6270, 6.2880], device='cuda:0'), covar=tensor([0.1284, 0.0878, 0.1042, 0.0843, 0.0916, 0.0738, 0.1213, 0.0863], device='cuda:0'), in_proj_covar=tensor([0.0681, 0.0841, 0.0688, 0.0632, 0.0535, 0.0537, 0.0703, 0.0647], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:45:02,030 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 21, batch 2800, loss[loss=0.1469, simple_loss=0.2403, pruned_loss=0.02668, over 17199.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2593, pruned_loss=0.04068, over 3322291.65 frames. ], batch size: 44, lr: 3.24e-03, grad_scale: 8.0 2023-05-01 05:45:23,861 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205805.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:35,279 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205813.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:45:38,232 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3971, 5.3582, 5.2012, 4.5784, 5.2114, 2.0247, 4.9616, 5.1974], device='cuda:0'), covar=tensor([0.0093, 0.0071, 0.0206, 0.0429, 0.0101, 0.2604, 0.0146, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0157, 0.0200, 0.0181, 0.0178, 0.0211, 0.0190, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:45:53,170 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6264, 2.3767, 2.3279, 4.4525, 2.4502, 2.7374, 2.4608, 2.5513], device='cuda:0'), covar=tensor([0.1195, 0.3741, 0.3147, 0.0498, 0.3949, 0.2774, 0.3449, 0.3726], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0449, 0.0370, 0.0332, 0.0436, 0.0516, 0.0417, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:46:29,085 INFO [train.py:904] (0/8) Epoch 21, batch 2850, loss[loss=0.1493, simple_loss=0.2457, pruned_loss=0.02642, over 17171.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2581, pruned_loss=0.04053, over 3330778.83 frames. ], batch size: 46, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:46:31,699 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=205854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:46:52,882 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-01 05:47:17,822 INFO [optim.py:368] (0/8) 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 05:47:19,299 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=205890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:36,339 INFO [train.py:904] (0/8) Epoch 21, batch 2900, loss[loss=0.1786, simple_loss=0.2478, pruned_loss=0.05466, over 16889.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2572, pruned_loss=0.04045, over 3333461.15 frames. ], batch size: 116, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:47:36,638 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=205902.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:47:50,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6050, 2.4425, 2.4883, 4.5653, 2.4207, 2.7950, 2.5017, 2.6116], device='cuda:0'), covar=tensor([0.1167, 0.3683, 0.2919, 0.0414, 0.3946, 0.2628, 0.3537, 0.3558], device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0449, 0.0369, 0.0331, 0.0436, 0.0516, 0.0418, 0.0526], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:48:44,868 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=205951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:48:45,572 INFO [train.py:904] (0/8) Epoch 21, batch 2950, loss[loss=0.1769, simple_loss=0.2525, pruned_loss=0.05064, over 16729.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2566, pruned_loss=0.04053, over 3334878.67 frames. ], batch size: 89, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:36,952 INFO [optim.py:368] (0/8) 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 05:49:51,237 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-206000.pt 2023-05-01 05:49:58,059 INFO [train.py:904] (0/8) Epoch 21, batch 3000, loss[loss=0.1614, simple_loss=0.2526, pruned_loss=0.03512, over 17228.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2566, pruned_loss=0.04117, over 3331090.37 frames. ], batch size: 44, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:49:58,060 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 05:50:05,556 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2932, 2.2638, 2.3189, 3.9263, 2.2112, 2.5821, 2.3722, 2.3803], device='cuda:0'), covar=tensor([0.1352, 0.3960, 0.3302, 0.0592, 0.4613, 0.2850, 0.3981, 0.3909], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0450, 0.0370, 0.0332, 0.0436, 0.0517, 0.0419, 0.0527], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:50:06,476 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 05:50:18,438 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206011.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:51:14,533 INFO [train.py:904] (0/8) Epoch 21, batch 3050, loss[loss=0.1485, simple_loss=0.2328, pruned_loss=0.0321, over 16814.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2571, pruned_loss=0.04199, over 3316448.73 frames. ], batch size: 39, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:51:21,153 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206056.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:51:30,488 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-01 05:51:38,414 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5433, 2.4108, 1.9076, 2.0047, 2.6331, 2.3606, 2.6036, 2.7382], device='cuda:0'), covar=tensor([0.0234, 0.0321, 0.0501, 0.0439, 0.0225, 0.0335, 0.0209, 0.0246], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0241, 0.0229, 0.0230, 0.0240, 0.0240, 0.0244, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:52:05,520 INFO [optim.py:368] (0/8) 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 05:52:21,249 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206100.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:52:24,080 INFO [train.py:904] (0/8) Epoch 21, batch 3100, loss[loss=0.1827, simple_loss=0.2707, pruned_loss=0.04736, over 16613.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2569, pruned_loss=0.04238, over 3320744.75 frames. ], batch size: 62, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:52:43,719 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206117.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:53:31,003 INFO [train.py:904] (0/8) Epoch 21, batch 3150, loss[loss=0.1984, simple_loss=0.2806, pruned_loss=0.05812, over 12059.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.255, pruned_loss=0.0416, over 3318947.78 frames. ], batch size: 246, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:53:53,020 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5303, 3.7328, 3.9823, 2.1813, 3.2058, 2.5516, 3.9066, 3.9895], device='cuda:0'), covar=tensor([0.0303, 0.0949, 0.0521, 0.2157, 0.0864, 0.1077, 0.0689, 0.1103], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0154, 0.0146, 0.0131, 0.0146, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-01 05:54:22,973 INFO [optim.py:368] (0/8) 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 05:54:34,340 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206197.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:54:41,550 INFO [train.py:904] (0/8) Epoch 21, batch 3200, loss[loss=0.1547, simple_loss=0.2527, pruned_loss=0.0283, over 17113.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2549, pruned_loss=0.04096, over 3314430.35 frames. ], batch size: 48, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:42,480 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:55:51,283 INFO [train.py:904] (0/8) Epoch 21, batch 3250, loss[loss=0.1335, simple_loss=0.2152, pruned_loss=0.02593, over 17014.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2552, pruned_loss=0.04113, over 3318838.17 frames. ], batch size: 41, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:55:59,442 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:56:40,075 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0792, 5.0300, 4.8523, 4.2202, 4.9112, 1.8646, 4.6751, 4.7553], device='cuda:0'), covar=tensor([0.0100, 0.0081, 0.0226, 0.0448, 0.0114, 0.2771, 0.0142, 0.0221], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0158, 0.0201, 0.0181, 0.0179, 0.0210, 0.0190, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:56:42,634 INFO [optim.py:368] (0/8) 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 05:56:59,771 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4596, 4.7902, 4.6091, 4.6079, 4.3579, 4.3648, 4.2945, 4.8892], device='cuda:0'), covar=tensor([0.1334, 0.1046, 0.1087, 0.0932, 0.0828, 0.1377, 0.1216, 0.0883], device='cuda:0'), in_proj_covar=tensor([0.0681, 0.0843, 0.0693, 0.0633, 0.0535, 0.0538, 0.0704, 0.0650], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 05:57:00,590 INFO [train.py:904] (0/8) Epoch 21, batch 3300, loss[loss=0.1865, simple_loss=0.2677, pruned_loss=0.05262, over 16188.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2569, pruned_loss=0.04205, over 3315821.94 frames. ], batch size: 165, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:57:13,826 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206311.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:58:09,896 INFO [train.py:904] (0/8) Epoch 21, batch 3350, loss[loss=0.1632, simple_loss=0.2514, pruned_loss=0.03746, over 16830.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2572, pruned_loss=0.04164, over 3323160.35 frames. ], batch size: 83, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:58:20,158 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:00,092 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.208e+02 2.654e+02 3.136e+02 4.728e+02, threshold=5.309e+02, percent-clipped=0.0 2023-05-01 05:59:16,468 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206400.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 05:59:16,705 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8052, 4.3121, 3.0957, 2.3711, 2.6993, 2.7024, 4.6843, 3.6050], device='cuda:0'), covar=tensor([0.2850, 0.0572, 0.1783, 0.2905, 0.2812, 0.1911, 0.0343, 0.1394], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0272, 0.0306, 0.0311, 0.0299, 0.0260, 0.0296, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 05:59:18,367 INFO [train.py:904] (0/8) Epoch 21, batch 3400, loss[loss=0.1758, simple_loss=0.2681, pruned_loss=0.04174, over 17132.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2572, pruned_loss=0.0414, over 3321901.82 frames. ], batch size: 48, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 05:59:32,310 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:00:21,657 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:00:26,897 INFO [train.py:904] (0/8) Epoch 21, batch 3450, loss[loss=0.1526, simple_loss=0.2447, pruned_loss=0.03026, over 17182.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2562, pruned_loss=0.0412, over 3310780.37 frames. ], batch size: 46, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:00:46,807 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 06:01:17,175 INFO [optim.py:368] (0/8) 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 06:01:36,862 INFO [train.py:904] (0/8) Epoch 21, batch 3500, loss[loss=0.2, simple_loss=0.2765, pruned_loss=0.06173, over 16680.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2552, pruned_loss=0.04057, over 3323102.25 frames. ], batch size: 134, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:01:58,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2560, 4.1232, 4.3247, 4.4419, 4.5655, 4.1645, 4.3596, 4.5607], device='cuda:0'), covar=tensor([0.1603, 0.1148, 0.1437, 0.0806, 0.0666, 0.1231, 0.2459, 0.0810], device='cuda:0'), in_proj_covar=tensor([0.0666, 0.0817, 0.0960, 0.0846, 0.0626, 0.0657, 0.0675, 0.0782], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:02:37,255 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:02:44,996 INFO [train.py:904] (0/8) Epoch 21, batch 3550, loss[loss=0.1759, simple_loss=0.2566, pruned_loss=0.04762, over 11906.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2548, pruned_loss=0.04074, over 3310646.65 frames. ], batch size: 246, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:02:47,187 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=206553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:18,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.34 vs. limit=5.0 2023-05-01 06:03:37,043 INFO [optim.py:368] (0/8) 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 06:03:43,639 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:03:55,939 INFO [train.py:904] (0/8) Epoch 21, batch 3600, loss[loss=0.1439, simple_loss=0.2358, pruned_loss=0.02603, over 17221.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2545, pruned_loss=0.04057, over 3316374.41 frames. ], batch size: 45, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:04:02,520 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7398, 2.6982, 2.4426, 2.6143, 2.9982, 2.7984, 3.3967, 3.2458], device='cuda:0'), covar=tensor([0.0138, 0.0436, 0.0475, 0.0435, 0.0294, 0.0389, 0.0216, 0.0285], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0240, 0.0229, 0.0229, 0.0241, 0.0240, 0.0243, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:05:06,986 INFO [train.py:904] (0/8) Epoch 21, batch 3650, loss[loss=0.1534, simple_loss=0.2287, pruned_loss=0.03901, over 16409.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2539, pruned_loss=0.04156, over 3300906.54 frames. ], batch size: 75, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:05:59,789 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.607e+02 2.224e+02 2.566e+02 3.274e+02 6.534e+02, threshold=5.132e+02, percent-clipped=5.0 2023-05-01 06:06:18,969 INFO [train.py:904] (0/8) Epoch 21, batch 3700, loss[loss=0.1712, simple_loss=0.2577, pruned_loss=0.04228, over 15438.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.253, pruned_loss=0.0431, over 3257378.73 frames. ], batch size: 190, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:06:34,657 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:07:29,066 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 06:07:32,578 INFO [train.py:904] (0/8) Epoch 21, batch 3750, loss[loss=0.1652, simple_loss=0.2489, pruned_loss=0.04077, over 15681.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2532, pruned_loss=0.0445, over 3258667.83 frames. ], batch size: 191, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:07:45,107 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206760.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:08:26,979 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.267e+02 2.726e+02 3.357e+02 9.074e+02, threshold=5.452e+02, percent-clipped=3.0 2023-05-01 06:08:45,468 INFO [train.py:904] (0/8) Epoch 21, batch 3800, loss[loss=0.17, simple_loss=0.2602, pruned_loss=0.03989, over 16252.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2546, pruned_loss=0.04556, over 3268710.31 frames. ], batch size: 35, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:09:32,867 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=206834.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:09:58,316 INFO [train.py:904] (0/8) Epoch 21, batch 3850, loss[loss=0.1974, simple_loss=0.278, pruned_loss=0.05834, over 12266.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2546, pruned_loss=0.0461, over 3254171.23 frames. ], batch size: 245, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:10:00,685 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=206853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:10:52,988 INFO [optim.py:368] (0/8) 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 06:11:00,912 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=206895.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 06:11:09,429 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=206901.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:11:10,722 INFO [train.py:904] (0/8) Epoch 21, batch 3900, loss[loss=0.1755, simple_loss=0.2569, pruned_loss=0.04707, over 16886.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.254, pruned_loss=0.04615, over 3261194.54 frames. ], batch size: 96, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:12:24,779 INFO [train.py:904] (0/8) Epoch 21, batch 3950, loss[loss=0.1776, simple_loss=0.2601, pruned_loss=0.0476, over 16473.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2535, pruned_loss=0.04678, over 3264554.82 frames. ], batch size: 68, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:12:35,812 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9497, 2.0968, 2.5422, 2.8459, 2.8413, 2.8522, 2.0775, 3.0805], device='cuda:0'), covar=tensor([0.0174, 0.0475, 0.0319, 0.0267, 0.0281, 0.0266, 0.0513, 0.0147], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0194, 0.0180, 0.0186, 0.0198, 0.0156, 0.0198, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:12:59,697 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0978, 2.1095, 2.3299, 3.7585, 2.1626, 2.3864, 2.2090, 2.2646], device='cuda:0'), covar=tensor([0.1519, 0.3617, 0.2860, 0.0668, 0.3877, 0.2698, 0.3798, 0.3004], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0452, 0.0371, 0.0334, 0.0439, 0.0521, 0.0420, 0.0529], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:13:16,334 INFO [optim.py:368] (0/8) 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 06:13:22,303 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9937, 3.1200, 3.2206, 2.1111, 2.8558, 2.3269, 3.4918, 3.5478], device='cuda:0'), covar=tensor([0.0221, 0.0797, 0.0595, 0.1869, 0.0807, 0.0978, 0.0520, 0.0705], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0165, 0.0168, 0.0152, 0.0145, 0.0130, 0.0144, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 06:13:22,455 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 06:13:35,030 INFO [train.py:904] (0/8) Epoch 21, batch 4000, loss[loss=0.1833, simple_loss=0.2669, pruned_loss=0.04983, over 16689.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2534, pruned_loss=0.04732, over 3264413.59 frames. ], batch size: 83, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:13:43,144 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 06:13:57,582 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1583, 4.4845, 4.7479, 4.7040, 4.7220, 4.4037, 4.1247, 4.2643], device='cuda:0'), covar=tensor([0.0621, 0.0745, 0.0555, 0.0693, 0.0760, 0.0658, 0.1443, 0.0660], device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0464, 0.0448, 0.0417, 0.0496, 0.0472, 0.0561, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 06:14:40,608 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.99 vs. limit=2.0 2023-05-01 06:14:45,537 INFO [train.py:904] (0/8) Epoch 21, batch 4050, loss[loss=0.1662, simple_loss=0.2551, pruned_loss=0.03865, over 17192.00 frames. ], tot_loss[loss=0.1739, simple_loss=0.2543, pruned_loss=0.04672, over 3268230.17 frames. ], batch size: 46, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:15:37,548 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.164e+02 1.811e+02 2.022e+02 2.362e+02 4.277e+02, threshold=4.045e+02, percent-clipped=0.0 2023-05-01 06:15:56,006 INFO [train.py:904] (0/8) Epoch 21, batch 4100, loss[loss=0.196, simple_loss=0.2806, pruned_loss=0.05573, over 16700.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2555, pruned_loss=0.04604, over 3261940.74 frames. ], batch size: 124, lr: 3.23e-03, grad_scale: 8.0 2023-05-01 06:16:27,169 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8695, 5.0466, 5.3200, 5.3253, 5.3482, 4.9976, 4.9305, 4.7010], device='cuda:0'), covar=tensor([0.0314, 0.0385, 0.0328, 0.0310, 0.0492, 0.0308, 0.0934, 0.0455], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0460, 0.0445, 0.0413, 0.0491, 0.0467, 0.0556, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 06:17:10,033 INFO [train.py:904] (0/8) Epoch 21, batch 4150, loss[loss=0.1879, simple_loss=0.2899, pruned_loss=0.043, over 16432.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2627, pruned_loss=0.0489, over 3219121.28 frames. ], batch size: 146, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:17:38,211 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1038, 5.0142, 4.7870, 3.9366, 5.0121, 1.6607, 4.7162, 4.5764], device='cuda:0'), covar=tensor([0.0107, 0.0100, 0.0220, 0.0506, 0.0102, 0.3244, 0.0135, 0.0275], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0158, 0.0200, 0.0182, 0.0180, 0.0211, 0.0191, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:17:45,664 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=207176.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:18:04,523 INFO [optim.py:368] (0/8) 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 06:18:06,788 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207190.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 06:18:24,677 INFO [train.py:904] (0/8) Epoch 21, batch 4200, loss[loss=0.2342, simple_loss=0.3138, pruned_loss=0.07723, over 11488.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2696, pruned_loss=0.05042, over 3198450.68 frames. ], batch size: 248, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:19:17,921 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=207237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:19:26,212 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3158, 3.4936, 3.6238, 3.5998, 3.6191, 3.4562, 3.4570, 3.5112], device='cuda:0'), covar=tensor([0.0415, 0.0672, 0.0527, 0.0493, 0.0570, 0.0535, 0.0971, 0.0567], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0458, 0.0443, 0.0412, 0.0489, 0.0466, 0.0554, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 06:19:40,051 INFO [train.py:904] (0/8) Epoch 21, batch 4250, loss[loss=0.1848, simple_loss=0.2846, pruned_loss=0.04254, over 17179.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2725, pruned_loss=0.04994, over 3181714.76 frames. ], batch size: 44, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:19:42,016 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 06:19:50,524 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2306, 2.2056, 2.3205, 3.9569, 2.0985, 2.3812, 2.2946, 2.3263], device='cuda:0'), covar=tensor([0.1536, 0.3931, 0.2876, 0.0644, 0.4728, 0.3113, 0.3594, 0.3822], device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0450, 0.0368, 0.0331, 0.0435, 0.0518, 0.0417, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:20:10,951 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 06:20:35,828 INFO [optim.py:368] (0/8) 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 06:20:55,810 INFO [train.py:904] (0/8) Epoch 21, batch 4300, loss[loss=0.1908, simple_loss=0.2831, pruned_loss=0.04922, over 16791.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2738, pruned_loss=0.04911, over 3183756.35 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:22:07,509 INFO [train.py:904] (0/8) Epoch 21, batch 4350, loss[loss=0.1978, simple_loss=0.2832, pruned_loss=0.05623, over 16694.00 frames. ], tot_loss[loss=0.1888, simple_loss=0.2771, pruned_loss=0.05025, over 3169931.16 frames. ], batch size: 62, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:22:57,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6588, 3.8371, 2.3774, 4.5179, 2.9995, 4.4057, 2.5781, 3.0756], device='cuda:0'), covar=tensor([0.0295, 0.0347, 0.1616, 0.0126, 0.0754, 0.0451, 0.1409, 0.0717], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0177, 0.0193, 0.0163, 0.0177, 0.0216, 0.0200, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 06:23:02,766 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.219e+02 2.598e+02 3.036e+02 7.307e+02, threshold=5.195e+02, percent-clipped=1.0 2023-05-01 06:23:22,045 INFO [train.py:904] (0/8) Epoch 21, batch 4400, loss[loss=0.2071, simple_loss=0.2945, pruned_loss=0.05987, over 16907.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2789, pruned_loss=0.05085, over 3196000.60 frames. ], batch size: 116, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:24:37,196 INFO [train.py:904] (0/8) Epoch 21, batch 4450, loss[loss=0.18, simple_loss=0.2806, pruned_loss=0.03974, over 16847.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2821, pruned_loss=0.05222, over 3192365.57 frames. ], batch size: 96, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:24:58,580 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 06:25:31,513 INFO [optim.py:368] (0/8) 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 06:25:33,771 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207490.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 06:25:50,492 INFO [train.py:904] (0/8) Epoch 21, batch 4500, loss[loss=0.1942, simple_loss=0.2837, pruned_loss=0.05235, over 16809.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2825, pruned_loss=0.05276, over 3195800.00 frames. ], batch size: 83, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:26:35,647 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=207532.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:26:43,151 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=207538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:26:59,875 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5645, 2.4626, 2.4313, 3.5629, 2.6256, 3.7873, 1.4644, 2.7774], device='cuda:0'), covar=tensor([0.1388, 0.0895, 0.1200, 0.0162, 0.0236, 0.0347, 0.1792, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0175, 0.0193, 0.0190, 0.0207, 0.0215, 0.0200, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 06:27:03,827 INFO [train.py:904] (0/8) Epoch 21, batch 4550, loss[loss=0.195, simple_loss=0.2808, pruned_loss=0.05461, over 16581.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2839, pruned_loss=0.05381, over 3215246.28 frames. ], batch size: 62, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:27:57,548 INFO [optim.py:368] (0/8) 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 06:28:11,204 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 06:28:16,272 INFO [train.py:904] (0/8) Epoch 21, batch 4600, loss[loss=0.1664, simple_loss=0.2619, pruned_loss=0.03544, over 16840.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2848, pruned_loss=0.05399, over 3213337.58 frames. ], batch size: 102, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:29:28,347 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4675, 4.8197, 4.4045, 4.7258, 4.3754, 4.2214, 4.3343, 4.8592], device='cuda:0'), covar=tensor([0.2313, 0.1352, 0.2159, 0.1270, 0.1575, 0.2356, 0.2485, 0.1528], device='cuda:0'), in_proj_covar=tensor([0.0662, 0.0809, 0.0672, 0.0615, 0.0514, 0.0522, 0.0681, 0.0633], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:29:29,139 INFO [train.py:904] (0/8) Epoch 21, batch 4650, loss[loss=0.175, simple_loss=0.2627, pruned_loss=0.04361, over 16380.00 frames. ], tot_loss[loss=0.1963, simple_loss=0.2839, pruned_loss=0.05435, over 3208069.04 frames. ], batch size: 146, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:29:52,714 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4689, 5.4272, 5.0868, 4.5683, 5.4085, 1.8767, 5.1170, 4.7617], device='cuda:0'), covar=tensor([0.0039, 0.0035, 0.0136, 0.0228, 0.0040, 0.2744, 0.0068, 0.0195], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0154, 0.0196, 0.0177, 0.0175, 0.0207, 0.0186, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:30:23,481 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 1.925e+02 2.221e+02 2.572e+02 4.933e+02, threshold=4.441e+02, percent-clipped=1.0 2023-05-01 06:30:42,427 INFO [train.py:904] (0/8) Epoch 21, batch 4700, loss[loss=0.1573, simple_loss=0.2462, pruned_loss=0.03426, over 16613.00 frames. ], tot_loss[loss=0.1936, simple_loss=0.2809, pruned_loss=0.05314, over 3203102.63 frames. ], batch size: 68, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:30:59,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4242, 3.3812, 2.6533, 2.2066, 2.2584, 2.2511, 3.6334, 3.0390], device='cuda:0'), covar=tensor([0.3063, 0.0750, 0.1940, 0.2706, 0.2644, 0.2220, 0.0499, 0.1338], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0269, 0.0303, 0.0310, 0.0297, 0.0257, 0.0295, 0.0337], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 06:31:12,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5099, 3.6155, 3.3295, 2.9752, 3.1860, 3.3995, 3.3112, 3.2600], device='cuda:0'), covar=tensor([0.0641, 0.0624, 0.0281, 0.0258, 0.0548, 0.0495, 0.1385, 0.0434], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0417, 0.0338, 0.0333, 0.0348, 0.0388, 0.0231, 0.0406], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:31:56,756 INFO [train.py:904] (0/8) Epoch 21, batch 4750, loss[loss=0.1557, simple_loss=0.2451, pruned_loss=0.03314, over 16895.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2765, pruned_loss=0.051, over 3200183.88 frames. ], batch size: 116, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:32:35,981 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5982, 4.4349, 4.5987, 4.7789, 4.9751, 4.4785, 4.9545, 4.9736], device='cuda:0'), covar=tensor([0.1836, 0.1224, 0.1823, 0.0742, 0.0561, 0.1087, 0.0568, 0.0605], device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0771, 0.0902, 0.0789, 0.0591, 0.0620, 0.0635, 0.0739], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:32:50,132 INFO [optim.py:368] (0/8) 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 06:33:10,676 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2023-05-01 06:33:11,161 INFO [train.py:904] (0/8) Epoch 21, batch 4800, loss[loss=0.1808, simple_loss=0.2653, pruned_loss=0.04812, over 16611.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.273, pruned_loss=0.04895, over 3196268.93 frames. ], batch size: 62, lr: 3.22e-03, grad_scale: 16.0 2023-05-01 06:33:50,339 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1550, 2.1748, 2.3372, 3.9112, 2.1031, 2.4840, 2.2771, 2.3574], device='cuda:0'), covar=tensor([0.1448, 0.3798, 0.2743, 0.0563, 0.4239, 0.2747, 0.3676, 0.3386], device='cuda:0'), in_proj_covar=tensor([0.0400, 0.0446, 0.0365, 0.0328, 0.0435, 0.0515, 0.0415, 0.0521], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:33:55,569 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=207832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:34:24,558 INFO [train.py:904] (0/8) Epoch 21, batch 4850, loss[loss=0.232, simple_loss=0.3134, pruned_loss=0.07532, over 12414.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2738, pruned_loss=0.04831, over 3177462.00 frames. ], batch size: 247, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:35:08,163 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=207880.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:35:12,886 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-05-01 06:35:22,070 INFO [optim.py:368] (0/8) 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 06:35:27,235 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5580, 4.4094, 4.6012, 4.7559, 4.9526, 4.5086, 4.9276, 4.9702], device='cuda:0'), covar=tensor([0.1742, 0.1228, 0.1705, 0.0827, 0.0528, 0.0878, 0.0633, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0764, 0.0894, 0.0783, 0.0585, 0.0614, 0.0632, 0.0733], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:35:40,336 INFO [train.py:904] (0/8) Epoch 21, batch 4900, loss[loss=0.1741, simple_loss=0.2591, pruned_loss=0.04454, over 16164.00 frames. ], tot_loss[loss=0.1833, simple_loss=0.2727, pruned_loss=0.04694, over 3165180.00 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:35:49,336 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2140, 4.2868, 4.5812, 4.5663, 4.5594, 4.2790, 4.2785, 4.2033], device='cuda:0'), covar=tensor([0.0327, 0.0539, 0.0325, 0.0373, 0.0463, 0.0377, 0.0887, 0.0469], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0446, 0.0432, 0.0401, 0.0476, 0.0454, 0.0542, 0.0364], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 06:36:52,786 INFO [train.py:904] (0/8) Epoch 21, batch 4950, loss[loss=0.1801, simple_loss=0.2686, pruned_loss=0.04575, over 17143.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2725, pruned_loss=0.0464, over 3182261.89 frames. ], batch size: 48, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:36:55,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2386, 4.3200, 4.5767, 4.5488, 4.5594, 4.2752, 4.2497, 4.1902], device='cuda:0'), covar=tensor([0.0311, 0.0471, 0.0346, 0.0394, 0.0429, 0.0367, 0.0869, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0445, 0.0432, 0.0400, 0.0475, 0.0452, 0.0539, 0.0363], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 06:37:47,807 INFO [optim.py:368] (0/8) 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 06:38:02,835 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-208000.pt 2023-05-01 06:38:08,412 INFO [train.py:904] (0/8) Epoch 21, batch 5000, loss[loss=0.182, simple_loss=0.2758, pruned_loss=0.04407, over 15315.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2737, pruned_loss=0.04592, over 3191846.62 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:38:52,098 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208032.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:39:21,549 INFO [train.py:904] (0/8) Epoch 21, batch 5050, loss[loss=0.1818, simple_loss=0.2769, pruned_loss=0.04336, over 16902.00 frames. ], tot_loss[loss=0.1831, simple_loss=0.2747, pruned_loss=0.04572, over 3205494.34 frames. ], batch size: 90, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:39:21,985 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7363, 4.8309, 4.6653, 2.9731, 3.9516, 4.6617, 3.9302, 2.6259], device='cuda:0'), covar=tensor([0.0540, 0.0033, 0.0031, 0.0391, 0.0097, 0.0071, 0.0111, 0.0423], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0082, 0.0082, 0.0133, 0.0097, 0.0108, 0.0093, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 06:39:36,354 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9603, 5.0219, 4.8504, 4.5245, 4.4972, 4.8934, 4.8335, 4.6313], device='cuda:0'), covar=tensor([0.0652, 0.0455, 0.0302, 0.0267, 0.1020, 0.0523, 0.0283, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0416, 0.0338, 0.0331, 0.0346, 0.0388, 0.0230, 0.0404], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:40:05,920 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6265, 4.7176, 4.5120, 4.1967, 4.1703, 4.5841, 4.4191, 4.3152], device='cuda:0'), covar=tensor([0.0635, 0.0438, 0.0322, 0.0281, 0.0997, 0.0492, 0.0479, 0.0633], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0414, 0.0336, 0.0330, 0.0345, 0.0386, 0.0230, 0.0403], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:40:18,554 INFO [optim.py:368] (0/8) 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 06:40:23,282 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208093.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 06:40:35,284 INFO [train.py:904] (0/8) Epoch 21, batch 5100, loss[loss=0.1662, simple_loss=0.2609, pruned_loss=0.03573, over 16639.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2734, pruned_loss=0.04526, over 3214312.47 frames. ], batch size: 76, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:41:48,629 INFO [train.py:904] (0/8) Epoch 21, batch 5150, loss[loss=0.1913, simple_loss=0.2927, pruned_loss=0.04493, over 16760.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.2733, pruned_loss=0.04489, over 3193159.70 frames. ], batch size: 124, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:42:00,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4099, 3.4104, 2.6356, 2.0816, 2.2853, 2.3335, 3.5373, 3.1252], device='cuda:0'), covar=tensor([0.2987, 0.0644, 0.1844, 0.2999, 0.2710, 0.2057, 0.0565, 0.1326], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0270, 0.0305, 0.0311, 0.0297, 0.0257, 0.0297, 0.0336], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 06:42:43,677 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 21, batch 5200, loss[loss=0.1949, simple_loss=0.2819, pruned_loss=0.054, over 15343.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.272, pruned_loss=0.04462, over 3189159.60 frames. ], batch size: 190, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:43:59,432 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0426, 2.7865, 2.8281, 2.1567, 2.6299, 2.1875, 2.7901, 2.9673], device='cuda:0'), covar=tensor([0.0289, 0.0707, 0.0565, 0.1662, 0.0816, 0.0888, 0.0568, 0.0593], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0163, 0.0167, 0.0152, 0.0145, 0.0130, 0.0143, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 06:44:11,874 INFO [train.py:904] (0/8) Epoch 21, batch 5250, loss[loss=0.1542, simple_loss=0.2544, pruned_loss=0.02695, over 16867.00 frames. ], tot_loss[loss=0.1783, simple_loss=0.2688, pruned_loss=0.04391, over 3213755.90 frames. ], batch size: 96, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:44:16,200 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 06:44:22,786 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7504, 4.1451, 3.0123, 2.4205, 2.7780, 2.7081, 4.5039, 3.5566], device='cuda:0'), covar=tensor([0.2802, 0.0595, 0.1827, 0.2696, 0.2656, 0.1818, 0.0437, 0.1272], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0269, 0.0304, 0.0309, 0.0295, 0.0256, 0.0296, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 06:44:31,816 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208265.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:44:58,064 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8432, 2.4608, 1.9571, 2.1350, 2.7159, 2.3998, 2.5652, 2.8597], device='cuda:0'), covar=tensor([0.0179, 0.0379, 0.0540, 0.0476, 0.0239, 0.0380, 0.0207, 0.0271], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0231, 0.0223, 0.0224, 0.0233, 0.0231, 0.0232, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:45:07,260 INFO [optim.py:368] (0/8) 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 06:45:25,349 INFO [train.py:904] (0/8) Epoch 21, batch 5300, loss[loss=0.1589, simple_loss=0.2406, pruned_loss=0.03857, over 16415.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2648, pruned_loss=0.04253, over 3228647.62 frames. ], batch size: 35, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:46:00,500 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208326.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:46:37,970 INFO [train.py:904] (0/8) Epoch 21, batch 5350, loss[loss=0.1855, simple_loss=0.2713, pruned_loss=0.04988, over 12455.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2632, pruned_loss=0.04173, over 3226282.88 frames. ], batch size: 246, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:47:32,349 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208388.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 06:47:34,964 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.374e+02 1.963e+02 2.291e+02 2.599e+02 4.161e+02, threshold=4.581e+02, percent-clipped=0.0 2023-05-01 06:47:53,658 INFO [train.py:904] (0/8) Epoch 21, batch 5400, loss[loss=0.173, simple_loss=0.258, pruned_loss=0.04396, over 17189.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2663, pruned_loss=0.04284, over 3215389.53 frames. ], batch size: 46, lr: 3.22e-03, grad_scale: 8.0 2023-05-01 06:48:25,258 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-05-01 06:48:56,961 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5382, 3.3522, 3.8006, 1.9088, 4.0037, 3.9965, 2.9908, 2.9259], device='cuda:0'), covar=tensor([0.0767, 0.0294, 0.0203, 0.1222, 0.0058, 0.0132, 0.0436, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0109, 0.0097, 0.0139, 0.0081, 0.0124, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 06:49:10,975 INFO [train.py:904] (0/8) Epoch 21, batch 5450, loss[loss=0.1764, simple_loss=0.2714, pruned_loss=0.04073, over 16848.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2692, pruned_loss=0.04418, over 3204253.03 frames. ], batch size: 102, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:49:48,947 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4811, 1.7042, 2.1065, 2.4484, 2.4470, 2.7239, 1.7949, 2.6808], device='cuda:0'), covar=tensor([0.0240, 0.0490, 0.0306, 0.0327, 0.0304, 0.0190, 0.0527, 0.0142], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0192, 0.0178, 0.0185, 0.0196, 0.0152, 0.0197, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:50:09,144 INFO [optim.py:368] (0/8) 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 06:50:28,306 INFO [train.py:904] (0/8) Epoch 21, batch 5500, loss[loss=0.1918, simple_loss=0.2862, pruned_loss=0.04866, over 16832.00 frames. ], tot_loss[loss=0.1863, simple_loss=0.2763, pruned_loss=0.04814, over 3209215.60 frames. ], batch size: 96, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:50:47,763 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4228, 3.4467, 2.5995, 2.1965, 2.3154, 2.2338, 3.6346, 3.2080], device='cuda:0'), covar=tensor([0.2784, 0.0626, 0.1841, 0.2478, 0.2583, 0.2075, 0.0493, 0.1169], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0267, 0.0301, 0.0307, 0.0293, 0.0254, 0.0294, 0.0332], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 06:50:57,303 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5685, 2.6682, 2.5780, 4.3045, 3.2090, 4.0774, 1.5053, 2.8508], device='cuda:0'), covar=tensor([0.1360, 0.0791, 0.1197, 0.0185, 0.0295, 0.0379, 0.1704, 0.0862], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0175, 0.0194, 0.0190, 0.0208, 0.0214, 0.0201, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 06:51:47,646 INFO [train.py:904] (0/8) Epoch 21, batch 5550, loss[loss=0.263, simple_loss=0.3289, pruned_loss=0.09856, over 11178.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2832, pruned_loss=0.05257, over 3196980.31 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:52:27,225 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=208576.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:52:27,772 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.58 vs. limit=2.0 2023-05-01 06:52:49,287 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.296e+02 3.115e+02 3.921e+02 4.748e+02 1.231e+03, threshold=7.841e+02, percent-clipped=12.0 2023-05-01 06:52:57,250 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0058, 5.2913, 5.0455, 5.0631, 4.7841, 4.6899, 4.7269, 5.3935], device='cuda:0'), covar=tensor([0.1226, 0.0855, 0.1025, 0.0979, 0.0798, 0.0998, 0.1187, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0653, 0.0794, 0.0662, 0.0602, 0.0505, 0.0511, 0.0666, 0.0621], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:53:07,713 INFO [train.py:904] (0/8) Epoch 21, batch 5600, loss[loss=0.2125, simple_loss=0.2979, pruned_loss=0.06356, over 16671.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2877, pruned_loss=0.05697, over 3146324.15 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:53:12,701 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5048, 4.0991, 4.0358, 2.6908, 3.6885, 4.1381, 3.6925, 2.2904], device='cuda:0'), covar=tensor([0.0496, 0.0050, 0.0050, 0.0378, 0.0093, 0.0098, 0.0089, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0082, 0.0082, 0.0133, 0.0097, 0.0108, 0.0093, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 06:53:39,442 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:53:41,282 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-01 06:54:06,498 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=208637.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:54:29,860 INFO [train.py:904] (0/8) Epoch 21, batch 5650, loss[loss=0.2186, simple_loss=0.3093, pruned_loss=0.06398, over 16680.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.293, pruned_loss=0.06114, over 3120462.99 frames. ], batch size: 76, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:54:33,290 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-01 06:55:28,202 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:55:30,665 INFO [optim.py:368] (0/8) 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 06:55:50,855 INFO [train.py:904] (0/8) Epoch 21, batch 5700, loss[loss=0.2296, simple_loss=0.3191, pruned_loss=0.07005, over 16240.00 frames. ], tot_loss[loss=0.2095, simple_loss=0.2943, pruned_loss=0.06239, over 3110389.67 frames. ], batch size: 165, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:56:45,715 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 06:57:10,984 INFO [train.py:904] (0/8) Epoch 21, batch 5750, loss[loss=0.223, simple_loss=0.3079, pruned_loss=0.06906, over 16264.00 frames. ], tot_loss[loss=0.2121, simple_loss=0.2967, pruned_loss=0.06378, over 3081228.05 frames. ], batch size: 165, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:57:30,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 06:57:48,536 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7721, 5.0678, 4.8131, 4.8522, 4.5841, 4.5670, 4.5203, 5.1445], device='cuda:0'), covar=tensor([0.1118, 0.0774, 0.1014, 0.0929, 0.0798, 0.0989, 0.1203, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0787, 0.0659, 0.0598, 0.0501, 0.0507, 0.0661, 0.0615], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 06:58:13,920 INFO [optim.py:368] (0/8) 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 06:58:33,866 INFO [train.py:904] (0/8) Epoch 21, batch 5800, loss[loss=0.2413, simple_loss=0.306, pruned_loss=0.08829, over 12057.00 frames. ], tot_loss[loss=0.211, simple_loss=0.2966, pruned_loss=0.06271, over 3085415.51 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:59:52,596 INFO [train.py:904] (0/8) Epoch 21, batch 5850, loss[loss=0.2046, simple_loss=0.2913, pruned_loss=0.05892, over 16681.00 frames. ], tot_loss[loss=0.2087, simple_loss=0.2946, pruned_loss=0.06138, over 3078943.45 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 06:59:57,324 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2909, 3.5588, 3.6620, 2.1469, 3.1726, 2.5030, 3.6520, 3.8908], device='cuda:0'), covar=tensor([0.0260, 0.0785, 0.0580, 0.2030, 0.0796, 0.0951, 0.0630, 0.0959], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0163, 0.0167, 0.0152, 0.0145, 0.0130, 0.0143, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 07:00:34,440 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9640, 5.3060, 5.0388, 5.0483, 4.7707, 4.7303, 4.6603, 5.3943], device='cuda:0'), covar=tensor([0.1257, 0.0837, 0.1055, 0.0896, 0.0801, 0.0962, 0.1302, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0650, 0.0789, 0.0660, 0.0600, 0.0502, 0.0509, 0.0663, 0.0616], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:00:53,530 INFO [optim.py:368] (0/8) 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 07:01:04,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0919, 2.2115, 2.2839, 3.7010, 2.1212, 2.5431, 2.3048, 2.3256], device='cuda:0'), covar=tensor([0.1325, 0.3239, 0.2691, 0.0531, 0.3892, 0.2337, 0.3298, 0.3025], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0442, 0.0362, 0.0324, 0.0432, 0.0508, 0.0412, 0.0516], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:01:12,711 INFO [train.py:904] (0/8) Epoch 21, batch 5900, loss[loss=0.2058, simple_loss=0.2941, pruned_loss=0.05878, over 16680.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2942, pruned_loss=0.06099, over 3100043.58 frames. ], batch size: 57, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:01:30,594 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 07:01:48,520 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=208921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:01:54,017 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6210, 4.4637, 4.6467, 4.8373, 4.9768, 4.4900, 4.9778, 4.9959], device='cuda:0'), covar=tensor([0.1942, 0.1195, 0.1646, 0.0717, 0.0683, 0.1054, 0.0734, 0.0647], device='cuda:0'), in_proj_covar=tensor([0.0619, 0.0763, 0.0892, 0.0780, 0.0585, 0.0616, 0.0631, 0.0733], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:02:04,526 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=208932.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:02:36,221 INFO [train.py:904] (0/8) Epoch 21, batch 5950, loss[loss=0.2096, simple_loss=0.2941, pruned_loss=0.06251, over 11920.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2948, pruned_loss=0.06065, over 3063606.83 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:02,965 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=208969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:03:36,550 INFO [optim.py:368] (0/8) 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 07:03:54,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8275, 2.4474, 2.0540, 2.1639, 2.7944, 2.4375, 2.6658, 2.9273], device='cuda:0'), covar=tensor([0.0200, 0.0407, 0.0492, 0.0493, 0.0243, 0.0359, 0.0212, 0.0252], device='cuda:0'), in_proj_covar=tensor([0.0205, 0.0232, 0.0223, 0.0224, 0.0233, 0.0231, 0.0232, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:03:56,994 INFO [train.py:904] (0/8) Epoch 21, batch 6000, loss[loss=0.2135, simple_loss=0.2928, pruned_loss=0.06708, over 11248.00 frames. ], tot_loss[loss=0.2072, simple_loss=0.2938, pruned_loss=0.06029, over 3067680.63 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:03:56,995 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 07:04:08,272 INFO [train.py:938] (0/8) Epoch 21, validation: loss=0.1512, simple_loss=0.2639, pruned_loss=0.01924, over 944034.00 frames. 2023-05-01 07:04:08,273 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 07:05:25,691 INFO [train.py:904] (0/8) Epoch 21, batch 6050, loss[loss=0.2425, simple_loss=0.3062, pruned_loss=0.08943, over 11779.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2922, pruned_loss=0.0595, over 3091712.60 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:05:29,758 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2023-05-01 07:06:26,993 INFO [optim.py:368] (0/8) 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 07:06:45,929 INFO [train.py:904] (0/8) Epoch 21, batch 6100, loss[loss=0.1831, simple_loss=0.2794, pruned_loss=0.04342, over 16780.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2917, pruned_loss=0.05845, over 3092945.21 frames. ], batch size: 96, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:06:55,553 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 07:07:38,942 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8834, 2.0951, 2.3746, 3.1451, 2.1995, 2.3180, 2.3321, 2.2566], device='cuda:0'), covar=tensor([0.1324, 0.3147, 0.2529, 0.0671, 0.3822, 0.2277, 0.2821, 0.3079], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0444, 0.0364, 0.0325, 0.0433, 0.0510, 0.0414, 0.0519], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:07:59,472 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 07:08:04,719 INFO [train.py:904] (0/8) Epoch 21, batch 6150, loss[loss=0.2196, simple_loss=0.3016, pruned_loss=0.06878, over 16879.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2905, pruned_loss=0.05831, over 3083242.11 frames. ], batch size: 116, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:04,304 INFO [optim.py:368] (0/8) 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 07:09:23,303 INFO [train.py:904] (0/8) Epoch 21, batch 6200, loss[loss=0.2129, simple_loss=0.2875, pruned_loss=0.06917, over 11674.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2886, pruned_loss=0.05785, over 3082176.90 frames. ], batch size: 248, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:09:23,914 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:09,571 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209232.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:16,383 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0279, 3.3179, 3.3185, 2.0861, 2.9953, 2.2887, 3.5013, 3.5220], device='cuda:0'), covar=tensor([0.0258, 0.0819, 0.0629, 0.1975, 0.0816, 0.0968, 0.0593, 0.0931], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 07:10:39,917 INFO [train.py:904] (0/8) Epoch 21, batch 6250, loss[loss=0.2179, simple_loss=0.3034, pruned_loss=0.06614, over 16679.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2882, pruned_loss=0.05761, over 3084938.06 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:10:45,645 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:10:57,086 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209263.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:22,155 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 07:11:49,917 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209298.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:11:51,866 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2434, 2.3355, 2.2534, 3.9791, 2.2064, 2.6519, 2.3857, 2.4879], device='cuda:0'), covar=tensor([0.1309, 0.3227, 0.2884, 0.0494, 0.3813, 0.2308, 0.3291, 0.3039], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0445, 0.0365, 0.0325, 0.0434, 0.0511, 0.0414, 0.0519], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:11:54,780 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-01 07:11:54,992 INFO [train.py:904] (0/8) Epoch 21, batch 6300, loss[loss=0.2048, simple_loss=0.2814, pruned_loss=0.0641, over 16600.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2877, pruned_loss=0.05703, over 3088768.70 frames. ], batch size: 57, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:12:07,901 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9698, 4.0497, 4.2966, 4.2719, 4.3008, 4.0211, 4.0424, 4.0137], device='cuda:0'), covar=tensor([0.0351, 0.0613, 0.0435, 0.0463, 0.0483, 0.0491, 0.0929, 0.0546], device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0445, 0.0434, 0.0402, 0.0479, 0.0456, 0.0541, 0.0366], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 07:12:17,994 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:13:13,582 INFO [train.py:904] (0/8) Epoch 21, batch 6350, loss[loss=0.1948, simple_loss=0.2832, pruned_loss=0.05322, over 16430.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2883, pruned_loss=0.05852, over 3059129.32 frames. ], batch size: 146, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:13:24,867 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209359.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:14:13,989 INFO [optim.py:368] (0/8) 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 07:14:31,832 INFO [train.py:904] (0/8) Epoch 21, batch 6400, loss[loss=0.2567, simple_loss=0.3247, pruned_loss=0.09435, over 10914.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.29, pruned_loss=0.06059, over 3035823.10 frames. ], batch size: 246, lr: 3.21e-03, grad_scale: 8.0 2023-05-01 07:15:06,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4543, 3.3538, 3.7299, 1.8238, 3.9259, 3.9184, 2.9740, 2.8591], device='cuda:0'), covar=tensor([0.0776, 0.0273, 0.0183, 0.1195, 0.0071, 0.0200, 0.0412, 0.0507], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0106, 0.0095, 0.0136, 0.0078, 0.0122, 0.0126, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 07:15:20,961 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 07:15:48,017 INFO [train.py:904] (0/8) Epoch 21, batch 6450, loss[loss=0.2065, simple_loss=0.2937, pruned_loss=0.05969, over 16687.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2903, pruned_loss=0.05976, over 3053209.22 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:16:10,227 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209466.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:16:52,761 INFO [optim.py:368] (0/8) 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 07:16:58,162 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9360, 2.0222, 2.2046, 3.4018, 2.0450, 2.3390, 2.1822, 2.1560], device='cuda:0'), covar=tensor([0.1497, 0.3745, 0.2876, 0.0679, 0.4443, 0.2608, 0.3496, 0.3566], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0444, 0.0363, 0.0325, 0.0433, 0.0511, 0.0413, 0.0519], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:17:02,818 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 07:17:08,456 INFO [train.py:904] (0/8) Epoch 21, batch 6500, loss[loss=0.1931, simple_loss=0.2743, pruned_loss=0.05596, over 16654.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2883, pruned_loss=0.05889, over 3075282.87 frames. ], batch size: 134, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:17:20,778 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209510.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:17:48,041 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209527.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:18:29,352 INFO [train.py:904] (0/8) Epoch 21, batch 6550, loss[loss=0.1983, simple_loss=0.3088, pruned_loss=0.04392, over 15387.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2908, pruned_loss=0.0598, over 3066144.63 frames. ], batch size: 191, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:18:40,053 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209558.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:00,591 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209571.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:19:22,037 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 07:19:33,744 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.848e+02 2.646e+02 3.318e+02 3.875e+02 1.016e+03, threshold=6.636e+02, percent-clipped=2.0 2023-05-01 07:19:49,526 INFO [train.py:904] (0/8) Epoch 21, batch 6600, loss[loss=0.2046, simple_loss=0.298, pruned_loss=0.05562, over 16774.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2925, pruned_loss=0.0602, over 3077943.50 frames. ], batch size: 76, lr: 3.21e-03, grad_scale: 4.0 2023-05-01 07:19:56,047 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209606.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:20:03,689 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:07,701 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 07:21:08,006 INFO [train.py:904] (0/8) Epoch 21, batch 6650, loss[loss=0.1747, simple_loss=0.2611, pruned_loss=0.04415, over 16776.00 frames. ], tot_loss[loss=0.208, simple_loss=0.2932, pruned_loss=0.06139, over 3063689.81 frames. ], batch size: 124, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:21:12,213 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209654.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:21:32,346 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209667.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:04,961 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=209688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:22:11,942 INFO [optim.py:368] (0/8) 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 07:22:25,669 INFO [train.py:904] (0/8) Epoch 21, batch 6700, loss[loss=0.1887, simple_loss=0.2775, pruned_loss=0.04996, over 16799.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2921, pruned_loss=0.06164, over 3056528.21 frames. ], batch size: 83, lr: 3.21e-03, grad_scale: 2.0 2023-05-01 07:22:31,709 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 07:23:40,234 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=209749.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:23:44,120 INFO [train.py:904] (0/8) Epoch 21, batch 6750, loss[loss=0.2055, simple_loss=0.2902, pruned_loss=0.06041, over 16738.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2907, pruned_loss=0.06137, over 3064155.92 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 2.0 2023-05-01 07:24:47,263 INFO [optim.py:368] (0/8) 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 07:25:01,552 INFO [train.py:904] (0/8) Epoch 21, batch 6800, loss[loss=0.2426, simple_loss=0.3169, pruned_loss=0.08415, over 11908.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2903, pruned_loss=0.061, over 3088627.30 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:25:33,303 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:19,673 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2133, 3.2495, 1.8693, 3.4178, 2.3996, 3.4770, 2.1311, 2.6360], device='cuda:0'), covar=tensor([0.0290, 0.0363, 0.1786, 0.0281, 0.0924, 0.0674, 0.1518, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0176, 0.0194, 0.0161, 0.0176, 0.0215, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 07:26:20,416 INFO [train.py:904] (0/8) Epoch 21, batch 6850, loss[loss=0.2002, simple_loss=0.2988, pruned_loss=0.05084, over 16273.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2912, pruned_loss=0.06068, over 3101434.36 frames. ], batch size: 165, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:26:28,946 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:26:29,585 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.49 vs. limit=5.0 2023-05-01 07:26:41,686 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:01,359 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0047, 2.8738, 2.4509, 2.5565, 3.3742, 3.0441, 3.5727, 3.6769], device='cuda:0'), covar=tensor([0.0107, 0.0513, 0.0577, 0.0471, 0.0279, 0.0361, 0.0220, 0.0249], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0231, 0.0223, 0.0224, 0.0232, 0.0230, 0.0231, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:27:21,615 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 21, batch 6900, loss[loss=0.2067, simple_loss=0.3009, pruned_loss=0.05629, over 16697.00 frames. ], tot_loss[loss=0.2075, simple_loss=0.2934, pruned_loss=0.06079, over 3081930.74 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:27:41,304 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:48,369 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:27:50,389 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3598, 2.8668, 2.6687, 2.2997, 2.2956, 2.2935, 2.8880, 2.8749], device='cuda:0'), covar=tensor([0.2230, 0.0727, 0.1453, 0.2240, 0.2074, 0.1979, 0.0467, 0.1240], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0271, 0.0303, 0.0311, 0.0296, 0.0257, 0.0294, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 07:27:50,582 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 07:28:51,101 INFO [train.py:904] (0/8) Epoch 21, batch 6950, loss[loss=0.2034, simple_loss=0.2842, pruned_loss=0.06126, over 11573.00 frames. ], tot_loss[loss=0.2103, simple_loss=0.2953, pruned_loss=0.06262, over 3063474.81 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:28:53,963 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=209954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:02,113 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=209959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:06,375 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=209962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:29:51,942 INFO [optim.py:368] (0/8) 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 07:30:03,430 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-210000.pt 2023-05-01 07:30:09,084 INFO [train.py:904] (0/8) Epoch 21, batch 7000, loss[loss=0.2139, simple_loss=0.2988, pruned_loss=0.06453, over 15426.00 frames. ], tot_loss[loss=0.2096, simple_loss=0.2953, pruned_loss=0.0619, over 3073572.97 frames. ], batch size: 190, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:30:09,335 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210002.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:30:49,949 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2589, 2.4228, 2.4754, 4.0810, 2.3211, 2.8264, 2.4639, 2.5575], device='cuda:0'), covar=tensor([0.1343, 0.3430, 0.2652, 0.0482, 0.3810, 0.2296, 0.3406, 0.3155], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0444, 0.0364, 0.0324, 0.0434, 0.0510, 0.0414, 0.0518], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:31:11,235 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:31:22,819 INFO [train.py:904] (0/8) Epoch 21, batch 7050, loss[loss=0.1908, simple_loss=0.2897, pruned_loss=0.04597, over 16598.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2961, pruned_loss=0.0618, over 3066889.75 frames. ], batch size: 62, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:31:30,576 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:32:24,160 INFO [optim.py:368] (0/8) 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 07:32:37,548 INFO [train.py:904] (0/8) Epoch 21, batch 7100, loss[loss=0.2357, simple_loss=0.2963, pruned_loss=0.08753, over 11239.00 frames. ], tot_loss[loss=0.2088, simple_loss=0.2947, pruned_loss=0.06151, over 3056352.08 frames. ], batch size: 248, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:33:03,283 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:08,085 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 07:33:09,194 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210122.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:33:11,666 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 07:33:55,441 INFO [train.py:904] (0/8) Epoch 21, batch 7150, loss[loss=0.1964, simple_loss=0.2814, pruned_loss=0.05575, over 16640.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.2929, pruned_loss=0.06091, over 3074990.21 frames. ], batch size: 62, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:34:16,445 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:21,261 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210170.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:34:47,410 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-01 07:34:53,947 INFO [optim.py:368] (0/8) 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 07:35:07,923 INFO [train.py:904] (0/8) Epoch 21, batch 7200, loss[loss=0.2075, simple_loss=0.2955, pruned_loss=0.05977, over 15356.00 frames. ], tot_loss[loss=0.205, simple_loss=0.291, pruned_loss=0.05944, over 3065996.26 frames. ], batch size: 191, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:35:13,843 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7634, 3.5921, 3.6301, 3.9001, 3.9654, 3.6276, 3.9100, 3.9648], device='cuda:0'), covar=tensor([0.1611, 0.1429, 0.1778, 0.0902, 0.0871, 0.2566, 0.1167, 0.1140], device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0769, 0.0890, 0.0783, 0.0589, 0.0618, 0.0639, 0.0735], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:35:26,244 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:35:36,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3968, 4.6783, 4.4871, 4.4703, 4.1713, 4.1553, 4.1883, 4.7103], device='cuda:0'), covar=tensor([0.1124, 0.0799, 0.0980, 0.0875, 0.0814, 0.1702, 0.1041, 0.0884], device='cuda:0'), in_proj_covar=tensor([0.0662, 0.0800, 0.0670, 0.0610, 0.0510, 0.0520, 0.0674, 0.0629], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:36:28,297 INFO [train.py:904] (0/8) Epoch 21, batch 7250, loss[loss=0.1857, simple_loss=0.2583, pruned_loss=0.05654, over 17025.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2886, pruned_loss=0.05844, over 3071092.58 frames. ], batch size: 55, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:36:43,120 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 07:36:43,781 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210262.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:37:31,775 INFO [optim.py:368] (0/8) 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 07:37:44,661 INFO [train.py:904] (0/8) Epoch 21, batch 7300, loss[loss=0.2067, simple_loss=0.2917, pruned_loss=0.06088, over 15123.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2881, pruned_loss=0.05838, over 3071960.78 frames. ], batch size: 190, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:37:50,521 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 07:37:58,791 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:44,666 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0187, 2.2807, 2.3064, 2.6264, 1.9318, 3.1327, 1.8234, 2.6614], device='cuda:0'), covar=tensor([0.1038, 0.0647, 0.1072, 0.0172, 0.0144, 0.0368, 0.1307, 0.0702], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0174, 0.0196, 0.0189, 0.0207, 0.0215, 0.0202, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 07:38:48,183 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210344.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:38:59,617 INFO [train.py:904] (0/8) Epoch 21, batch 7350, loss[loss=0.2032, simple_loss=0.2875, pruned_loss=0.05947, over 17282.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2894, pruned_loss=0.05933, over 3050614.45 frames. ], batch size: 52, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:40:00,130 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210392.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:40:02,087 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.675e+02 2.648e+02 3.204e+02 3.841e+02 6.410e+02, threshold=6.409e+02, percent-clipped=2.0 2023-05-01 07:40:11,056 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3756, 3.3480, 3.4170, 3.5016, 3.5182, 3.2865, 3.4819, 3.5713], device='cuda:0'), covar=tensor([0.1165, 0.0929, 0.1032, 0.0637, 0.0684, 0.2391, 0.1106, 0.0742], device='cuda:0'), in_proj_covar=tensor([0.0618, 0.0762, 0.0882, 0.0777, 0.0584, 0.0612, 0.0632, 0.0727], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:40:14,865 INFO [train.py:904] (0/8) Epoch 21, batch 7400, loss[loss=0.1907, simple_loss=0.2822, pruned_loss=0.04961, over 16526.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2896, pruned_loss=0.05937, over 3068766.03 frames. ], batch size: 68, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:40:26,990 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4728, 3.5291, 3.3192, 2.9432, 3.1743, 3.4459, 3.3271, 3.2773], device='cuda:0'), covar=tensor([0.0578, 0.0619, 0.0273, 0.0247, 0.0493, 0.0445, 0.1232, 0.0475], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0406, 0.0327, 0.0322, 0.0337, 0.0376, 0.0226, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:40:32,236 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210413.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:40:44,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0971, 2.1468, 2.2190, 3.7897, 2.0736, 2.5425, 2.2486, 2.2848], device='cuda:0'), covar=tensor([0.1411, 0.3567, 0.2983, 0.0543, 0.4282, 0.2446, 0.3610, 0.3190], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0445, 0.0362, 0.0324, 0.0435, 0.0512, 0.0414, 0.0518], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:41:02,781 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.84 vs. limit=5.0 2023-05-01 07:41:32,236 INFO [train.py:904] (0/8) Epoch 21, batch 7450, loss[loss=0.2146, simple_loss=0.3053, pruned_loss=0.06191, over 16143.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2909, pruned_loss=0.06038, over 3063286.91 frames. ], batch size: 165, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:41:57,721 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4085, 2.8761, 3.0727, 1.9092, 2.7466, 2.0821, 2.9695, 3.0588], device='cuda:0'), covar=tensor([0.0345, 0.0809, 0.0602, 0.2123, 0.0846, 0.1064, 0.0783, 0.0934], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0163, 0.0167, 0.0152, 0.0144, 0.0129, 0.0143, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 07:42:42,646 INFO [optim.py:368] (0/8) 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 07:42:48,266 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3298, 4.3387, 4.2273, 3.5115, 4.2509, 1.7551, 4.0256, 3.9191], device='cuda:0'), covar=tensor([0.0129, 0.0099, 0.0186, 0.0320, 0.0106, 0.2722, 0.0134, 0.0240], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0149, 0.0192, 0.0173, 0.0170, 0.0202, 0.0181, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:42:53,264 INFO [train.py:904] (0/8) Epoch 21, batch 7500, loss[loss=0.2016, simple_loss=0.2937, pruned_loss=0.05476, over 16798.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.291, pruned_loss=0.05956, over 3066098.23 frames. ], batch size: 102, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:43:25,801 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 07:43:49,983 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6294, 2.2684, 1.8113, 1.9455, 2.5956, 2.2106, 2.3984, 2.7451], device='cuda:0'), covar=tensor([0.0217, 0.0423, 0.0627, 0.0564, 0.0270, 0.0399, 0.0231, 0.0245], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0229, 0.0222, 0.0222, 0.0230, 0.0227, 0.0228, 0.0224], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:44:09,163 INFO [train.py:904] (0/8) Epoch 21, batch 7550, loss[loss=0.1862, simple_loss=0.2789, pruned_loss=0.0467, over 16752.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2903, pruned_loss=0.05958, over 3071708.30 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 4.0 2023-05-01 07:45:11,351 INFO [optim.py:368] (0/8) 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 07:45:23,241 INFO [train.py:904] (0/8) Epoch 21, batch 7600, loss[loss=0.2158, simple_loss=0.3007, pruned_loss=0.06548, over 16699.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2891, pruned_loss=0.05927, over 3083679.44 frames. ], batch size: 134, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:45:29,918 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 07:46:08,189 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210632.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:46:37,232 INFO [train.py:904] (0/8) Epoch 21, batch 7650, loss[loss=0.2011, simple_loss=0.2882, pruned_loss=0.05702, over 16574.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2893, pruned_loss=0.05967, over 3068501.59 frames. ], batch size: 75, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:46:59,531 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210666.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:25,405 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,082 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:47:42,762 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.936e+02 3.052e+02 3.513e+02 4.198e+02 7.732e+02, threshold=7.025e+02, percent-clipped=2.0 2023-05-01 07:47:55,361 INFO [train.py:904] (0/8) Epoch 21, batch 7700, loss[loss=0.1958, simple_loss=0.2783, pruned_loss=0.05666, over 16715.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2892, pruned_loss=0.05993, over 3063768.40 frames. ], batch size: 124, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:48:12,330 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=210713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:34,128 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210727.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:48:58,939 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:12,079 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-05-01 07:49:12,539 INFO [train.py:904] (0/8) Epoch 21, batch 7750, loss[loss=0.2262, simple_loss=0.3093, pruned_loss=0.07157, over 15495.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.2896, pruned_loss=0.05972, over 3080373.76 frames. ], batch size: 191, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:49:27,317 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=210761.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:32,463 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210764.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:49:36,151 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5258, 5.5607, 5.4032, 5.0361, 5.0422, 5.4609, 5.4329, 5.1166], device='cuda:0'), covar=tensor([0.0630, 0.0487, 0.0250, 0.0267, 0.0936, 0.0486, 0.0234, 0.0692], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0405, 0.0326, 0.0320, 0.0335, 0.0375, 0.0226, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 07:49:36,705 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 07:50:18,620 INFO [optim.py:368] (0/8) 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 07:50:31,011 INFO [train.py:904] (0/8) Epoch 21, batch 7800, loss[loss=0.1896, simple_loss=0.2784, pruned_loss=0.05042, over 16373.00 frames. ], tot_loss[loss=0.2059, simple_loss=0.2904, pruned_loss=0.06066, over 3073039.39 frames. ], batch size: 146, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:51:07,534 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=210825.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:51:48,920 INFO [train.py:904] (0/8) Epoch 21, batch 7850, loss[loss=0.2031, simple_loss=0.2962, pruned_loss=0.055, over 16471.00 frames. ], tot_loss[loss=0.2057, simple_loss=0.291, pruned_loss=0.06025, over 3081733.70 frames. ], batch size: 75, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:52:54,105 INFO [optim.py:368] (0/8) 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 07:53:05,689 INFO [train.py:904] (0/8) Epoch 21, batch 7900, loss[loss=0.2083, simple_loss=0.3035, pruned_loss=0.05653, over 16778.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2899, pruned_loss=0.05989, over 3077277.94 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:53:59,030 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 07:54:03,516 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-05-01 07:54:24,394 INFO [train.py:904] (0/8) Epoch 21, batch 7950, loss[loss=0.2325, simple_loss=0.3174, pruned_loss=0.07382, over 16873.00 frames. ], tot_loss[loss=0.205, simple_loss=0.29, pruned_loss=0.05998, over 3094712.76 frames. ], batch size: 116, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:54:33,163 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7307, 4.7631, 5.1156, 5.0885, 5.1343, 4.8091, 4.7612, 4.5828], device='cuda:0'), covar=tensor([0.0261, 0.0537, 0.0337, 0.0401, 0.0429, 0.0373, 0.0881, 0.0471], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0443, 0.0431, 0.0401, 0.0478, 0.0453, 0.0539, 0.0361], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 07:55:13,959 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=210984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:20,268 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=210988.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:55:29,114 INFO [optim.py:368] (0/8) 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 07:55:41,333 INFO [train.py:904] (0/8) Epoch 21, batch 8000, loss[loss=0.1863, simple_loss=0.2866, pruned_loss=0.04302, over 16782.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2914, pruned_loss=0.06084, over 3087746.75 frames. ], batch size: 83, lr: 3.20e-03, grad_scale: 8.0 2023-05-01 07:56:12,599 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211022.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:35,351 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211038.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:56:45,499 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 21, batch 8050, loss[loss=0.1848, simple_loss=0.2759, pruned_loss=0.04688, over 16788.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2906, pruned_loss=0.05983, over 3105322.62 frames. ], batch size: 83, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:57:56,778 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6267, 3.3935, 3.7781, 1.8633, 3.9524, 3.9886, 3.0007, 2.9267], device='cuda:0'), covar=tensor([0.0750, 0.0266, 0.0193, 0.1283, 0.0070, 0.0186, 0.0441, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0107, 0.0096, 0.0137, 0.0079, 0.0123, 0.0128, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 07:57:59,279 INFO [optim.py:368] (0/8) 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 07:58:08,122 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2108, 3.2892, 2.0121, 3.5906, 2.4833, 3.6069, 2.0881, 2.6685], device='cuda:0'), covar=tensor([0.0327, 0.0389, 0.1685, 0.0200, 0.0791, 0.0552, 0.1569, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0174, 0.0192, 0.0158, 0.0174, 0.0214, 0.0199, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 07:58:10,499 INFO [train.py:904] (0/8) Epoch 21, batch 8100, loss[loss=0.1939, simple_loss=0.2914, pruned_loss=0.04823, over 16750.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2901, pruned_loss=0.05894, over 3114064.72 frames. ], batch size: 76, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:58:38,228 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211120.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 07:59:05,224 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4025, 4.0833, 4.1219, 2.5647, 3.7043, 4.1015, 3.6403, 2.3294], device='cuda:0'), covar=tensor([0.0575, 0.0049, 0.0048, 0.0454, 0.0102, 0.0103, 0.0092, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0082, 0.0083, 0.0134, 0.0096, 0.0108, 0.0093, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 07:59:22,870 INFO [train.py:904] (0/8) Epoch 21, batch 8150, loss[loss=0.1938, simple_loss=0.2725, pruned_loss=0.0575, over 16632.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2878, pruned_loss=0.05855, over 3095204.66 frames. ], batch size: 134, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 07:59:36,398 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 08:00:18,998 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8640, 4.6748, 4.9113, 5.0854, 5.2876, 4.6975, 5.2342, 5.2644], device='cuda:0'), covar=tensor([0.1951, 0.1217, 0.1605, 0.0695, 0.0559, 0.0943, 0.0613, 0.0616], device='cuda:0'), in_proj_covar=tensor([0.0616, 0.0759, 0.0879, 0.0772, 0.0581, 0.0611, 0.0631, 0.0728], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:00:27,470 INFO [optim.py:368] (0/8) 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 08:00:34,038 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5273, 3.4962, 3.4618, 2.7823, 3.3553, 2.1712, 3.1629, 2.8270], device='cuda:0'), covar=tensor([0.0172, 0.0123, 0.0194, 0.0224, 0.0096, 0.2177, 0.0137, 0.0241], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0151, 0.0195, 0.0175, 0.0172, 0.0204, 0.0182, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:00:40,755 INFO [train.py:904] (0/8) Epoch 21, batch 8200, loss[loss=0.2237, simple_loss=0.293, pruned_loss=0.07725, over 11474.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2857, pruned_loss=0.05811, over 3085875.65 frames. ], batch size: 246, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:01:59,684 INFO [train.py:904] (0/8) Epoch 21, batch 8250, loss[loss=0.1705, simple_loss=0.2762, pruned_loss=0.03242, over 16821.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2843, pruned_loss=0.05547, over 3082829.22 frames. ], batch size: 102, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:02:37,451 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 08:02:56,792 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211288.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:03:06,692 INFO [optim.py:368] (0/8) 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 08:03:18,676 INFO [train.py:904] (0/8) Epoch 21, batch 8300, loss[loss=0.1756, simple_loss=0.2729, pruned_loss=0.03916, over 16902.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2814, pruned_loss=0.05256, over 3073397.64 frames. ], batch size: 109, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:03:51,730 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211322.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:12,349 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211336.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:15,556 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211338.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:18,460 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:04:24,860 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7967, 5.0920, 4.8472, 4.8787, 4.6820, 4.6428, 4.4876, 5.1475], device='cuda:0'), covar=tensor([0.1111, 0.0820, 0.0968, 0.0876, 0.0789, 0.1025, 0.1338, 0.0821], device='cuda:0'), in_proj_covar=tensor([0.0657, 0.0792, 0.0662, 0.0605, 0.0504, 0.0517, 0.0669, 0.0623], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:04:36,936 INFO [train.py:904] (0/8) Epoch 21, batch 8350, loss[loss=0.184, simple_loss=0.2813, pruned_loss=0.04341, over 16549.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2807, pruned_loss=0.05078, over 3066269.40 frames. ], batch size: 68, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:05:05,462 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,341 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:30,442 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211386.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:05:43,493 INFO [optim.py:368] (0/8) 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 08:05:55,520 INFO [train.py:904] (0/8) Epoch 21, batch 8400, loss[loss=0.1728, simple_loss=0.2673, pruned_loss=0.03913, over 16240.00 frames. ], tot_loss[loss=0.1879, simple_loss=0.2784, pruned_loss=0.04872, over 3067804.17 frames. ], batch size: 165, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:06:01,483 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2474, 2.1797, 2.1547, 3.9004, 2.1232, 2.4708, 2.3060, 2.3466], device='cuda:0'), covar=tensor([0.1205, 0.3821, 0.3103, 0.0495, 0.4347, 0.2622, 0.3673, 0.3444], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0436, 0.0357, 0.0316, 0.0427, 0.0499, 0.0405, 0.0509], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:06:10,883 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=211412.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:06:22,411 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:06:39,764 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4026, 4.3711, 4.7376, 4.7077, 4.7247, 4.4610, 4.4188, 4.3741], device='cuda:0'), covar=tensor([0.0305, 0.0687, 0.0389, 0.0458, 0.0422, 0.0418, 0.0943, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0445, 0.0430, 0.0401, 0.0477, 0.0451, 0.0537, 0.0362], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 08:07:03,731 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:10,973 INFO [train.py:904] (0/8) Epoch 21, batch 8450, loss[loss=0.1857, simple_loss=0.2668, pruned_loss=0.05229, over 12257.00 frames. ], tot_loss[loss=0.1856, simple_loss=0.2766, pruned_loss=0.04727, over 3063177.23 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:07:36,106 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:07:44,922 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=211473.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:08:15,870 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7458, 3.2638, 3.4092, 2.1082, 2.9360, 2.2097, 3.3591, 3.4239], device='cuda:0'), covar=tensor([0.0303, 0.0830, 0.0558, 0.2127, 0.0852, 0.1096, 0.0685, 0.0910], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0159, 0.0163, 0.0149, 0.0141, 0.0126, 0.0139, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 08:08:18,593 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.143e+02 2.501e+02 3.093e+02 5.428e+02, threshold=5.001e+02, percent-clipped=1.0 2023-05-01 08:08:29,990 INFO [train.py:904] (0/8) Epoch 21, batch 8500, loss[loss=0.166, simple_loss=0.2567, pruned_loss=0.03761, over 16493.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2731, pruned_loss=0.04482, over 3061461.59 frames. ], batch size: 146, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:09:33,169 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9243, 2.6737, 2.6411, 2.0400, 2.5083, 2.7841, 2.6696, 1.9701], device='cuda:0'), covar=tensor([0.0395, 0.0083, 0.0070, 0.0311, 0.0130, 0.0102, 0.0096, 0.0417], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0079, 0.0080, 0.0130, 0.0094, 0.0104, 0.0090, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 08:09:42,922 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 08:09:54,396 INFO [train.py:904] (0/8) Epoch 21, batch 8550, loss[loss=0.1688, simple_loss=0.2497, pruned_loss=0.04397, over 12018.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2706, pruned_loss=0.0439, over 3035031.04 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:11:18,422 INFO [optim.py:368] (0/8) 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 08:11:27,796 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4000, 3.5944, 3.3364, 3.0684, 3.0054, 3.5167, 3.2657, 3.2558], device='cuda:0'), covar=tensor([0.0820, 0.0610, 0.0414, 0.0416, 0.0888, 0.0507, 0.1842, 0.0628], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0409, 0.0328, 0.0324, 0.0335, 0.0377, 0.0227, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:11:32,299 INFO [train.py:904] (0/8) Epoch 21, batch 8600, loss[loss=0.1813, simple_loss=0.2643, pruned_loss=0.04915, over 12623.00 frames. ], tot_loss[loss=0.1786, simple_loss=0.2705, pruned_loss=0.04331, over 3013323.81 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:12:48,719 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=211640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:13:09,697 INFO [train.py:904] (0/8) Epoch 21, batch 8650, loss[loss=0.1681, simple_loss=0.2567, pruned_loss=0.03976, over 12280.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2682, pruned_loss=0.04193, over 2996724.90 frames. ], batch size: 250, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:13:47,527 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4164, 3.3350, 3.4908, 3.5525, 3.5907, 3.2942, 3.5619, 3.6407], device='cuda:0'), covar=tensor([0.1192, 0.0954, 0.1006, 0.0603, 0.0579, 0.2552, 0.0783, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0601, 0.0744, 0.0865, 0.0759, 0.0572, 0.0601, 0.0618, 0.0716], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:14:12,373 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-01 08:14:30,265 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=211688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:14:43,487 INFO [optim.py:368] (0/8) 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 08:14:57,335 INFO [train.py:904] (0/8) Epoch 21, batch 8700, loss[loss=0.1692, simple_loss=0.25, pruned_loss=0.04423, over 12227.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2654, pruned_loss=0.04047, over 3006266.49 frames. ], batch size: 247, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:15:24,217 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0429, 4.1241, 3.9186, 3.6261, 3.6644, 4.0422, 3.7239, 3.8021], device='cuda:0'), covar=tensor([0.0600, 0.0628, 0.0309, 0.0316, 0.0657, 0.0539, 0.0973, 0.0580], device='cuda:0'), in_proj_covar=tensor([0.0280, 0.0407, 0.0326, 0.0323, 0.0335, 0.0377, 0.0226, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:16:09,885 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:16:31,431 INFO [train.py:904] (0/8) Epoch 21, batch 8750, loss[loss=0.1792, simple_loss=0.2808, pruned_loss=0.03882, over 16595.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2656, pruned_loss=0.03985, over 3029333.68 frames. ], batch size: 62, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:17:13,767 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=211768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:18:09,823 INFO [optim.py:368] (0/8) 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 08:18:23,768 INFO [train.py:904] (0/8) Epoch 21, batch 8800, loss[loss=0.1667, simple_loss=0.2647, pruned_loss=0.03434, over 15334.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2644, pruned_loss=0.03888, over 3036564.14 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:20:07,844 INFO [train.py:904] (0/8) Epoch 21, batch 8850, loss[loss=0.1663, simple_loss=0.2555, pruned_loss=0.03858, over 12543.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2663, pruned_loss=0.03783, over 3032211.50 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:20:36,669 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1094, 3.5467, 3.5965, 2.3671, 3.1806, 3.6174, 3.4167, 2.1205], device='cuda:0'), covar=tensor([0.0547, 0.0050, 0.0052, 0.0417, 0.0121, 0.0084, 0.0086, 0.0464], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0079, 0.0081, 0.0132, 0.0095, 0.0105, 0.0091, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 08:21:38,909 INFO [optim.py:368] (0/8) 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 08:21:54,333 INFO [train.py:904] (0/8) Epoch 21, batch 8900, loss[loss=0.1904, simple_loss=0.2805, pruned_loss=0.05014, over 16543.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2672, pruned_loss=0.03732, over 3044455.27 frames. ], batch size: 62, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:23:57,817 INFO [train.py:904] (0/8) Epoch 21, batch 8950, loss[loss=0.1822, simple_loss=0.2685, pruned_loss=0.04795, over 12830.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2667, pruned_loss=0.03747, over 3052331.78 frames. ], batch size: 248, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:29,297 INFO [optim.py:368] (0/8) 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 08:25:40,948 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-212000.pt 2023-05-01 08:25:46,911 INFO [train.py:904] (0/8) Epoch 21, batch 9000, loss[loss=0.1651, simple_loss=0.2582, pruned_loss=0.03598, over 15149.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2633, pruned_loss=0.03604, over 3061866.62 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:25:46,912 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 08:25:57,433 INFO [train.py:938] (0/8) Epoch 21, validation: loss=0.1457, simple_loss=0.2498, pruned_loss=0.02077, over 944034.00 frames. 2023-05-01 08:25:57,434 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 08:27:19,831 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212042.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:27:39,213 INFO [train.py:904] (0/8) Epoch 21, batch 9050, loss[loss=0.1725, simple_loss=0.2605, pruned_loss=0.04222, over 15470.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2651, pruned_loss=0.03665, over 3073773.99 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:28:14,916 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:28:58,638 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212090.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:29:10,428 INFO [optim.py:368] (0/8) 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 08:29:26,261 INFO [train.py:904] (0/8) Epoch 21, batch 9100, loss[loss=0.1693, simple_loss=0.2702, pruned_loss=0.03416, over 16904.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2649, pruned_loss=0.03733, over 3078226.85 frames. ], batch size: 102, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:29:30,920 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212104.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:29:53,446 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:30:50,946 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.21 vs. limit=5.0 2023-05-01 08:31:15,061 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2039, 3.5157, 3.6148, 2.4364, 3.2248, 3.6038, 3.3833, 2.0873], device='cuda:0'), covar=tensor([0.0529, 0.0056, 0.0052, 0.0394, 0.0117, 0.0080, 0.0103, 0.0481], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0080, 0.0081, 0.0132, 0.0095, 0.0105, 0.0092, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 08:31:22,753 INFO [train.py:904] (0/8) Epoch 21, batch 9150, loss[loss=0.173, simple_loss=0.2652, pruned_loss=0.04035, over 16258.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2655, pruned_loss=0.03692, over 3088009.75 frames. ], batch size: 166, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:31:52,285 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212165.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:31:57,181 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4417, 4.7091, 4.5043, 4.5249, 4.2354, 4.2169, 4.1718, 4.7206], device='cuda:0'), covar=tensor([0.1030, 0.0912, 0.0902, 0.0807, 0.0785, 0.1496, 0.1106, 0.0914], device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0781, 0.0649, 0.0594, 0.0497, 0.0507, 0.0655, 0.0612], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:31:57,349 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8305, 2.6259, 2.3300, 3.5805, 2.0501, 3.6694, 1.6390, 2.8642], device='cuda:0'), covar=tensor([0.1322, 0.0692, 0.1193, 0.0153, 0.0105, 0.0353, 0.1575, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0171, 0.0192, 0.0184, 0.0200, 0.0211, 0.0200, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 08:32:13,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8386, 4.8360, 4.6023, 4.0809, 4.7356, 1.9725, 4.4928, 4.4291], device='cuda:0'), covar=tensor([0.0236, 0.0193, 0.0264, 0.0332, 0.0227, 0.2425, 0.0148, 0.0243], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0147, 0.0189, 0.0168, 0.0167, 0.0200, 0.0177, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:32:49,429 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.82 vs. limit=5.0 2023-05-01 08:32:57,280 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.621e+02 2.079e+02 2.542e+02 3.144e+02 5.396e+02, threshold=5.084e+02, percent-clipped=1.0 2023-05-01 08:33:02,172 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2957, 4.1612, 4.3463, 4.5023, 4.6295, 4.1653, 4.5998, 4.6312], device='cuda:0'), covar=tensor([0.1662, 0.1115, 0.1449, 0.0657, 0.0527, 0.1215, 0.0679, 0.0713], device='cuda:0'), in_proj_covar=tensor([0.0594, 0.0734, 0.0853, 0.0748, 0.0565, 0.0594, 0.0610, 0.0707], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:33:09,952 INFO [train.py:904] (0/8) Epoch 21, batch 9200, loss[loss=0.1803, simple_loss=0.2721, pruned_loss=0.04425, over 15376.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2611, pruned_loss=0.03598, over 3102236.28 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:13,793 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3954, 3.0328, 2.7157, 2.2590, 2.1868, 2.2944, 2.9867, 2.8145], device='cuda:0'), covar=tensor([0.2646, 0.0658, 0.1599, 0.2901, 0.2635, 0.2218, 0.0438, 0.1368], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0262, 0.0298, 0.0304, 0.0285, 0.0252, 0.0285, 0.0325], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 08:34:48,829 INFO [train.py:904] (0/8) Epoch 21, batch 9250, loss[loss=0.1561, simple_loss=0.256, pruned_loss=0.02807, over 15319.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2603, pruned_loss=0.03583, over 3092747.33 frames. ], batch size: 191, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:34:51,539 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212253.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:36:25,640 INFO [optim.py:368] (0/8) 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 08:36:39,758 INFO [train.py:904] (0/8) Epoch 21, batch 9300, loss[loss=0.151, simple_loss=0.2447, pruned_loss=0.02863, over 16791.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2592, pruned_loss=0.03559, over 3086379.38 frames. ], batch size: 83, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:37:07,701 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212314.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:38:01,787 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0669, 5.1055, 4.9253, 4.5518, 4.6197, 5.0496, 4.8965, 4.6598], device='cuda:0'), covar=tensor([0.0624, 0.0654, 0.0319, 0.0316, 0.1064, 0.0516, 0.0304, 0.0757], device='cuda:0'), in_proj_covar=tensor([0.0277, 0.0401, 0.0324, 0.0319, 0.0330, 0.0372, 0.0224, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:38:15,131 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 08:38:22,845 INFO [train.py:904] (0/8) Epoch 21, batch 9350, loss[loss=0.1897, simple_loss=0.2811, pruned_loss=0.04917, over 15329.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2591, pruned_loss=0.03582, over 3091929.96 frames. ], batch size: 190, lr: 3.19e-03, grad_scale: 8.0 2023-05-01 08:39:47,853 INFO [optim.py:368] (0/8) 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 08:39:57,746 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5741, 2.4233, 2.4097, 3.6308, 2.0833, 3.7746, 1.4588, 2.8622], device='cuda:0'), covar=tensor([0.1527, 0.0845, 0.1224, 0.0186, 0.0103, 0.0332, 0.1754, 0.0768], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0170, 0.0191, 0.0182, 0.0197, 0.0210, 0.0199, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 08:40:03,006 INFO [train.py:904] (0/8) Epoch 21, batch 9400, loss[loss=0.1822, simple_loss=0.2813, pruned_loss=0.04156, over 16904.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2591, pruned_loss=0.03561, over 3083310.98 frames. ], batch size: 116, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:14,893 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 08:41:18,522 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212439.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:41:42,927 INFO [train.py:904] (0/8) Epoch 21, batch 9450, loss[loss=0.1717, simple_loss=0.2691, pruned_loss=0.03709, over 16672.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2607, pruned_loss=0.03586, over 3079211.26 frames. ], batch size: 134, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:41:58,676 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212460.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 08:43:11,248 INFO [optim.py:368] (0/8) 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 08:43:21,161 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:43:25,078 INFO [train.py:904] (0/8) Epoch 21, batch 9500, loss[loss=0.189, simple_loss=0.2876, pruned_loss=0.04516, over 16127.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2599, pruned_loss=0.0358, over 3055081.00 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:45:02,004 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5891, 2.0620, 1.7938, 1.8225, 2.3829, 2.0521, 1.9617, 2.4858], device='cuda:0'), covar=tensor([0.0186, 0.0473, 0.0549, 0.0570, 0.0283, 0.0468, 0.0204, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0225, 0.0218, 0.0217, 0.0225, 0.0224, 0.0221, 0.0218], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:45:12,079 INFO [train.py:904] (0/8) Epoch 21, batch 9550, loss[loss=0.154, simple_loss=0.2441, pruned_loss=0.03196, over 17021.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2599, pruned_loss=0.03607, over 3069016.45 frames. ], batch size: 50, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:45:19,569 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0253, 4.9749, 4.7329, 4.2874, 4.8659, 1.9994, 4.6052, 4.5875], device='cuda:0'), covar=tensor([0.0084, 0.0115, 0.0229, 0.0334, 0.0103, 0.2521, 0.0140, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0145, 0.0185, 0.0164, 0.0164, 0.0197, 0.0174, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:46:39,908 INFO [optim.py:368] (0/8) 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 08:46:50,971 INFO [train.py:904] (0/8) Epoch 21, batch 9600, loss[loss=0.1576, simple_loss=0.2454, pruned_loss=0.03493, over 12375.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2615, pruned_loss=0.03673, over 3060689.53 frames. ], batch size: 247, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:47:05,845 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212609.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 08:48:38,610 INFO [train.py:904] (0/8) Epoch 21, batch 9650, loss[loss=0.1671, simple_loss=0.2534, pruned_loss=0.04035, over 12691.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2632, pruned_loss=0.03701, over 3044646.90 frames. ], batch size: 250, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:48:54,747 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8323, 3.1960, 3.4933, 2.0531, 2.9125, 2.1733, 3.3076, 3.2701], device='cuda:0'), covar=tensor([0.0246, 0.0805, 0.0462, 0.1991, 0.0778, 0.0989, 0.0692, 0.0977], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0154, 0.0161, 0.0147, 0.0139, 0.0125, 0.0138, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 08:49:24,727 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9839, 2.6744, 2.8720, 2.0247, 2.6625, 2.0598, 2.6900, 2.7952], device='cuda:0'), covar=tensor([0.0302, 0.0894, 0.0466, 0.1937, 0.0820, 0.0952, 0.0699, 0.0911], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0154, 0.0161, 0.0147, 0.0139, 0.0125, 0.0138, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 08:49:47,551 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6490, 4.5062, 4.6909, 4.8169, 4.9912, 4.4553, 5.0399, 5.0036], device='cuda:0'), covar=tensor([0.1884, 0.1201, 0.1609, 0.0713, 0.0528, 0.0938, 0.0495, 0.0711], device='cuda:0'), in_proj_covar=tensor([0.0593, 0.0728, 0.0849, 0.0744, 0.0564, 0.0591, 0.0609, 0.0704], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:50:11,920 INFO [optim.py:368] (0/8) 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 08:50:20,714 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212698.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:50:28,039 INFO [train.py:904] (0/8) Epoch 21, batch 9700, loss[loss=0.1536, simple_loss=0.2541, pruned_loss=0.02651, over 16868.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.262, pruned_loss=0.03669, over 3050063.13 frames. ], batch size: 102, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:50:30,780 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1162, 1.5465, 1.9120, 2.0781, 2.2373, 2.3108, 1.7827, 2.2151], device='cuda:0'), covar=tensor([0.0287, 0.0554, 0.0309, 0.0356, 0.0318, 0.0238, 0.0522, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0188, 0.0172, 0.0176, 0.0190, 0.0147, 0.0190, 0.0143], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 08:50:46,013 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:10,272 INFO [train.py:904] (0/8) Epoch 21, batch 9750, loss[loss=0.1604, simple_loss=0.2598, pruned_loss=0.03047, over 16356.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.261, pruned_loss=0.03673, over 3048735.50 frames. ], batch size: 165, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:52:24,895 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:26,946 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212760.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 08:52:29,384 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 08:52:49,215 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:52:53,145 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=212774.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:36,940 INFO [optim.py:368] (0/8) 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 08:53:37,581 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=212795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:53:46,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1610, 2.8689, 3.0677, 1.6125, 3.2529, 3.3831, 2.7263, 2.5884], device='cuda:0'), covar=tensor([0.0839, 0.0320, 0.0204, 0.1375, 0.0099, 0.0162, 0.0466, 0.0476], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0103, 0.0091, 0.0134, 0.0076, 0.0117, 0.0123, 0.0124], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 08:53:48,995 INFO [train.py:904] (0/8) Epoch 21, batch 9800, loss[loss=0.1566, simple_loss=0.2424, pruned_loss=0.03542, over 12576.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2614, pruned_loss=0.03642, over 3049054.44 frames. ], batch size: 247, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:54:01,432 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=212808.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 08:54:05,601 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-01 08:54:10,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7847, 3.7734, 4.0204, 3.7419, 3.9780, 4.3398, 3.9438, 3.6455], device='cuda:0'), covar=tensor([0.2313, 0.2518, 0.2228, 0.2324, 0.2339, 0.1474, 0.1630, 0.2560], device='cuda:0'), in_proj_covar=tensor([0.0380, 0.0556, 0.0614, 0.0463, 0.0613, 0.0645, 0.0485, 0.0620], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 08:54:35,146 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 08:54:51,521 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=212835.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:55:31,552 INFO [train.py:904] (0/8) Epoch 21, batch 9850, loss[loss=0.1747, simple_loss=0.2661, pruned_loss=0.04169, over 16779.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2627, pruned_loss=0.03637, over 3053294.40 frames. ], batch size: 124, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:56:28,019 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3621, 2.9848, 2.6493, 2.2265, 2.1584, 2.2757, 2.9747, 2.7925], device='cuda:0'), covar=tensor([0.2650, 0.0693, 0.1684, 0.2886, 0.2726, 0.2189, 0.0466, 0.1351], device='cuda:0'), in_proj_covar=tensor([0.0316, 0.0259, 0.0296, 0.0301, 0.0282, 0.0249, 0.0282, 0.0320], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 08:57:07,915 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.603e+02 2.176e+02 2.812e+02 3.349e+02 5.684e+02, threshold=5.624e+02, percent-clipped=7.0 2023-05-01 08:57:09,106 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4683, 3.4815, 1.8177, 3.8433, 2.5002, 3.7831, 1.9565, 2.7808], device='cuda:0'), covar=tensor([0.0284, 0.0380, 0.2082, 0.0244, 0.0963, 0.0507, 0.2026, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0169, 0.0189, 0.0152, 0.0172, 0.0207, 0.0196, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-01 08:57:21,372 INFO [train.py:904] (0/8) Epoch 21, batch 9900, loss[loss=0.1952, simple_loss=0.2987, pruned_loss=0.04588, over 16929.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2631, pruned_loss=0.03637, over 3046225.52 frames. ], batch size: 125, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:57:38,700 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=212909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 08:58:16,459 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 08:59:16,823 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.41 vs. limit=5.0 2023-05-01 08:59:17,577 INFO [train.py:904] (0/8) Epoch 21, batch 9950, loss[loss=0.1757, simple_loss=0.2849, pruned_loss=0.03328, over 15364.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.265, pruned_loss=0.0363, over 3050864.65 frames. ], batch size: 190, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 08:59:29,582 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 09:01:17,071 INFO [train.py:904] (0/8) Epoch 21, batch 10000, loss[loss=0.1713, simple_loss=0.27, pruned_loss=0.03626, over 15480.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2635, pruned_loss=0.03594, over 3064551.06 frames. ], batch size: 192, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:01:51,257 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4997, 3.6952, 2.7330, 2.1402, 2.2794, 2.3780, 3.8990, 3.1263], device='cuda:0'), covar=tensor([0.2970, 0.0497, 0.1815, 0.2937, 0.2817, 0.2103, 0.0421, 0.1262], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0258, 0.0295, 0.0300, 0.0280, 0.0249, 0.0282, 0.0319], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 09:02:58,021 INFO [train.py:904] (0/8) Epoch 21, batch 10050, loss[loss=0.1691, simple_loss=0.2648, pruned_loss=0.03675, over 16560.00 frames. ], tot_loss[loss=0.168, simple_loss=0.264, pruned_loss=0.03598, over 3071969.29 frames. ], batch size: 62, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:03:02,071 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213054.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:03:28,736 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:04:21,052 INFO [optim.py:368] (0/8) 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 09:04:21,594 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:04:32,169 INFO [train.py:904] (0/8) Epoch 21, batch 10100, loss[loss=0.1515, simple_loss=0.2525, pruned_loss=0.02529, over 16888.00 frames. ], tot_loss[loss=0.168, simple_loss=0.264, pruned_loss=0.03599, over 3040120.93 frames. ], batch size: 102, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:05:27,302 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213130.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:42,672 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:05:51,221 INFO [train.py:904] (0/8) Epoch 21, batch 10150, loss[loss=0.1705, simple_loss=0.2574, pruned_loss=0.04182, over 12554.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2629, pruned_loss=0.03609, over 3041717.13 frames. ], batch size: 246, lr: 3.18e-03, grad_scale: 8.0 2023-05-01 09:05:53,946 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-21.pt 2023-05-01 09:06:16,181 INFO [train.py:904] (0/8) Epoch 22, batch 0, loss[loss=0.2697, simple_loss=0.3303, pruned_loss=0.1045, over 16758.00 frames. ], tot_loss[loss=0.2697, simple_loss=0.3303, pruned_loss=0.1045, over 16758.00 frames. ], batch size: 124, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:06:16,182 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 09:06:23,633 INFO [train.py:938] (0/8) Epoch 22, validation: loss=0.1457, simple_loss=0.2489, pruned_loss=0.0212, over 944034.00 frames. 2023-05-01 09:06:23,634 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 09:06:31,408 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.02 vs. limit=5.0 2023-05-01 09:07:26,348 INFO [optim.py:368] (0/8) 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 09:07:34,023 INFO [train.py:904] (0/8) Epoch 22, batch 50, loss[loss=0.1862, simple_loss=0.2686, pruned_loss=0.05195, over 15927.00 frames. ], tot_loss[loss=0.1873, simple_loss=0.2695, pruned_loss=0.05255, over 753302.77 frames. ], batch size: 35, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:08:03,025 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-05-01 09:08:41,859 INFO [train.py:904] (0/8) Epoch 22, batch 100, loss[loss=0.2079, simple_loss=0.2797, pruned_loss=0.06801, over 16674.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2655, pruned_loss=0.04899, over 1323000.23 frames. ], batch size: 134, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:09:28,945 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2257, 3.5055, 3.8611, 2.2272, 3.1324, 2.3231, 3.6915, 3.6751], device='cuda:0'), covar=tensor([0.0291, 0.0893, 0.0499, 0.1903, 0.0801, 0.1037, 0.0562, 0.1021], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0156, 0.0162, 0.0149, 0.0141, 0.0126, 0.0139, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:09:44,708 INFO [optim.py:368] (0/8) 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 09:09:51,957 INFO [train.py:904] (0/8) Epoch 22, batch 150, loss[loss=0.1741, simple_loss=0.2531, pruned_loss=0.04756, over 16873.00 frames. ], tot_loss[loss=0.1788, simple_loss=0.2639, pruned_loss=0.04684, over 1757856.95 frames. ], batch size: 90, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:00,689 INFO [train.py:904] (0/8) Epoch 22, batch 200, loss[loss=0.1667, simple_loss=0.2425, pruned_loss=0.04541, over 16908.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2634, pruned_loss=0.04572, over 2113793.83 frames. ], batch size: 96, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:11:01,662 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.94 vs. limit=2.0 2023-05-01 09:11:02,273 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:11:12,738 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8361, 5.2261, 4.9692, 4.9569, 4.6930, 4.7049, 4.6899, 5.3341], device='cuda:0'), covar=tensor([0.1459, 0.0979, 0.1289, 0.0943, 0.1007, 0.1112, 0.1215, 0.0997], device='cuda:0'), in_proj_covar=tensor([0.0650, 0.0791, 0.0655, 0.0601, 0.0503, 0.0513, 0.0662, 0.0618], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:11:20,364 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.291e+02 2.770e+02 3.341e+02 7.349e+02, threshold=5.539e+02, percent-clipped=1.0 2023-05-01 09:12:07,195 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213402.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:08,111 INFO [train.py:904] (0/8) Epoch 22, batch 250, loss[loss=0.1683, simple_loss=0.2558, pruned_loss=0.04037, over 17115.00 frames. ], tot_loss[loss=0.1766, simple_loss=0.2615, pruned_loss=0.04588, over 2374542.64 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:12:24,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0382, 2.3377, 2.3516, 2.8068, 2.0706, 3.1304, 1.8763, 2.7186], device='cuda:0'), covar=tensor([0.1146, 0.0685, 0.1072, 0.0178, 0.0124, 0.0349, 0.1392, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0172, 0.0193, 0.0185, 0.0199, 0.0213, 0.0201, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:12:25,570 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213415.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:12:46,427 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=213430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:13:17,499 INFO [train.py:904] (0/8) Epoch 22, batch 300, loss[loss=0.1881, simple_loss=0.2634, pruned_loss=0.05636, over 16828.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2596, pruned_loss=0.04534, over 2577573.06 frames. ], batch size: 102, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:13:20,920 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3627, 2.3671, 2.3932, 4.1626, 2.3019, 2.6808, 2.4063, 2.5163], device='cuda:0'), covar=tensor([0.1360, 0.3401, 0.2922, 0.0639, 0.3939, 0.2520, 0.3595, 0.3111], device='cuda:0'), in_proj_covar=tensor([0.0394, 0.0442, 0.0364, 0.0324, 0.0434, 0.0504, 0.0413, 0.0517], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:13:52,065 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=213478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:13:54,570 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1521, 3.2847, 3.5052, 2.3408, 3.2222, 3.5696, 3.2990, 2.1518], device='cuda:0'), covar=tensor([0.0531, 0.0150, 0.0058, 0.0414, 0.0119, 0.0092, 0.0101, 0.0441], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0082, 0.0082, 0.0133, 0.0097, 0.0106, 0.0093, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 09:14:03,275 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 09:14:19,530 INFO [optim.py:368] (0/8) 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 09:14:25,256 INFO [train.py:904] (0/8) Epoch 22, batch 350, loss[loss=0.1537, simple_loss=0.2488, pruned_loss=0.0293, over 17210.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2573, pruned_loss=0.04353, over 2747025.83 frames. ], batch size: 46, lr: 3.10e-03, grad_scale: 1.0 2023-05-01 09:14:57,806 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9687, 5.4911, 5.5306, 5.3321, 5.3957, 5.9780, 5.4124, 5.1488], device='cuda:0'), covar=tensor([0.1119, 0.1942, 0.2279, 0.2265, 0.2905, 0.1101, 0.1782, 0.2475], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0587, 0.0651, 0.0490, 0.0648, 0.0683, 0.0511, 0.0653], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 09:15:18,059 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0908, 2.2663, 2.6833, 2.9726, 2.8854, 3.5525, 2.5637, 3.4693], device='cuda:0'), covar=tensor([0.0264, 0.0500, 0.0355, 0.0335, 0.0350, 0.0183, 0.0463, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0191, 0.0177, 0.0181, 0.0195, 0.0150, 0.0194, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:15:34,180 INFO [train.py:904] (0/8) Epoch 22, batch 400, loss[loss=0.1869, simple_loss=0.26, pruned_loss=0.05692, over 16638.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2557, pruned_loss=0.04296, over 2879539.72 frames. ], batch size: 89, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:15:52,117 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=213566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:15:56,527 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3436, 5.3409, 5.0654, 4.4448, 5.1125, 1.8240, 4.8233, 4.9124], device='cuda:0'), covar=tensor([0.0084, 0.0078, 0.0208, 0.0450, 0.0119, 0.2913, 0.0156, 0.0252], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0150, 0.0190, 0.0168, 0.0169, 0.0202, 0.0179, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:16:12,661 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 09:16:26,898 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-01 09:16:36,631 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.265e+02 2.651e+02 3.318e+02 1.993e+03, threshold=5.302e+02, percent-clipped=3.0 2023-05-01 09:16:43,117 INFO [train.py:904] (0/8) Epoch 22, batch 450, loss[loss=0.1665, simple_loss=0.2445, pruned_loss=0.04425, over 16732.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2544, pruned_loss=0.04207, over 2974164.35 frames. ], batch size: 124, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:17:16,900 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=213627.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:17:53,038 INFO [train.py:904] (0/8) Epoch 22, batch 500, loss[loss=0.1639, simple_loss=0.2526, pruned_loss=0.03758, over 16893.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2524, pruned_loss=0.04137, over 3050817.04 frames. ], batch size: 90, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:18:54,886 INFO [optim.py:368] (0/8) 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 09:19:01,585 INFO [train.py:904] (0/8) Epoch 22, batch 550, loss[loss=0.1491, simple_loss=0.2425, pruned_loss=0.02785, over 17184.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2519, pruned_loss=0.0411, over 3107490.32 frames. ], batch size: 46, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:19:23,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3522, 4.3620, 4.6752, 4.6864, 4.7382, 4.4459, 4.4458, 4.2555], device='cuda:0'), covar=tensor([0.0393, 0.1106, 0.0514, 0.0537, 0.0488, 0.0556, 0.0881, 0.0693], device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0448, 0.0434, 0.0404, 0.0482, 0.0456, 0.0537, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 09:19:30,783 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 09:20:10,673 INFO [train.py:904] (0/8) Epoch 22, batch 600, loss[loss=0.1491, simple_loss=0.2488, pruned_loss=0.02472, over 17115.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2519, pruned_loss=0.04133, over 3161475.82 frames. ], batch size: 47, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:21:13,522 INFO [optim.py:368] (0/8) 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 09:21:17,453 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 09:21:21,579 INFO [train.py:904] (0/8) Epoch 22, batch 650, loss[loss=0.148, simple_loss=0.2393, pruned_loss=0.02836, over 16840.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.25, pruned_loss=0.04046, over 3198407.50 frames. ], batch size: 42, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:21:23,130 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8242, 2.9074, 2.6264, 2.8398, 3.2489, 2.9762, 3.4282, 3.3659], device='cuda:0'), covar=tensor([0.0146, 0.0403, 0.0435, 0.0394, 0.0273, 0.0375, 0.0316, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0236, 0.0227, 0.0228, 0.0237, 0.0236, 0.0236, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:22:30,265 INFO [train.py:904] (0/8) Epoch 22, batch 700, loss[loss=0.1922, simple_loss=0.2657, pruned_loss=0.05938, over 16753.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2497, pruned_loss=0.04015, over 3227247.33 frames. ], batch size: 124, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:22:46,330 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-05-01 09:23:35,484 INFO [optim.py:368] (0/8) 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 09:23:41,805 INFO [train.py:904] (0/8) Epoch 22, batch 750, loss[loss=0.1421, simple_loss=0.2355, pruned_loss=0.0243, over 17187.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2507, pruned_loss=0.04037, over 3243106.83 frames. ], batch size: 46, lr: 3.10e-03, grad_scale: 2.0 2023-05-01 09:23:47,437 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 09:24:08,372 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=213922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:24:30,111 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-01 09:24:53,416 INFO [train.py:904] (0/8) Epoch 22, batch 800, loss[loss=0.1714, simple_loss=0.2463, pruned_loss=0.04827, over 16488.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.251, pruned_loss=0.0406, over 3242919.26 frames. ], batch size: 146, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:25:32,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1735, 3.0836, 3.3734, 2.1636, 3.0186, 2.3079, 3.5986, 3.4775], device='cuda:0'), covar=tensor([0.0239, 0.1022, 0.0620, 0.2008, 0.0851, 0.1070, 0.0535, 0.0902], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0161, 0.0166, 0.0153, 0.0144, 0.0129, 0.0143, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:25:56,896 INFO [optim.py:368] (0/8) 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 09:26:00,278 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-214000.pt 2023-05-01 09:26:06,618 INFO [train.py:904] (0/8) Epoch 22, batch 850, loss[loss=0.1668, simple_loss=0.266, pruned_loss=0.03386, over 17288.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2501, pruned_loss=0.04008, over 3264612.38 frames. ], batch size: 52, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:26:15,369 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8076, 3.9728, 2.7412, 4.6617, 3.2276, 4.5439, 2.8445, 3.3998], device='cuda:0'), covar=tensor([0.0337, 0.0445, 0.1521, 0.0197, 0.0807, 0.0576, 0.1375, 0.0740], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0178, 0.0197, 0.0164, 0.0179, 0.0219, 0.0204, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:27:15,833 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 09:27:17,256 INFO [train.py:904] (0/8) Epoch 22, batch 900, loss[loss=0.1581, simple_loss=0.2545, pruned_loss=0.03082, over 17034.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2489, pruned_loss=0.03898, over 3280651.14 frames. ], batch size: 50, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:28:19,802 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.353e+02 1.995e+02 2.382e+02 2.694e+02 6.707e+02, threshold=4.763e+02, percent-clipped=1.0 2023-05-01 09:28:27,537 INFO [train.py:904] (0/8) Epoch 22, batch 950, loss[loss=0.168, simple_loss=0.2565, pruned_loss=0.0397, over 16524.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2489, pruned_loss=0.03896, over 3294897.74 frames. ], batch size: 75, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:29:03,275 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214129.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:29:05,856 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.16 vs. limit=5.0 2023-05-01 09:29:35,595 INFO [train.py:904] (0/8) Epoch 22, batch 1000, loss[loss=0.1593, simple_loss=0.2311, pruned_loss=0.04372, over 16890.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.248, pruned_loss=0.03965, over 3290131.77 frames. ], batch size: 116, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:30:29,082 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214190.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:30:30,308 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9035, 2.0287, 2.4137, 2.7823, 2.7603, 2.8552, 2.0676, 3.0109], device='cuda:0'), covar=tensor([0.0194, 0.0483, 0.0362, 0.0286, 0.0316, 0.0254, 0.0527, 0.0186], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0194, 0.0179, 0.0183, 0.0198, 0.0154, 0.0196, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:30:39,709 INFO [optim.py:368] (0/8) 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 09:30:46,644 INFO [train.py:904] (0/8) Epoch 22, batch 1050, loss[loss=0.164, simple_loss=0.2429, pruned_loss=0.04255, over 16767.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2479, pruned_loss=0.03935, over 3288943.42 frames. ], batch size: 83, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:31:07,841 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=214218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:13,766 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:31:56,109 INFO [train.py:904] (0/8) Epoch 22, batch 1100, loss[loss=0.1542, simple_loss=0.2296, pruned_loss=0.03934, over 16866.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2478, pruned_loss=0.03943, over 3292733.89 frames. ], batch size: 90, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:32:19,650 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214270.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:31,256 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=214279.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:32:57,756 INFO [optim.py:368] (0/8) 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 09:33:03,850 INFO [train.py:904] (0/8) Epoch 22, batch 1150, loss[loss=0.1797, simple_loss=0.259, pruned_loss=0.0502, over 16594.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2471, pruned_loss=0.03878, over 3305442.22 frames. ], batch size: 62, lr: 3.10e-03, grad_scale: 4.0 2023-05-01 09:34:15,644 INFO [train.py:904] (0/8) Epoch 22, batch 1200, loss[loss=0.147, simple_loss=0.2284, pruned_loss=0.0328, over 16787.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2461, pruned_loss=0.03817, over 3318685.05 frames. ], batch size: 39, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:35:02,137 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-01 09:35:18,113 INFO [optim.py:368] (0/8) 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 09:35:25,093 INFO [train.py:904] (0/8) Epoch 22, batch 1250, loss[loss=0.2023, simple_loss=0.2678, pruned_loss=0.06839, over 16503.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2468, pruned_loss=0.03893, over 3319686.12 frames. ], batch size: 146, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:36:35,051 INFO [train.py:904] (0/8) Epoch 22, batch 1300, loss[loss=0.1772, simple_loss=0.254, pruned_loss=0.05022, over 16342.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2468, pruned_loss=0.03865, over 3309611.93 frames. ], batch size: 165, lr: 3.10e-03, grad_scale: 8.0 2023-05-01 09:36:49,638 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.59 vs. limit=2.0 2023-05-01 09:37:18,461 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214485.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:37:34,615 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.423e+02 2.092e+02 2.483e+02 2.893e+02 4.900e+02, threshold=4.965e+02, percent-clipped=1.0 2023-05-01 09:37:42,346 INFO [train.py:904] (0/8) Epoch 22, batch 1350, loss[loss=0.1654, simple_loss=0.2463, pruned_loss=0.04226, over 16847.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2471, pruned_loss=0.03862, over 3301371.75 frames. ], batch size: 96, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:37:51,389 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1863, 5.1997, 4.9225, 4.4099, 5.0045, 1.8482, 4.7330, 4.7590], device='cuda:0'), covar=tensor([0.0102, 0.0080, 0.0231, 0.0427, 0.0114, 0.3104, 0.0165, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0156, 0.0199, 0.0176, 0.0177, 0.0208, 0.0188, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:38:42,615 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-05-01 09:38:48,216 INFO [train.py:904] (0/8) Epoch 22, batch 1400, loss[loss=0.1765, simple_loss=0.2487, pruned_loss=0.05219, over 16841.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2475, pruned_loss=0.03868, over 3305040.68 frames. ], batch size: 109, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:38:52,233 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 09:39:17,562 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=214574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:39:19,991 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-01 09:39:50,089 INFO [optim.py:368] (0/8) 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 09:39:57,059 INFO [train.py:904] (0/8) Epoch 22, batch 1450, loss[loss=0.1578, simple_loss=0.2515, pruned_loss=0.03205, over 17106.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2474, pruned_loss=0.03812, over 3306458.26 frames. ], batch size: 47, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:41:07,201 INFO [train.py:904] (0/8) Epoch 22, batch 1500, loss[loss=0.2005, simple_loss=0.2764, pruned_loss=0.0623, over 17027.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2465, pruned_loss=0.03817, over 3296354.46 frames. ], batch size: 55, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:42:10,193 INFO [optim.py:368] (0/8) 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 09:42:16,365 INFO [train.py:904] (0/8) Epoch 22, batch 1550, loss[loss=0.165, simple_loss=0.258, pruned_loss=0.03602, over 17044.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2478, pruned_loss=0.03893, over 3310306.08 frames. ], batch size: 50, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:42:41,763 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3863, 4.3963, 4.7339, 4.7266, 4.7715, 4.4569, 4.4718, 4.3114], device='cuda:0'), covar=tensor([0.0378, 0.0659, 0.0482, 0.0448, 0.0572, 0.0470, 0.0891, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0460, 0.0447, 0.0415, 0.0493, 0.0470, 0.0552, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 09:42:50,996 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8027, 4.7973, 5.1735, 5.1719, 5.2234, 4.8819, 4.8124, 4.6575], device='cuda:0'), covar=tensor([0.0368, 0.0611, 0.0437, 0.0474, 0.0586, 0.0431, 0.1023, 0.0541], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0460, 0.0448, 0.0415, 0.0494, 0.0470, 0.0552, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 09:42:52,058 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5614, 4.6897, 4.8122, 4.6405, 4.6150, 5.2516, 4.7282, 4.4620], device='cuda:0'), covar=tensor([0.1523, 0.1913, 0.2330, 0.2113, 0.3068, 0.1115, 0.1695, 0.2641], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0603, 0.0663, 0.0502, 0.0667, 0.0696, 0.0524, 0.0667], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 09:43:28,006 INFO [train.py:904] (0/8) Epoch 22, batch 1600, loss[loss=0.18, simple_loss=0.2815, pruned_loss=0.03922, over 17168.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2492, pruned_loss=0.03907, over 3315478.01 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:44:03,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6588, 3.8559, 4.1034, 2.8373, 3.6369, 4.1268, 3.7646, 2.4596], device='cuda:0'), covar=tensor([0.0515, 0.0341, 0.0053, 0.0397, 0.0120, 0.0091, 0.0099, 0.0465], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0134, 0.0098, 0.0109, 0.0095, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 09:44:12,575 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214785.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:44:32,065 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.192e+02 2.616e+02 3.002e+02 5.148e+02, threshold=5.232e+02, percent-clipped=2.0 2023-05-01 09:44:37,870 INFO [train.py:904] (0/8) Epoch 22, batch 1650, loss[loss=0.1419, simple_loss=0.2299, pruned_loss=0.02699, over 17203.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2497, pruned_loss=0.03972, over 3315392.17 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:45:04,220 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 09:45:07,978 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1739, 5.7804, 5.9049, 5.6504, 5.6969, 6.2759, 5.8323, 5.4847], device='cuda:0'), covar=tensor([0.0943, 0.1851, 0.2581, 0.2160, 0.2901, 0.1091, 0.1573, 0.2512], device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0610, 0.0671, 0.0507, 0.0674, 0.0705, 0.0530, 0.0673], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 09:45:13,483 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7834, 2.7549, 2.6016, 4.3623, 3.5715, 4.1340, 1.6254, 3.0018], device='cuda:0'), covar=tensor([0.1396, 0.0734, 0.1144, 0.0196, 0.0201, 0.0411, 0.1617, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0175, 0.0195, 0.0192, 0.0204, 0.0218, 0.0203, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:45:20,148 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:45:20,251 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1923, 4.1736, 4.4969, 4.4825, 4.5257, 4.2245, 4.2434, 4.1394], device='cuda:0'), covar=tensor([0.0374, 0.0685, 0.0444, 0.0475, 0.0510, 0.0488, 0.0842, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0465, 0.0451, 0.0419, 0.0499, 0.0474, 0.0557, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 09:45:25,449 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 09:45:47,643 INFO [train.py:904] (0/8) Epoch 22, batch 1700, loss[loss=0.1506, simple_loss=0.2443, pruned_loss=0.02845, over 17174.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2519, pruned_loss=0.04077, over 3309889.30 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:46:17,911 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=214874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:46:53,106 INFO [optim.py:368] (0/8) 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 09:46:58,561 INFO [train.py:904] (0/8) Epoch 22, batch 1750, loss[loss=0.1609, simple_loss=0.2526, pruned_loss=0.03459, over 17176.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2533, pruned_loss=0.04103, over 3312648.01 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:47:25,708 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=214922.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:48:06,732 INFO [train.py:904] (0/8) Epoch 22, batch 1800, loss[loss=0.1577, simple_loss=0.2494, pruned_loss=0.03295, over 17168.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2538, pruned_loss=0.04055, over 3321725.45 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:48:20,439 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7452, 4.5284, 4.7851, 4.9586, 5.1369, 4.5472, 5.0959, 5.1257], device='cuda:0'), covar=tensor([0.1897, 0.1395, 0.1751, 0.0824, 0.0615, 0.1120, 0.0746, 0.0691], device='cuda:0'), in_proj_covar=tensor([0.0655, 0.0806, 0.0943, 0.0817, 0.0616, 0.0648, 0.0669, 0.0780], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:49:13,318 INFO [optim.py:368] (0/8) 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 09:49:18,212 INFO [train.py:904] (0/8) Epoch 22, batch 1850, loss[loss=0.1867, simple_loss=0.2633, pruned_loss=0.05507, over 16709.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2546, pruned_loss=0.04082, over 3324281.18 frames. ], batch size: 134, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:49:30,319 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8090, 4.6581, 4.7001, 4.3703, 4.4340, 4.7475, 4.5732, 4.4724], device='cuda:0'), covar=tensor([0.0662, 0.1067, 0.0334, 0.0302, 0.0890, 0.0544, 0.0486, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0439, 0.0353, 0.0351, 0.0362, 0.0407, 0.0242, 0.0422], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:50:27,231 INFO [train.py:904] (0/8) Epoch 22, batch 1900, loss[loss=0.1567, simple_loss=0.2596, pruned_loss=0.02686, over 17064.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2548, pruned_loss=0.0406, over 3328578.73 frames. ], batch size: 50, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:50:47,105 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 09:51:31,800 INFO [optim.py:368] (0/8) 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 09:51:36,133 INFO [train.py:904] (0/8) Epoch 22, batch 1950, loss[loss=0.1565, simple_loss=0.2549, pruned_loss=0.02907, over 17114.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2556, pruned_loss=0.04073, over 3322459.30 frames. ], batch size: 47, lr: 3.09e-03, grad_scale: 4.0 2023-05-01 09:51:41,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9419, 3.3616, 3.0226, 5.2513, 4.4249, 4.6812, 1.6845, 3.5680], device='cuda:0'), covar=tensor([0.1297, 0.0667, 0.1073, 0.0168, 0.0198, 0.0331, 0.1590, 0.0611], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0173, 0.0194, 0.0190, 0.0202, 0.0216, 0.0201, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:52:03,019 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7657, 2.7569, 2.3218, 2.5511, 3.0237, 2.8173, 3.3716, 3.2554], device='cuda:0'), covar=tensor([0.0183, 0.0453, 0.0598, 0.0495, 0.0331, 0.0449, 0.0283, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0217, 0.0240, 0.0230, 0.0231, 0.0241, 0.0240, 0.0242, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:52:21,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5893, 3.8653, 4.0383, 3.9939, 4.0376, 3.8232, 3.5986, 3.8486], device='cuda:0'), covar=tensor([0.0630, 0.0763, 0.0587, 0.0675, 0.0708, 0.0726, 0.1301, 0.0658], device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0462, 0.0448, 0.0417, 0.0494, 0.0472, 0.0555, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 09:52:44,784 INFO [train.py:904] (0/8) Epoch 22, batch 2000, loss[loss=0.1539, simple_loss=0.2445, pruned_loss=0.03167, over 17206.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2558, pruned_loss=0.04058, over 3326179.61 frames. ], batch size: 46, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:52:59,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9110, 2.9948, 2.7039, 2.8480, 3.2456, 3.0566, 3.5946, 3.4632], device='cuda:0'), covar=tensor([0.0156, 0.0387, 0.0438, 0.0400, 0.0268, 0.0346, 0.0248, 0.0258], device='cuda:0'), in_proj_covar=tensor([0.0217, 0.0240, 0.0231, 0.0231, 0.0241, 0.0240, 0.0243, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:53:19,698 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7550, 3.5566, 3.9091, 2.0608, 3.9804, 4.0618, 3.1933, 2.9618], device='cuda:0'), covar=tensor([0.0708, 0.0210, 0.0164, 0.1084, 0.0094, 0.0169, 0.0367, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0140, 0.0081, 0.0127, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:53:47,110 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 09:53:48,736 INFO [optim.py:368] (0/8) 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 09:53:53,927 INFO [train.py:904] (0/8) Epoch 22, batch 2050, loss[loss=0.1739, simple_loss=0.2509, pruned_loss=0.04849, over 16838.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2551, pruned_loss=0.04091, over 3326963.14 frames. ], batch size: 83, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:53:56,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7881, 4.0436, 2.8042, 4.5615, 3.2084, 4.5906, 2.8729, 3.4184], device='cuda:0'), covar=tensor([0.0348, 0.0431, 0.1395, 0.0405, 0.0815, 0.0503, 0.1382, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0181, 0.0198, 0.0168, 0.0180, 0.0223, 0.0206, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:54:31,446 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215230.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:55:01,198 INFO [train.py:904] (0/8) Epoch 22, batch 2100, loss[loss=0.1859, simple_loss=0.268, pruned_loss=0.0519, over 16675.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2554, pruned_loss=0.04105, over 3328409.09 frames. ], batch size: 89, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:55:54,411 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215291.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 09:56:04,632 INFO [optim.py:368] (0/8) 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 09:56:09,030 INFO [train.py:904] (0/8) Epoch 22, batch 2150, loss[loss=0.1778, simple_loss=0.2701, pruned_loss=0.04278, over 16270.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2567, pruned_loss=0.04137, over 3323313.37 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:56:39,609 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215325.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 09:56:51,080 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 09:57:18,610 INFO [train.py:904] (0/8) Epoch 22, batch 2200, loss[loss=0.1845, simple_loss=0.2633, pruned_loss=0.05285, over 16343.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2573, pruned_loss=0.04211, over 3318028.19 frames. ], batch size: 165, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:57:29,759 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8398, 3.1971, 2.6964, 5.0155, 4.0120, 4.4693, 1.8029, 3.2156], device='cuda:0'), covar=tensor([0.1426, 0.0739, 0.1278, 0.0185, 0.0243, 0.0428, 0.1690, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0174, 0.0195, 0.0191, 0.0205, 0.0218, 0.0203, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:58:02,271 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215386.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 09:58:14,366 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215394.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 09:58:20,118 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0829, 3.0646, 1.9116, 3.2542, 2.4327, 3.3044, 2.1255, 2.6205], device='cuda:0'), covar=tensor([0.0316, 0.0451, 0.1660, 0.0349, 0.0844, 0.0699, 0.1455, 0.0755], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0181, 0.0198, 0.0168, 0.0180, 0.0223, 0.0206, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:58:20,774 INFO [optim.py:368] (0/8) 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 09:58:24,661 INFO [train.py:904] (0/8) Epoch 22, batch 2250, loss[loss=0.1576, simple_loss=0.2548, pruned_loss=0.03019, over 17142.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2578, pruned_loss=0.04252, over 3320438.77 frames. ], batch size: 48, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:58:25,154 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5507, 5.5365, 5.2993, 4.8288, 5.4056, 2.4574, 5.1568, 5.1924], device='cuda:0'), covar=tensor([0.0078, 0.0071, 0.0180, 0.0316, 0.0091, 0.2309, 0.0109, 0.0164], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0157, 0.0200, 0.0179, 0.0179, 0.0209, 0.0189, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 09:58:27,092 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2427, 3.8928, 4.4644, 2.2908, 4.7172, 4.8354, 3.4852, 3.6602], device='cuda:0'), covar=tensor([0.0656, 0.0269, 0.0216, 0.1177, 0.0067, 0.0148, 0.0398, 0.0392], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0109, 0.0098, 0.0140, 0.0081, 0.0128, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 09:59:36,621 INFO [train.py:904] (0/8) Epoch 22, batch 2300, loss[loss=0.1488, simple_loss=0.2362, pruned_loss=0.03071, over 16994.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2578, pruned_loss=0.04241, over 3308887.08 frames. ], batch size: 41, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 09:59:59,688 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6021, 4.6363, 4.8444, 4.5975, 4.6960, 5.3110, 4.8580, 4.4745], device='cuda:0'), covar=tensor([0.1676, 0.2194, 0.2651, 0.2378, 0.2903, 0.1227, 0.1730, 0.2883], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0619, 0.0680, 0.0515, 0.0684, 0.0714, 0.0535, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:00:17,443 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6930, 2.8669, 2.5228, 4.9163, 3.9217, 4.4487, 1.5711, 3.2371], device='cuda:0'), covar=tensor([0.1466, 0.0813, 0.1328, 0.0197, 0.0215, 0.0392, 0.1774, 0.0786], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0173, 0.0193, 0.0191, 0.0204, 0.0217, 0.0201, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:00:35,868 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 10:00:42,879 INFO [optim.py:368] (0/8) 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 10:00:46,507 INFO [train.py:904] (0/8) Epoch 22, batch 2350, loss[loss=0.1858, simple_loss=0.2638, pruned_loss=0.05396, over 16806.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2583, pruned_loss=0.04261, over 3310445.44 frames. ], batch size: 134, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:00:46,879 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215503.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:01:30,561 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4460, 4.3729, 4.3312, 4.0867, 4.1358, 4.4195, 4.1196, 4.1683], device='cuda:0'), covar=tensor([0.0649, 0.0832, 0.0291, 0.0276, 0.0715, 0.0469, 0.0685, 0.0639], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0442, 0.0355, 0.0355, 0.0365, 0.0410, 0.0244, 0.0426], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:01:48,901 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:01:55,012 INFO [train.py:904] (0/8) Epoch 22, batch 2400, loss[loss=0.1641, simple_loss=0.267, pruned_loss=0.03064, over 17138.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2599, pruned_loss=0.04287, over 3305045.19 frames. ], batch size: 49, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:02:00,688 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-01 10:02:11,603 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215564.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:12,760 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215565.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:02:41,726 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215586.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 10:02:59,711 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.274e+02 2.111e+02 2.344e+02 3.035e+02 7.498e+02, threshold=4.688e+02, percent-clipped=4.0 2023-05-01 10:03:04,616 INFO [train.py:904] (0/8) Epoch 22, batch 2450, loss[loss=0.1486, simple_loss=0.2526, pruned_loss=0.02225, over 17046.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2607, pruned_loss=0.04324, over 3302971.33 frames. ], batch size: 50, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:03:08,839 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-01 10:03:11,977 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215609.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 10:03:35,729 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 22, batch 2500, loss[loss=0.1624, simple_loss=0.2615, pruned_loss=0.03166, over 17102.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.26, pruned_loss=0.04248, over 3304837.29 frames. ], batch size: 47, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:04:51,433 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215681.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:05:01,824 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-05-01 10:05:02,675 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215689.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:05:12,948 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2462, 5.2173, 5.0819, 4.5769, 4.7223, 5.1494, 5.1393, 4.7301], device='cuda:0'), covar=tensor([0.0653, 0.0612, 0.0346, 0.0386, 0.1178, 0.0558, 0.0301, 0.0848], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0444, 0.0356, 0.0356, 0.0367, 0.0411, 0.0244, 0.0428], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:05:15,512 INFO [optim.py:368] (0/8) 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 10:05:20,586 INFO [train.py:904] (0/8) Epoch 22, batch 2550, loss[loss=0.1544, simple_loss=0.2385, pruned_loss=0.03512, over 16730.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2598, pruned_loss=0.04226, over 3304968.27 frames. ], batch size: 39, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:06:30,403 INFO [train.py:904] (0/8) Epoch 22, batch 2600, loss[loss=0.1563, simple_loss=0.2404, pruned_loss=0.03608, over 16781.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2591, pruned_loss=0.04166, over 3314049.58 frames. ], batch size: 102, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:06:35,451 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7750, 4.8020, 5.1902, 5.1410, 5.2144, 4.8918, 4.8289, 4.6681], device='cuda:0'), covar=tensor([0.0334, 0.0611, 0.0353, 0.0418, 0.0527, 0.0399, 0.0993, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0459, 0.0447, 0.0414, 0.0493, 0.0470, 0.0552, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 10:07:36,124 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.223e+02 2.543e+02 2.941e+02 8.287e+02, threshold=5.085e+02, percent-clipped=5.0 2023-05-01 10:07:40,013 INFO [train.py:904] (0/8) Epoch 22, batch 2650, loss[loss=0.1851, simple_loss=0.2653, pruned_loss=0.0525, over 16940.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2593, pruned_loss=0.04119, over 3318623.81 frames. ], batch size: 116, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:07:42,030 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-01 10:08:06,442 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-01 10:08:15,788 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=215829.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:08:48,050 INFO [train.py:904] (0/8) Epoch 22, batch 2700, loss[loss=0.182, simple_loss=0.2723, pruned_loss=0.04581, over 16901.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2592, pruned_loss=0.04076, over 3317719.79 frames. ], batch size: 96, lr: 3.09e-03, grad_scale: 8.0 2023-05-01 10:08:55,052 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4614, 5.4256, 5.2894, 4.7818, 4.9553, 5.3769, 5.3471, 4.9566], device='cuda:0'), covar=tensor([0.0593, 0.0493, 0.0299, 0.0341, 0.1071, 0.0423, 0.0279, 0.0759], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0447, 0.0358, 0.0358, 0.0369, 0.0413, 0.0245, 0.0431], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:08:57,510 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:01,081 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-01 10:09:34,760 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215886.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:09:39,291 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9622, 3.7749, 4.2978, 2.1013, 4.4825, 4.5823, 3.2697, 3.4520], device='cuda:0'), covar=tensor([0.0722, 0.0240, 0.0214, 0.1188, 0.0090, 0.0184, 0.0421, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0139, 0.0080, 0.0128, 0.0130, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:09:39,319 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=215890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:09:53,925 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 22, batch 2750, loss[loss=0.1716, simple_loss=0.2625, pruned_loss=0.04032, over 17122.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2592, pruned_loss=0.04084, over 3315376.36 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:09:59,453 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215904.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 10:10:21,977 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=215921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:10:33,150 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2790, 5.2577, 5.1340, 4.6588, 4.7721, 5.2004, 5.0979, 4.7750], device='cuda:0'), covar=tensor([0.0589, 0.0513, 0.0318, 0.0356, 0.1078, 0.0461, 0.0336, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0446, 0.0356, 0.0356, 0.0368, 0.0412, 0.0245, 0.0430], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:10:38,796 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=215934.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:11:05,018 INFO [train.py:904] (0/8) Epoch 22, batch 2800, loss[loss=0.1381, simple_loss=0.2245, pruned_loss=0.02588, over 16816.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2594, pruned_loss=0.04083, over 3313275.01 frames. ], batch size: 39, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:11:11,242 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 10:11:42,305 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215981.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:11:53,934 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=215989.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:12:07,422 INFO [optim.py:368] (0/8) 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 10:12:07,798 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-216000.pt 2023-05-01 10:12:14,956 INFO [train.py:904] (0/8) Epoch 22, batch 2850, loss[loss=0.1405, simple_loss=0.2249, pruned_loss=0.02801, over 17043.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2588, pruned_loss=0.04095, over 3311675.56 frames. ], batch size: 41, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:12:17,722 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5826, 2.4099, 2.4397, 4.4573, 2.3896, 2.7941, 2.5279, 2.5942], device='cuda:0'), covar=tensor([0.1258, 0.3744, 0.3084, 0.0515, 0.4088, 0.2680, 0.3526, 0.3770], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0452, 0.0372, 0.0332, 0.0440, 0.0521, 0.0424, 0.0531], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:12:51,469 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216028.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:12:52,495 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216029.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:13:03,976 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:13:26,231 INFO [train.py:904] (0/8) Epoch 22, batch 2900, loss[loss=0.1652, simple_loss=0.2499, pruned_loss=0.04025, over 15876.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2586, pruned_loss=0.04229, over 3296093.43 frames. ], batch size: 35, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:13:32,103 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.01 vs. limit=5.0 2023-05-01 10:14:16,947 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216089.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:14:24,301 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9170, 4.7742, 4.9708, 5.1955, 5.3819, 4.7377, 5.3420, 5.3751], device='cuda:0'), covar=tensor([0.1934, 0.1355, 0.1874, 0.0761, 0.0634, 0.0996, 0.0687, 0.0616], device='cuda:0'), in_proj_covar=tensor([0.0666, 0.0827, 0.0963, 0.0834, 0.0631, 0.0665, 0.0683, 0.0793], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:14:31,573 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 22, batch 2950, loss[loss=0.1675, simple_loss=0.26, pruned_loss=0.03748, over 17116.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2576, pruned_loss=0.04193, over 3311943.49 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:14:54,356 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9377, 2.7971, 2.7239, 5.0229, 4.0595, 4.3487, 1.5861, 3.2748], device='cuda:0'), covar=tensor([0.1289, 0.0811, 0.1194, 0.0155, 0.0224, 0.0435, 0.1650, 0.0779], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0192, 0.0204, 0.0217, 0.0202, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:15:45,388 INFO [train.py:904] (0/8) Epoch 22, batch 3000, loss[loss=0.1764, simple_loss=0.2664, pruned_loss=0.04323, over 15673.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2571, pruned_loss=0.04161, over 3326225.00 frames. ], batch size: 191, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:15:45,389 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 10:15:54,105 INFO [train.py:938] (0/8) Epoch 22, validation: loss=0.1347, simple_loss=0.2399, pruned_loss=0.0148, over 944034.00 frames. 2023-05-01 10:15:54,105 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 10:16:02,810 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:38,735 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:16:59,839 INFO [optim.py:368] (0/8) 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 10:17:03,742 INFO [train.py:904] (0/8) Epoch 22, batch 3050, loss[loss=0.1477, simple_loss=0.2362, pruned_loss=0.02966, over 17206.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2565, pruned_loss=0.04124, over 3323734.09 frames. ], batch size: 44, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:17:05,188 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216204.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:17:09,671 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216207.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:25,836 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 10:17:29,253 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216221.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:17:40,193 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 10:17:56,723 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0051, 4.4574, 4.4767, 3.2856, 3.7855, 4.4283, 4.0172, 2.8340], device='cuda:0'), covar=tensor([0.0467, 0.0058, 0.0043, 0.0336, 0.0137, 0.0094, 0.0085, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0084, 0.0084, 0.0133, 0.0099, 0.0109, 0.0095, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:18:10,766 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216252.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:18:12,229 INFO [train.py:904] (0/8) Epoch 22, batch 3100, loss[loss=0.1615, simple_loss=0.2387, pruned_loss=0.04215, over 16899.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.256, pruned_loss=0.04164, over 3318757.38 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:18:33,893 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216269.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:19:16,928 INFO [optim.py:368] (0/8) 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 10:19:21,088 INFO [train.py:904] (0/8) Epoch 22, batch 3150, loss[loss=0.1695, simple_loss=0.267, pruned_loss=0.03604, over 17094.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2563, pruned_loss=0.04206, over 3327260.40 frames. ], batch size: 55, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:19:43,632 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2315, 4.2376, 4.5634, 4.5449, 4.5891, 4.3067, 4.3355, 4.2618], device='cuda:0'), covar=tensor([0.0386, 0.0605, 0.0368, 0.0411, 0.0533, 0.0458, 0.0829, 0.0531], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0471, 0.0457, 0.0424, 0.0505, 0.0482, 0.0566, 0.0384], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 10:20:08,016 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 10:20:17,593 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 10:20:29,459 INFO [train.py:904] (0/8) Epoch 22, batch 3200, loss[loss=0.1527, simple_loss=0.2504, pruned_loss=0.02749, over 17060.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2546, pruned_loss=0.04097, over 3329722.80 frames. ], batch size: 50, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:20:33,572 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 10:21:13,885 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216384.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:21:24,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8962, 2.4760, 1.7504, 2.0775, 2.8515, 2.6418, 3.0357, 2.9566], device='cuda:0'), covar=tensor([0.0279, 0.0491, 0.0793, 0.0652, 0.0314, 0.0417, 0.0276, 0.0347], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0240, 0.0230, 0.0231, 0.0241, 0.0241, 0.0245, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:21:36,412 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.593e+02 2.182e+02 2.419e+02 2.918e+02 5.281e+02, threshold=4.837e+02, percent-clipped=1.0 2023-05-01 10:21:40,330 INFO [train.py:904] (0/8) Epoch 22, batch 3250, loss[loss=0.1754, simple_loss=0.2638, pruned_loss=0.04354, over 16710.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2551, pruned_loss=0.04139, over 3336571.05 frames. ], batch size: 57, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:21:47,438 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 10:22:34,198 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.25 vs. limit=5.0 2023-05-01 10:22:52,229 INFO [train.py:904] (0/8) Epoch 22, batch 3300, loss[loss=0.2092, simple_loss=0.2857, pruned_loss=0.06634, over 16467.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2562, pruned_loss=0.04182, over 3333617.16 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:23:26,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3164, 2.3204, 2.4448, 4.1107, 2.3173, 2.7635, 2.3380, 2.5834], device='cuda:0'), covar=tensor([0.1515, 0.3816, 0.2845, 0.0604, 0.3877, 0.2558, 0.4036, 0.2790], device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0452, 0.0373, 0.0333, 0.0441, 0.0522, 0.0425, 0.0531], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:23:35,899 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 10:24:00,076 INFO [train.py:904] (0/8) Epoch 22, batch 3350, loss[loss=0.1748, simple_loss=0.2544, pruned_loss=0.04757, over 16483.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2568, pruned_loss=0.04196, over 3326839.42 frames. ], batch size: 75, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:24:05,809 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8556, 4.9868, 5.1850, 4.9332, 4.9333, 5.6170, 5.0528, 4.7904], device='cuda:0'), covar=tensor([0.1402, 0.1987, 0.2392, 0.2416, 0.3166, 0.1149, 0.1883, 0.2780], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0616, 0.0675, 0.0512, 0.0682, 0.0711, 0.0532, 0.0681], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:24:36,517 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4462, 3.6129, 3.8783, 2.6922, 3.5382, 3.9166, 3.5628, 2.2521], device='cuda:0'), covar=tensor([0.0503, 0.0237, 0.0057, 0.0364, 0.0111, 0.0097, 0.0108, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0133, 0.0099, 0.0110, 0.0095, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 10:24:42,590 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:25:10,678 INFO [train.py:904] (0/8) Epoch 22, batch 3400, loss[loss=0.1504, simple_loss=0.2426, pruned_loss=0.02909, over 17115.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2565, pruned_loss=0.0419, over 3315087.20 frames. ], batch size: 47, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:25:53,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8259, 2.7471, 2.4251, 2.7489, 3.0552, 2.8824, 3.4206, 3.3344], device='cuda:0'), covar=tensor([0.0143, 0.0449, 0.0537, 0.0449, 0.0299, 0.0384, 0.0247, 0.0272], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0241, 0.0231, 0.0232, 0.0242, 0.0242, 0.0246, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:26:02,854 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:26:07,075 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9985, 2.1099, 2.6319, 2.8913, 2.8060, 3.5201, 2.5181, 3.3963], device='cuda:0'), covar=tensor([0.0261, 0.0536, 0.0357, 0.0361, 0.0364, 0.0176, 0.0454, 0.0195], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0198, 0.0184, 0.0189, 0.0201, 0.0158, 0.0200, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:26:15,744 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.115e+02 2.508e+02 2.941e+02 4.033e+02, threshold=5.017e+02, percent-clipped=0.0 2023-05-01 10:26:19,665 INFO [train.py:904] (0/8) Epoch 22, batch 3450, loss[loss=0.1527, simple_loss=0.2502, pruned_loss=0.02761, over 17119.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2557, pruned_loss=0.04161, over 3314673.45 frames. ], batch size: 49, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:26:33,539 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1793, 2.1826, 2.3069, 3.7506, 2.1935, 2.5080, 2.2900, 2.3490], device='cuda:0'), covar=tensor([0.1399, 0.3516, 0.2810, 0.0652, 0.3940, 0.2500, 0.3705, 0.3362], device='cuda:0'), in_proj_covar=tensor([0.0408, 0.0453, 0.0374, 0.0334, 0.0442, 0.0523, 0.0426, 0.0533], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:26:55,835 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 10:27:26,415 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4367, 2.7479, 3.0168, 2.0141, 2.7147, 2.0759, 3.1582, 3.1140], device='cuda:0'), covar=tensor([0.0264, 0.0979, 0.0601, 0.1983, 0.0898, 0.1030, 0.0563, 0.0845], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0166, 0.0167, 0.0154, 0.0146, 0.0131, 0.0145, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:27:29,262 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216652.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:27:29,992 INFO [train.py:904] (0/8) Epoch 22, batch 3500, loss[loss=0.135, simple_loss=0.2164, pruned_loss=0.02681, over 16759.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2541, pruned_loss=0.04129, over 3313437.26 frames. ], batch size: 39, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:27:32,837 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216655.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 10:27:33,173 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 10:28:11,460 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8742, 2.5746, 2.5835, 1.9387, 2.5926, 2.7016, 2.5405, 1.9170], device='cuda:0'), covar=tensor([0.0458, 0.0119, 0.0099, 0.0393, 0.0148, 0.0142, 0.0148, 0.0412], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0084, 0.0084, 0.0133, 0.0098, 0.0109, 0.0095, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 10:28:12,635 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=216684.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:35,767 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.548e+02 1.992e+02 2.337e+02 2.852e+02 6.760e+02, threshold=4.673e+02, percent-clipped=2.0 2023-05-01 10:28:39,283 INFO [train.py:904] (0/8) Epoch 22, batch 3550, loss[loss=0.1702, simple_loss=0.2446, pruned_loss=0.04787, over 16894.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2527, pruned_loss=0.04062, over 3309016.26 frames. ], batch size: 116, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:28:42,591 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216705.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:28:56,716 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216716.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:29:17,553 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8717, 5.0316, 5.2032, 4.9952, 5.0009, 5.6305, 5.1896, 4.8270], device='cuda:0'), covar=tensor([0.1349, 0.2084, 0.2126, 0.2147, 0.2681, 0.1007, 0.1450, 0.2443], device='cuda:0'), in_proj_covar=tensor([0.0424, 0.0615, 0.0675, 0.0511, 0.0682, 0.0710, 0.0530, 0.0681], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:29:20,501 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=216732.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:29:49,497 INFO [train.py:904] (0/8) Epoch 22, batch 3600, loss[loss=0.1471, simple_loss=0.2305, pruned_loss=0.03188, over 16781.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2513, pruned_loss=0.04014, over 3315083.05 frames. ], batch size: 83, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:30:08,456 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=216766.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:30:24,183 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7603, 2.6874, 2.2607, 2.6267, 3.0624, 2.8852, 3.3647, 3.2997], device='cuda:0'), covar=tensor([0.0138, 0.0429, 0.0549, 0.0447, 0.0275, 0.0365, 0.0246, 0.0261], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0242, 0.0232, 0.0232, 0.0242, 0.0242, 0.0246, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:31:00,617 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 22, batch 3650, loss[loss=0.1416, simple_loss=0.2271, pruned_loss=0.02804, over 16803.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2505, pruned_loss=0.04021, over 3309841.67 frames. ], batch size: 39, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:31:32,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1679, 2.1955, 2.2642, 3.8633, 2.2330, 2.4983, 2.2375, 2.3805], device='cuda:0'), covar=tensor([0.1487, 0.3924, 0.3050, 0.0628, 0.3891, 0.2813, 0.4102, 0.3120], device='cuda:0'), in_proj_covar=tensor([0.0408, 0.0453, 0.0374, 0.0335, 0.0442, 0.0524, 0.0426, 0.0533], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:32:08,767 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6972, 2.7411, 2.8699, 2.0978, 2.6605, 2.1180, 2.8072, 2.8664], device='cuda:0'), covar=tensor([0.0245, 0.0816, 0.0508, 0.1956, 0.0802, 0.0945, 0.0531, 0.0744], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0166, 0.0167, 0.0155, 0.0146, 0.0131, 0.0145, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:32:18,435 INFO [train.py:904] (0/8) Epoch 22, batch 3700, loss[loss=0.1871, simple_loss=0.2761, pruned_loss=0.04909, over 16693.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2496, pruned_loss=0.04187, over 3280215.79 frames. ], batch size: 57, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:32:31,647 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 10:33:19,768 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7609, 3.5505, 3.9516, 2.0879, 4.0565, 4.0720, 3.3038, 2.9682], device='cuda:0'), covar=tensor([0.0793, 0.0281, 0.0193, 0.1252, 0.0113, 0.0193, 0.0369, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0110, 0.0100, 0.0140, 0.0082, 0.0129, 0.0131, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:33:31,450 INFO [optim.py:368] (0/8) 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 10:33:32,651 INFO [train.py:904] (0/8) Epoch 22, batch 3750, loss[loss=0.1663, simple_loss=0.2455, pruned_loss=0.04356, over 16211.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2507, pruned_loss=0.0434, over 3281598.94 frames. ], batch size: 165, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:33:35,180 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 10:34:36,649 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=216947.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:34:44,490 INFO [train.py:904] (0/8) Epoch 22, batch 3800, loss[loss=0.1496, simple_loss=0.2301, pruned_loss=0.03457, over 16508.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2517, pruned_loss=0.04477, over 3278619.63 frames. ], batch size: 75, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:35:03,012 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=216965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:35:55,034 INFO [optim.py:368] (0/8) 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 10:35:56,830 INFO [train.py:904] (0/8) Epoch 22, batch 3850, loss[loss=0.1788, simple_loss=0.2618, pruned_loss=0.04789, over 16389.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2522, pruned_loss=0.04566, over 3279621.00 frames. ], batch size: 35, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:36:08,366 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217011.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:36:10,571 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0999, 2.1366, 2.2671, 3.7392, 2.1942, 2.4338, 2.2513, 2.3321], device='cuda:0'), covar=tensor([0.1561, 0.3849, 0.3032, 0.0649, 0.3982, 0.2711, 0.4175, 0.3149], device='cuda:0'), in_proj_covar=tensor([0.0408, 0.0456, 0.0374, 0.0334, 0.0442, 0.0525, 0.0427, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:36:29,838 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:37:09,496 INFO [train.py:904] (0/8) Epoch 22, batch 3900, loss[loss=0.166, simple_loss=0.2439, pruned_loss=0.04408, over 16886.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2521, pruned_loss=0.0461, over 3293661.72 frames. ], batch size: 96, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:37:22,076 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217061.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:38:21,560 INFO [optim.py:368] (0/8) 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 10:38:22,857 INFO [train.py:904] (0/8) Epoch 22, batch 3950, loss[loss=0.1752, simple_loss=0.251, pruned_loss=0.04968, over 16343.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2516, pruned_loss=0.04654, over 3286487.98 frames. ], batch size: 68, lr: 3.08e-03, grad_scale: 4.0 2023-05-01 10:39:35,777 INFO [train.py:904] (0/8) Epoch 22, batch 4000, loss[loss=0.1853, simple_loss=0.2728, pruned_loss=0.04889, over 16378.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2524, pruned_loss=0.04693, over 3284284.94 frames. ], batch size: 146, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:40:00,567 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1597, 5.7742, 6.0323, 5.5863, 5.8330, 6.2560, 5.8387, 5.5186], device='cuda:0'), covar=tensor([0.0848, 0.1619, 0.1765, 0.2009, 0.2317, 0.0875, 0.1331, 0.2216], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0607, 0.0665, 0.0503, 0.0671, 0.0698, 0.0522, 0.0670], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:40:48,054 INFO [optim.py:368] (0/8) 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 10:40:49,971 INFO [train.py:904] (0/8) Epoch 22, batch 4050, loss[loss=0.1717, simple_loss=0.2524, pruned_loss=0.04553, over 16531.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2535, pruned_loss=0.04665, over 3286039.04 frames. ], batch size: 62, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:41:06,473 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6959, 1.8540, 2.3899, 2.5799, 2.6866, 2.9992, 1.9853, 2.9290], device='cuda:0'), covar=tensor([0.0214, 0.0524, 0.0320, 0.0326, 0.0320, 0.0183, 0.0541, 0.0117], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0195, 0.0181, 0.0187, 0.0199, 0.0156, 0.0198, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:41:51,981 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3001, 3.4507, 3.4320, 1.8046, 2.8915, 1.9896, 3.7871, 3.8167], device='cuda:0'), covar=tensor([0.0231, 0.0803, 0.0766, 0.2479, 0.1045, 0.1249, 0.0532, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0165, 0.0166, 0.0153, 0.0145, 0.0130, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:41:55,341 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:41:56,827 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7894, 4.0331, 3.0122, 2.4143, 2.7164, 2.6977, 4.4365, 3.6174], device='cuda:0'), covar=tensor([0.2882, 0.0611, 0.1868, 0.2674, 0.2757, 0.1888, 0.0391, 0.1147], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0272, 0.0308, 0.0316, 0.0301, 0.0263, 0.0297, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:42:04,912 INFO [train.py:904] (0/8) Epoch 22, batch 4100, loss[loss=0.1898, simple_loss=0.2735, pruned_loss=0.05304, over 16781.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2547, pruned_loss=0.0459, over 3277132.15 frames. ], batch size: 124, lr: 3.08e-03, grad_scale: 8.0 2023-05-01 10:43:02,147 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 10:43:04,918 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9056, 3.3256, 3.4624, 1.9639, 2.9423, 2.2493, 3.3416, 3.5595], device='cuda:0'), covar=tensor([0.0273, 0.0840, 0.0564, 0.2193, 0.0862, 0.1038, 0.0738, 0.0973], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0165, 0.0167, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:43:11,033 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217295.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:43:21,264 INFO [optim.py:368] (0/8) 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 10:43:23,211 INFO [train.py:904] (0/8) Epoch 22, batch 4150, loss[loss=0.1955, simple_loss=0.2925, pruned_loss=0.04924, over 16651.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.262, pruned_loss=0.04821, over 3236203.28 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:43:36,095 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217311.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 10:43:51,598 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:44:39,666 INFO [train.py:904] (0/8) Epoch 22, batch 4200, loss[loss=0.2241, simple_loss=0.3127, pruned_loss=0.06776, over 16496.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2694, pruned_loss=0.04987, over 3225807.63 frames. ], batch size: 75, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:44:50,073 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217359.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:44:52,811 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:32,908 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217388.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:45:39,702 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 10:45:53,907 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.404e+02 2.187e+02 2.459e+02 3.018e+02 5.838e+02, threshold=4.919e+02, percent-clipped=1.0 2023-05-01 10:45:55,232 INFO [train.py:904] (0/8) Epoch 22, batch 4250, loss[loss=0.1806, simple_loss=0.2817, pruned_loss=0.0398, over 17027.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2728, pruned_loss=0.0496, over 3211497.89 frames. ], batch size: 50, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:46:04,074 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217409.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:46:08,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2239, 1.5089, 2.0010, 2.1170, 2.3062, 2.4433, 1.8227, 2.2734], device='cuda:0'), covar=tensor([0.0208, 0.0487, 0.0276, 0.0314, 0.0283, 0.0173, 0.0468, 0.0132], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0187, 0.0198, 0.0155, 0.0198, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:46:23,537 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3800, 4.3014, 4.4176, 4.5761, 4.7585, 4.2838, 4.6720, 4.7756], device='cuda:0'), covar=tensor([0.1898, 0.1262, 0.1511, 0.0726, 0.0529, 0.1176, 0.0856, 0.0570], device='cuda:0'), in_proj_covar=tensor([0.0648, 0.0801, 0.0930, 0.0814, 0.0616, 0.0643, 0.0666, 0.0770], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:47:04,282 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217449.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:47:09,162 INFO [train.py:904] (0/8) Epoch 22, batch 4300, loss[loss=0.2032, simple_loss=0.2969, pruned_loss=0.05474, over 16710.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2738, pruned_loss=0.0482, over 3221110.84 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:47:12,303 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0335, 3.0356, 2.6601, 2.8891, 3.4095, 3.0678, 3.5203, 3.5560], device='cuda:0'), covar=tensor([0.0065, 0.0353, 0.0452, 0.0359, 0.0217, 0.0313, 0.0219, 0.0203], device='cuda:0'), in_proj_covar=tensor([0.0216, 0.0237, 0.0229, 0.0229, 0.0239, 0.0238, 0.0241, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:48:07,882 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217492.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:48:23,055 INFO [optim.py:368] (0/8) 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 10:48:24,313 INFO [train.py:904] (0/8) Epoch 22, batch 4350, loss[loss=0.1995, simple_loss=0.286, pruned_loss=0.05648, over 16931.00 frames. ], tot_loss[loss=0.1876, simple_loss=0.277, pruned_loss=0.04915, over 3230865.25 frames. ], batch size: 109, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:28,134 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7757, 3.8573, 2.1559, 4.8334, 2.9281, 4.6383, 2.1459, 2.9880], device='cuda:0'), covar=tensor([0.0285, 0.0354, 0.1941, 0.0127, 0.0848, 0.0367, 0.1938, 0.0879], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0177, 0.0194, 0.0164, 0.0177, 0.0219, 0.0201, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:49:39,992 INFO [train.py:904] (0/8) Epoch 22, batch 4400, loss[loss=0.2074, simple_loss=0.2934, pruned_loss=0.06072, over 16725.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2799, pruned_loss=0.05119, over 3208389.42 frames. ], batch size: 62, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:49:40,524 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217553.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:50:52,395 INFO [optim.py:368] (0/8) 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 10:50:53,489 INFO [train.py:904] (0/8) Epoch 22, batch 4450, loss[loss=0.2012, simple_loss=0.2917, pruned_loss=0.05535, over 15326.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2828, pruned_loss=0.05231, over 3213606.45 frames. ], batch size: 191, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:51:20,394 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=217621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:52:08,357 INFO [train.py:904] (0/8) Epoch 22, batch 4500, loss[loss=0.2054, simple_loss=0.2906, pruned_loss=0.06014, over 15494.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.283, pruned_loss=0.05268, over 3236197.21 frames. ], batch size: 191, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:52:24,065 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:52:32,193 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=217669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:52:35,424 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7159, 2.4083, 2.2934, 3.0935, 2.3528, 3.4552, 1.5078, 2.6658], device='cuda:0'), covar=tensor([0.1356, 0.0839, 0.1294, 0.0155, 0.0171, 0.0356, 0.1756, 0.0840], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0174, 0.0195, 0.0192, 0.0205, 0.0216, 0.0202, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:53:18,213 INFO [optim.py:368] (0/8) 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 10:53:19,304 INFO [train.py:904] (0/8) Epoch 22, batch 4550, loss[loss=0.1956, simple_loss=0.2791, pruned_loss=0.05604, over 17039.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2835, pruned_loss=0.05372, over 3233246.89 frames. ], batch size: 41, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:53:52,645 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:19,814 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217744.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:54:31,799 INFO [train.py:904] (0/8) Epoch 22, batch 4600, loss[loss=0.1959, simple_loss=0.2855, pruned_loss=0.05314, over 16539.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2844, pruned_loss=0.05397, over 3239528.68 frames. ], batch size: 35, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:54:49,222 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9329, 2.1517, 2.2062, 3.4531, 2.0422, 2.4507, 2.2655, 2.2652], device='cuda:0'), covar=tensor([0.1434, 0.3350, 0.2781, 0.0613, 0.4192, 0.2380, 0.3271, 0.3448], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0452, 0.0369, 0.0330, 0.0439, 0.0520, 0.0423, 0.0529], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:54:52,515 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6478, 4.5213, 4.7168, 4.8570, 4.9737, 4.5355, 4.9715, 5.0117], device='cuda:0'), covar=tensor([0.1546, 0.1092, 0.1274, 0.0582, 0.0459, 0.0836, 0.0554, 0.0536], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0791, 0.0919, 0.0802, 0.0608, 0.0636, 0.0657, 0.0761], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:55:25,194 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4288, 4.5607, 4.7537, 4.5141, 4.5859, 5.1520, 4.6734, 4.3391], device='cuda:0'), covar=tensor([0.1438, 0.2004, 0.2176, 0.1928, 0.2653, 0.0989, 0.1477, 0.2526], device='cuda:0'), in_proj_covar=tensor([0.0409, 0.0587, 0.0642, 0.0486, 0.0648, 0.0680, 0.0506, 0.0653], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 10:55:25,277 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1020, 3.9929, 4.1495, 4.2691, 4.3697, 4.0149, 4.2841, 4.4098], device='cuda:0'), covar=tensor([0.1422, 0.1080, 0.1274, 0.0629, 0.0488, 0.1397, 0.0858, 0.0607], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0791, 0.0919, 0.0802, 0.0608, 0.0636, 0.0658, 0.0760], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:55:41,528 INFO [optim.py:368] (0/8) 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 10:55:42,805 INFO [train.py:904] (0/8) Epoch 22, batch 4650, loss[loss=0.1814, simple_loss=0.2735, pruned_loss=0.0446, over 15420.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2836, pruned_loss=0.05414, over 3231350.60 frames. ], batch size: 191, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:55:55,882 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9416, 4.9112, 4.6907, 4.0810, 4.8472, 1.7278, 4.5774, 4.2251], device='cuda:0'), covar=tensor([0.0055, 0.0051, 0.0143, 0.0280, 0.0053, 0.3020, 0.0082, 0.0229], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0158, 0.0203, 0.0181, 0.0180, 0.0211, 0.0192, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:56:46,087 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=217848.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 10:56:52,992 INFO [train.py:904] (0/8) Epoch 22, batch 4700, loss[loss=0.1842, simple_loss=0.2768, pruned_loss=0.04581, over 15434.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2805, pruned_loss=0.05289, over 3219906.24 frames. ], batch size: 190, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:57:32,367 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8066, 4.8145, 4.7046, 3.9044, 4.7489, 1.6472, 4.5105, 4.4174], device='cuda:0'), covar=tensor([0.0108, 0.0116, 0.0163, 0.0530, 0.0109, 0.3058, 0.0136, 0.0262], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0159, 0.0203, 0.0182, 0.0180, 0.0212, 0.0192, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 10:57:44,787 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5559, 2.5359, 2.2507, 3.9532, 2.5116, 3.8261, 1.4675, 2.5966], device='cuda:0'), covar=tensor([0.1562, 0.0952, 0.1493, 0.0216, 0.0257, 0.0445, 0.1891, 0.1053], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0191, 0.0204, 0.0215, 0.0201, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 10:57:59,228 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=217898.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 10:58:04,666 INFO [optim.py:368] (0/8) 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 10:58:05,934 INFO [train.py:904] (0/8) Epoch 22, batch 4750, loss[loss=0.178, simple_loss=0.2577, pruned_loss=0.0492, over 16254.00 frames. ], tot_loss[loss=0.1895, simple_loss=0.277, pruned_loss=0.05104, over 3199832.13 frames. ], batch size: 35, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:59:17,336 INFO [train.py:904] (0/8) Epoch 22, batch 4800, loss[loss=0.1794, simple_loss=0.2716, pruned_loss=0.04364, over 16439.00 frames. ], tot_loss[loss=0.1854, simple_loss=0.273, pruned_loss=0.04888, over 3199899.13 frames. ], batch size: 75, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 10:59:27,572 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=217959.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:00:09,080 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7227, 2.6182, 2.4733, 4.0479, 2.5150, 3.9341, 1.4743, 2.8776], device='cuda:0'), covar=tensor([0.1373, 0.0828, 0.1214, 0.0127, 0.0158, 0.0369, 0.1726, 0.0780], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0191, 0.0204, 0.0214, 0.0200, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:00:28,201 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-218000.pt 2023-05-01 11:00:36,343 INFO [optim.py:368] (0/8) 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 11:00:36,359 INFO [train.py:904] (0/8) Epoch 22, batch 4850, loss[loss=0.1803, simple_loss=0.2758, pruned_loss=0.04236, over 15411.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2737, pruned_loss=0.04796, over 3177041.50 frames. ], batch size: 190, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:01:01,610 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218020.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:27,702 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7922, 4.0564, 4.4576, 2.1925, 4.7431, 4.8917, 3.3795, 3.4549], device='cuda:0'), covar=tensor([0.1084, 0.0222, 0.0170, 0.1290, 0.0067, 0.0087, 0.0409, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0108, 0.0098, 0.0138, 0.0081, 0.0125, 0.0128, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 11:01:38,625 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:01:43,712 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.62 vs. limit=5.0 2023-05-01 11:01:51,560 INFO [train.py:904] (0/8) Epoch 22, batch 4900, loss[loss=0.1509, simple_loss=0.2472, pruned_loss=0.02732, over 17232.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2724, pruned_loss=0.04651, over 3175854.06 frames. ], batch size: 45, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:02:17,294 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.4558, 2.4481, 2.3093, 3.9890, 2.5719, 3.8855, 1.4142, 2.7571], device='cuda:0'), covar=tensor([0.1488, 0.0885, 0.1350, 0.0148, 0.0209, 0.0392, 0.1774, 0.0872], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0190, 0.0203, 0.0214, 0.0200, 0.0190], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:02:35,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7517, 4.8215, 4.6597, 4.3268, 4.2927, 4.7547, 4.5521, 4.4456], device='cuda:0'), covar=tensor([0.0625, 0.0519, 0.0309, 0.0297, 0.0995, 0.0554, 0.0386, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0291, 0.0426, 0.0341, 0.0341, 0.0350, 0.0393, 0.0234, 0.0407], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:02:49,276 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218092.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:01,685 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218100.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:03:05,450 INFO [optim.py:368] (0/8) 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 11:03:05,471 INFO [train.py:904] (0/8) Epoch 22, batch 4950, loss[loss=0.1799, simple_loss=0.2678, pruned_loss=0.04597, over 16603.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2721, pruned_loss=0.04601, over 3188324.45 frames. ], batch size: 57, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:03:28,888 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5042, 1.7200, 2.1582, 2.4191, 2.5054, 2.8561, 1.8359, 2.7183], device='cuda:0'), covar=tensor([0.0243, 0.0597, 0.0360, 0.0372, 0.0361, 0.0197, 0.0636, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0196, 0.0181, 0.0187, 0.0200, 0.0155, 0.0199, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:03:46,725 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 11:04:10,898 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:04:18,920 INFO [train.py:904] (0/8) Epoch 22, batch 5000, loss[loss=0.1851, simple_loss=0.2798, pruned_loss=0.04523, over 16688.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2732, pruned_loss=0.04593, over 3194979.00 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:04:30,088 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:00,950 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218196.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:05:32,361 INFO [optim.py:368] (0/8) 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 11:05:32,377 INFO [train.py:904] (0/8) Epoch 22, batch 5050, loss[loss=0.1723, simple_loss=0.2645, pruned_loss=0.04006, over 17098.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2741, pruned_loss=0.046, over 3199192.17 frames. ], batch size: 49, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:05:42,341 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218210.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:06:29,683 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:06:44,158 INFO [train.py:904] (0/8) Epoch 22, batch 5100, loss[loss=0.1477, simple_loss=0.2321, pruned_loss=0.03165, over 16339.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.272, pruned_loss=0.04522, over 3205567.65 frames. ], batch size: 35, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:06:45,778 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218254.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 11:07:00,173 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.13 vs. limit=5.0 2023-05-01 11:07:10,611 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218271.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:07:57,332 INFO [optim.py:368] (0/8) 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 11:07:57,353 INFO [train.py:904] (0/8) Epoch 22, batch 5150, loss[loss=0.1792, simple_loss=0.2708, pruned_loss=0.04378, over 16679.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.2715, pruned_loss=0.04442, over 3194401.97 frames. ], batch size: 134, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:08:06,628 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4337, 4.4901, 4.7676, 4.7254, 4.7383, 4.5179, 4.4320, 4.3887], device='cuda:0'), covar=tensor([0.0317, 0.0542, 0.0372, 0.0424, 0.0439, 0.0347, 0.0859, 0.0418], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0440, 0.0427, 0.0396, 0.0473, 0.0445, 0.0532, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 11:08:23,170 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:08:29,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4078, 2.0622, 1.6425, 1.7445, 2.3484, 2.1036, 2.2421, 2.6105], device='cuda:0'), covar=tensor([0.0221, 0.0527, 0.0741, 0.0631, 0.0293, 0.0493, 0.0222, 0.0361], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0233, 0.0225, 0.0224, 0.0234, 0.0233, 0.0235, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:08:36,102 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 11:08:58,740 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 11:09:02,159 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-01 11:09:11,245 INFO [train.py:904] (0/8) Epoch 22, batch 5200, loss[loss=0.1624, simple_loss=0.2649, pruned_loss=0.02991, over 16824.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2703, pruned_loss=0.04355, over 3195301.61 frames. ], batch size: 102, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:09:33,043 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218368.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:09:34,448 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7400, 3.7068, 4.2013, 1.9912, 4.3871, 4.3654, 3.0709, 3.1669], device='cuda:0'), covar=tensor([0.0803, 0.0259, 0.0148, 0.1265, 0.0063, 0.0116, 0.0422, 0.0482], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0110, 0.0099, 0.0140, 0.0082, 0.0127, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:10:23,969 INFO [optim.py:368] (0/8) 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 11:10:23,984 INFO [train.py:904] (0/8) Epoch 22, batch 5250, loss[loss=0.1701, simple_loss=0.2612, pruned_loss=0.0395, over 16890.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2684, pruned_loss=0.04358, over 3193750.37 frames. ], batch size: 109, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:10:32,828 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1010, 2.0328, 2.6268, 3.0135, 2.8665, 3.4565, 2.1698, 3.4623], device='cuda:0'), covar=tensor([0.0205, 0.0533, 0.0334, 0.0316, 0.0316, 0.0164, 0.0578, 0.0125], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0196, 0.0181, 0.0187, 0.0200, 0.0155, 0.0199, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:11:37,158 INFO [train.py:904] (0/8) Epoch 22, batch 5300, loss[loss=0.1389, simple_loss=0.2344, pruned_loss=0.0217, over 16841.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.2654, pruned_loss=0.04281, over 3193436.37 frames. ], batch size: 102, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:11:41,568 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218456.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:12:51,258 INFO [optim.py:368] (0/8) 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 11:12:51,279 INFO [train.py:904] (0/8) Epoch 22, batch 5350, loss[loss=0.1927, simple_loss=0.2799, pruned_loss=0.05274, over 17021.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2645, pruned_loss=0.04251, over 3205952.30 frames. ], batch size: 53, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:13:00,640 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:13:27,168 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0694, 2.2391, 2.7127, 3.0631, 2.9894, 3.6266, 2.3349, 3.5363], device='cuda:0'), covar=tensor([0.0227, 0.0461, 0.0326, 0.0310, 0.0293, 0.0142, 0.0478, 0.0113], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0186, 0.0199, 0.0154, 0.0199, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:13:43,231 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:03,270 INFO [train.py:904] (0/8) Epoch 22, batch 5400, loss[loss=0.1814, simple_loss=0.2692, pruned_loss=0.0468, over 15375.00 frames. ], tot_loss[loss=0.1762, simple_loss=0.2663, pruned_loss=0.04311, over 3187626.01 frames. ], batch size: 190, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:14:06,150 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218554.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:14:23,166 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218566.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:14:28,905 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:34,236 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218574.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:14:50,399 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7333, 1.3269, 1.6790, 1.6810, 1.8216, 1.9778, 1.6309, 1.8632], device='cuda:0'), covar=tensor([0.0266, 0.0417, 0.0216, 0.0320, 0.0282, 0.0167, 0.0433, 0.0128], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0195, 0.0181, 0.0186, 0.0199, 0.0154, 0.0198, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:15:18,770 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218602.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 11:15:19,494 INFO [optim.py:368] (0/8) 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 11:15:19,514 INFO [train.py:904] (0/8) Epoch 22, batch 5450, loss[loss=0.2039, simple_loss=0.2919, pruned_loss=0.0579, over 16672.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2697, pruned_loss=0.04453, over 3175015.08 frames. ], batch size: 76, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:15:35,420 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3729, 3.3206, 3.3311, 3.4460, 3.4976, 3.2706, 3.4491, 3.5414], device='cuda:0'), covar=tensor([0.1246, 0.1067, 0.1133, 0.0854, 0.0778, 0.2546, 0.1241, 0.1199], device='cuda:0'), in_proj_covar=tensor([0.0633, 0.0786, 0.0912, 0.0797, 0.0598, 0.0631, 0.0651, 0.0752], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:16:10,919 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=218635.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:16:37,997 INFO [train.py:904] (0/8) Epoch 22, batch 5500, loss[loss=0.2035, simple_loss=0.2898, pruned_loss=0.05866, over 16342.00 frames. ], tot_loss[loss=0.1875, simple_loss=0.277, pruned_loss=0.04905, over 3143322.64 frames. ], batch size: 146, lr: 3.07e-03, grad_scale: 8.0 2023-05-01 11:17:57,929 INFO [optim.py:368] (0/8) 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 11:17:57,949 INFO [train.py:904] (0/8) Epoch 22, batch 5550, loss[loss=0.2128, simple_loss=0.3007, pruned_loss=0.06243, over 16871.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2835, pruned_loss=0.05302, over 3156981.45 frames. ], batch size: 116, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:18:12,597 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 11:19:20,445 INFO [train.py:904] (0/8) Epoch 22, batch 5600, loss[loss=0.2165, simple_loss=0.2981, pruned_loss=0.06741, over 16728.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.288, pruned_loss=0.05659, over 3132041.93 frames. ], batch size: 124, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:19:25,850 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218756.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:20:41,876 INFO [optim.py:368] (0/8) 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 11:20:41,899 INFO [train.py:904] (0/8) Epoch 22, batch 5650, loss[loss=0.221, simple_loss=0.3014, pruned_loss=0.07033, over 16316.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2921, pruned_loss=0.05969, over 3116352.02 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:20:43,479 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218804.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:21:35,405 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218838.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:21:56,893 INFO [train.py:904] (0/8) Epoch 22, batch 5700, loss[loss=0.2089, simple_loss=0.2952, pruned_loss=0.06126, over 16623.00 frames. ], tot_loss[loss=0.2089, simple_loss=0.2939, pruned_loss=0.06192, over 3076842.16 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:22:12,170 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 11:22:16,168 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218865.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:17,600 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=218866.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:22:48,049 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218886.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:13,876 INFO [optim.py:368] (0/8) 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 11:23:13,898 INFO [train.py:904] (0/8) Epoch 22, batch 5750, loss[loss=0.1918, simple_loss=0.2818, pruned_loss=0.05085, over 16281.00 frames. ], tot_loss[loss=0.2119, simple_loss=0.297, pruned_loss=0.06336, over 3072822.14 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:23:32,118 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=218914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:23:55,816 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8916, 2.8035, 2.8270, 2.1694, 2.6990, 2.2038, 2.6886, 2.9319], device='cuda:0'), covar=tensor([0.0288, 0.0717, 0.0552, 0.1708, 0.0794, 0.0942, 0.0559, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0166, 0.0153, 0.0145, 0.0130, 0.0143, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:23:57,124 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=218930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:24:34,389 INFO [train.py:904] (0/8) Epoch 22, batch 5800, loss[loss=0.2023, simple_loss=0.2918, pruned_loss=0.05639, over 16429.00 frames. ], tot_loss[loss=0.2105, simple_loss=0.2964, pruned_loss=0.06226, over 3070995.45 frames. ], batch size: 146, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:24:54,949 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:05,556 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 11:25:15,104 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1467, 3.5924, 3.5744, 2.2220, 3.3526, 3.7146, 3.3977, 1.6329], device='cuda:0'), covar=tensor([0.0648, 0.0099, 0.0096, 0.0577, 0.0138, 0.0168, 0.0147, 0.0713], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0083, 0.0084, 0.0133, 0.0098, 0.0110, 0.0095, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 11:25:24,080 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=218984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:25:53,554 INFO [optim.py:368] (0/8) 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 11:25:53,574 INFO [train.py:904] (0/8) Epoch 22, batch 5850, loss[loss=0.1984, simple_loss=0.288, pruned_loss=0.05441, over 16602.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.294, pruned_loss=0.06062, over 3073798.84 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:26:21,609 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:26:30,047 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219026.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:02,377 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:27:14,936 INFO [train.py:904] (0/8) Epoch 22, batch 5900, loss[loss=0.2069, simple_loss=0.293, pruned_loss=0.06039, over 16729.00 frames. ], tot_loss[loss=0.2078, simple_loss=0.2939, pruned_loss=0.06085, over 3082413.75 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:27:49,359 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219072.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:04,732 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:28:35,887 INFO [optim.py:368] (0/8) 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 11:28:35,908 INFO [train.py:904] (0/8) Epoch 22, batch 5950, loss[loss=0.1822, simple_loss=0.2853, pruned_loss=0.03954, over 16481.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2944, pruned_loss=0.0594, over 3077785.33 frames. ], batch size: 68, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:28:51,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9169, 2.9040, 2.4614, 4.7027, 3.4196, 4.1137, 1.7322, 3.0363], device='cuda:0'), covar=tensor([0.1267, 0.0758, 0.1349, 0.0180, 0.0363, 0.0431, 0.1615, 0.0834], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0173, 0.0194, 0.0190, 0.0205, 0.0214, 0.0201, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:29:24,523 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:29:57,674 INFO [train.py:904] (0/8) Epoch 22, batch 6000, loss[loss=0.2151, simple_loss=0.2962, pruned_loss=0.06696, over 15497.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2936, pruned_loss=0.05933, over 3087127.27 frames. ], batch size: 191, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:29:57,674 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 11:30:07,617 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 11:30:27,750 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219165.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:28,923 INFO [optim.py:368] (0/8) 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 11:31:28,944 INFO [train.py:904] (0/8) Epoch 22, batch 6050, loss[loss=0.1931, simple_loss=0.2817, pruned_loss=0.05219, over 15285.00 frames. ], tot_loss[loss=0.2045, simple_loss=0.292, pruned_loss=0.05848, over 3094712.84 frames. ], batch size: 190, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:31:29,592 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:31:44,912 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:03,243 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 11:32:10,555 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219230.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:32:46,228 INFO [train.py:904] (0/8) Epoch 22, batch 6100, loss[loss=0.2117, simple_loss=0.2849, pruned_loss=0.06929, over 11491.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2909, pruned_loss=0.05743, over 3089101.00 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:33:05,456 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219264.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:26,351 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:33:26,556 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5595, 1.8513, 2.2194, 2.5256, 2.5538, 2.8670, 1.8844, 2.7474], device='cuda:0'), covar=tensor([0.0235, 0.0519, 0.0303, 0.0361, 0.0315, 0.0198, 0.0554, 0.0137], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0193, 0.0179, 0.0183, 0.0197, 0.0152, 0.0196, 0.0151], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:33:33,768 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 11:34:03,662 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-01 11:34:04,119 INFO [train.py:904] (0/8) Epoch 22, batch 6150, loss[loss=0.1912, simple_loss=0.282, pruned_loss=0.0502, over 16566.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2895, pruned_loss=0.05719, over 3082888.64 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:34:05,867 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.700e+02 2.531e+02 2.938e+02 3.735e+02 5.885e+02, threshold=5.877e+02, percent-clipped=0.0 2023-05-01 11:34:33,729 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219321.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:03,742 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:35:06,296 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219342.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:35:22,811 INFO [train.py:904] (0/8) Epoch 22, batch 6200, loss[loss=0.2283, simple_loss=0.2947, pruned_loss=0.08093, over 11394.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2883, pruned_loss=0.05756, over 3076436.58 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:02,468 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219377.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:36:13,825 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 11:36:42,474 INFO [train.py:904] (0/8) Epoch 22, batch 6250, loss[loss=0.2354, simple_loss=0.306, pruned_loss=0.0824, over 11429.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2887, pruned_loss=0.05806, over 3067656.50 frames. ], batch size: 248, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:36:43,145 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219403.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 11:36:43,737 INFO [optim.py:368] (0/8) 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 11:37:20,352 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:37:57,578 INFO [train.py:904] (0/8) Epoch 22, batch 6300, loss[loss=0.1993, simple_loss=0.2796, pruned_loss=0.05954, over 16649.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2885, pruned_loss=0.05686, over 3100526.99 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:38:50,312 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6523, 1.7843, 1.5959, 1.5232, 1.9593, 1.6763, 1.6138, 1.9479], device='cuda:0'), covar=tensor([0.0205, 0.0297, 0.0414, 0.0369, 0.0209, 0.0239, 0.0181, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0207, 0.0230, 0.0223, 0.0222, 0.0231, 0.0230, 0.0231, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:38:55,967 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219490.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:39:15,309 INFO [train.py:904] (0/8) Epoch 22, batch 6350, loss[loss=0.2048, simple_loss=0.2868, pruned_loss=0.0614, over 16591.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2891, pruned_loss=0.05803, over 3103782.35 frames. ], batch size: 62, lr: 3.06e-03, grad_scale: 4.0 2023-05-01 11:39:16,414 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.781e+02 2.840e+02 3.669e+02 4.695e+02 9.321e+02, threshold=7.339e+02, percent-clipped=9.0 2023-05-01 11:39:21,304 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3480, 3.1856, 3.5473, 1.8287, 3.6289, 3.7185, 2.8239, 2.7377], device='cuda:0'), covar=tensor([0.0831, 0.0285, 0.0193, 0.1196, 0.0094, 0.0191, 0.0475, 0.0488], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0110, 0.0099, 0.0141, 0.0082, 0.0128, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:40:28,591 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:40:31,091 INFO [train.py:904] (0/8) Epoch 22, batch 6400, loss[loss=0.2611, simple_loss=0.3305, pruned_loss=0.09584, over 11408.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.29, pruned_loss=0.05976, over 3080914.99 frames. ], batch size: 246, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:40:40,476 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219559.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:40:49,482 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1898, 3.2927, 3.6347, 2.1211, 3.1831, 2.3630, 3.6633, 3.5909], device='cuda:0'), covar=tensor([0.0237, 0.0773, 0.0540, 0.1990, 0.0757, 0.0938, 0.0540, 0.0895], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0164, 0.0167, 0.0153, 0.0145, 0.0130, 0.0143, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:41:35,047 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7086, 3.7563, 2.3185, 4.3558, 2.9261, 4.3075, 2.4888, 3.1061], device='cuda:0'), covar=tensor([0.0274, 0.0412, 0.1670, 0.0217, 0.0775, 0.0572, 0.1497, 0.0720], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0161, 0.0176, 0.0216, 0.0200, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:41:45,428 INFO [train.py:904] (0/8) Epoch 22, batch 6450, loss[loss=0.2009, simple_loss=0.2866, pruned_loss=0.05766, over 16893.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2893, pruned_loss=0.05794, over 3103804.41 frames. ], batch size: 116, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:41:47,190 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.783e+02 2.795e+02 3.528e+02 4.334e+02 7.123e+02, threshold=7.056e+02, percent-clipped=0.0 2023-05-01 11:42:13,399 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:42:45,455 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:43:04,573 INFO [train.py:904] (0/8) Epoch 22, batch 6500, loss[loss=0.2033, simple_loss=0.2888, pruned_loss=0.05887, over 16686.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2867, pruned_loss=0.05719, over 3101866.42 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:43:20,097 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 11:43:30,551 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:43:42,922 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:00,774 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:44:04,831 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5644, 4.7451, 4.8841, 4.6626, 4.7860, 5.2881, 4.7170, 4.4433], device='cuda:0'), covar=tensor([0.1292, 0.1930, 0.2725, 0.1984, 0.2194, 0.0860, 0.1771, 0.2690], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0593, 0.0654, 0.0495, 0.0658, 0.0683, 0.0515, 0.0665], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 11:44:19,786 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219698.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:44:26,649 INFO [train.py:904] (0/8) Epoch 22, batch 6550, loss[loss=0.2227, simple_loss=0.3183, pruned_loss=0.06359, over 16729.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2894, pruned_loss=0.05807, over 3108306.88 frames. ], batch size: 124, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:44:28,428 INFO [optim.py:368] (0/8) 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 11:44:44,117 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 11:45:03,115 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219725.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:08,248 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219728.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:45:46,182 INFO [train.py:904] (0/8) Epoch 22, batch 6600, loss[loss=0.2257, simple_loss=0.3118, pruned_loss=0.06977, over 16372.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.292, pruned_loss=0.05911, over 3081426.10 frames. ], batch size: 165, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:46:02,121 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219763.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:23,297 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219776.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:46:29,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-01 11:47:06,517 INFO [train.py:904] (0/8) Epoch 22, batch 6650, loss[loss=0.2041, simple_loss=0.29, pruned_loss=0.05906, over 16858.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2912, pruned_loss=0.05905, over 3106661.74 frames. ], batch size: 116, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:47:07,643 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.837e+02 2.943e+02 3.397e+02 4.500e+02 8.184e+02, threshold=6.793e+02, percent-clipped=7.0 2023-05-01 11:47:39,198 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219824.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:10,737 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=219846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:20,386 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8520, 5.1285, 5.3183, 5.0764, 5.1281, 5.6747, 5.1903, 4.9525], device='cuda:0'), covar=tensor([0.1002, 0.1802, 0.2105, 0.1644, 0.2120, 0.0836, 0.1428, 0.2127], device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0586, 0.0649, 0.0488, 0.0651, 0.0676, 0.0510, 0.0656], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 11:48:21,251 INFO [train.py:904] (0/8) Epoch 22, batch 6700, loss[loss=0.2081, simple_loss=0.3079, pruned_loss=0.05413, over 16759.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2898, pruned_loss=0.0585, over 3133602.88 frames. ], batch size: 89, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:48:27,548 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=219857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:48:30,125 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219859.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:49:07,039 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 11:49:36,043 INFO [train.py:904] (0/8) Epoch 22, batch 6750, loss[loss=0.1802, simple_loss=0.2672, pruned_loss=0.04659, over 16508.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2891, pruned_loss=0.05902, over 3116318.76 frames. ], batch size: 75, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:49:37,874 INFO [optim.py:368] (0/8) 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 11:49:43,852 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=219907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:00,726 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=219918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:50:30,191 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 11:50:54,132 INFO [train.py:904] (0/8) Epoch 22, batch 6800, loss[loss=0.2024, simple_loss=0.2934, pruned_loss=0.0557, over 16658.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2891, pruned_loss=0.05892, over 3113588.91 frames. ], batch size: 134, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:51:07,235 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3507, 3.1859, 3.5702, 1.8013, 3.6867, 3.7729, 2.8272, 2.7101], device='cuda:0'), covar=tensor([0.0938, 0.0291, 0.0222, 0.1296, 0.0090, 0.0172, 0.0498, 0.0565], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0109, 0.0099, 0.0139, 0.0081, 0.0127, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:51:58,208 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 11:52:04,363 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=219998.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:52:07,920 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-220000.pt 2023-05-01 11:52:15,005 INFO [train.py:904] (0/8) Epoch 22, batch 6850, loss[loss=0.2037, simple_loss=0.3089, pruned_loss=0.04928, over 16662.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2899, pruned_loss=0.05892, over 3122730.11 frames. ], batch size: 57, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:52:16,800 INFO [optim.py:368] (0/8) 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 11:52:31,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0318, 5.3928, 5.7424, 5.6771, 5.6012, 5.3699, 4.9124, 5.1147], device='cuda:0'), covar=tensor([0.0628, 0.0575, 0.0521, 0.0594, 0.0676, 0.0540, 0.1707, 0.0500], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0448, 0.0434, 0.0402, 0.0483, 0.0456, 0.0540, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 11:53:03,300 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-01 11:53:20,651 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220046.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 11:53:31,123 INFO [train.py:904] (0/8) Epoch 22, batch 6900, loss[loss=0.205, simple_loss=0.2939, pruned_loss=0.0581, over 16478.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2923, pruned_loss=0.0585, over 3139136.28 frames. ], batch size: 75, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:54:00,676 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 11:54:24,034 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220087.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:41,863 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220098.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:54:49,993 INFO [train.py:904] (0/8) Epoch 22, batch 6950, loss[loss=0.1812, simple_loss=0.273, pruned_loss=0.04467, over 16481.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2941, pruned_loss=0.06054, over 3114698.77 frames. ], batch size: 68, lr: 3.06e-03, grad_scale: 8.0 2023-05-01 11:54:51,083 INFO [optim.py:368] (0/8) 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 11:55:03,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7773, 3.1578, 3.3022, 2.0091, 2.8572, 2.1190, 3.2518, 3.3140], device='cuda:0'), covar=tensor([0.0277, 0.0785, 0.0608, 0.2125, 0.0886, 0.1084, 0.0670, 0.0945], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0165, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 11:55:14,108 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220119.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:22,686 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4032, 3.3973, 3.4288, 3.5230, 3.5596, 3.3003, 3.5361, 3.6164], device='cuda:0'), covar=tensor([0.1348, 0.1027, 0.1054, 0.0655, 0.0693, 0.2693, 0.1101, 0.0821], device='cuda:0'), in_proj_covar=tensor([0.0632, 0.0784, 0.0908, 0.0791, 0.0601, 0.0625, 0.0653, 0.0754], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:55:52,302 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6609, 4.7228, 5.0841, 5.0500, 5.0482, 4.7657, 4.7154, 4.6297], device='cuda:0'), covar=tensor([0.0410, 0.0601, 0.0509, 0.0473, 0.0504, 0.0535, 0.1046, 0.0504], device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0449, 0.0434, 0.0403, 0.0483, 0.0457, 0.0542, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 11:55:54,614 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:55:57,620 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:56:03,780 INFO [train.py:904] (0/8) Epoch 22, batch 7000, loss[loss=0.1967, simple_loss=0.3027, pruned_loss=0.04537, over 16537.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2941, pruned_loss=0.05965, over 3113613.42 frames. ], batch size: 75, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 11:56:14,305 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:56:32,702 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4940, 2.1583, 1.7607, 1.9615, 2.5078, 2.1536, 2.2757, 2.6396], device='cuda:0'), covar=tensor([0.0219, 0.0416, 0.0579, 0.0484, 0.0237, 0.0389, 0.0207, 0.0285], device='cuda:0'), in_proj_covar=tensor([0.0207, 0.0231, 0.0223, 0.0222, 0.0232, 0.0230, 0.0230, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 11:57:05,652 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220194.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:18,842 INFO [train.py:904] (0/8) Epoch 22, batch 7050, loss[loss=0.2044, simple_loss=0.2962, pruned_loss=0.05625, over 16181.00 frames. ], tot_loss[loss=0.2074, simple_loss=0.295, pruned_loss=0.0599, over 3100316.84 frames. ], batch size: 165, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:57:21,955 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.068e+02 2.577e+02 3.204e+02 3.781e+02 6.030e+02, threshold=6.408e+02, percent-clipped=0.0 2023-05-01 11:57:34,762 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220213.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:57:58,647 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:58:36,516 INFO [train.py:904] (0/8) Epoch 22, batch 7100, loss[loss=0.2147, simple_loss=0.3029, pruned_loss=0.06327, over 16454.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2935, pruned_loss=0.05944, over 3103145.99 frames. ], batch size: 146, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:58:45,485 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:35,720 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 11:59:54,492 INFO [train.py:904] (0/8) Epoch 22, batch 7150, loss[loss=0.2121, simple_loss=0.2958, pruned_loss=0.06416, over 16900.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2917, pruned_loss=0.05956, over 3095997.99 frames. ], batch size: 109, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 11:59:58,138 INFO [optim.py:368] (0/8) 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 12:00:20,307 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220319.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:00:48,189 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1605, 4.0611, 4.2364, 4.3515, 4.4947, 4.0464, 4.4222, 4.4970], device='cuda:0'), covar=tensor([0.1763, 0.1224, 0.1360, 0.0734, 0.0548, 0.1427, 0.0759, 0.0705], device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0778, 0.0899, 0.0787, 0.0597, 0.0620, 0.0649, 0.0748], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:01:07,958 INFO [train.py:904] (0/8) Epoch 22, batch 7200, loss[loss=0.1717, simple_loss=0.2593, pruned_loss=0.04206, over 17080.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2894, pruned_loss=0.0576, over 3098450.94 frames. ], batch size: 53, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:28,587 INFO [train.py:904] (0/8) Epoch 22, batch 7250, loss[loss=0.1574, simple_loss=0.2482, pruned_loss=0.03328, over 16699.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2868, pruned_loss=0.056, over 3107120.57 frames. ], batch size: 89, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:02:30,896 INFO [optim.py:368] (0/8) 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 12:02:53,048 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220419.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:30,268 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220443.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:45,049 INFO [train.py:904] (0/8) Epoch 22, batch 7300, loss[loss=0.1923, simple_loss=0.2899, pruned_loss=0.04733, over 16701.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2863, pruned_loss=0.05594, over 3104282.54 frames. ], batch size: 76, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:03:46,869 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220454.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:03:52,203 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1981, 4.2738, 4.0676, 3.7972, 3.8221, 4.2031, 3.8842, 3.9609], device='cuda:0'), covar=tensor([0.0575, 0.0430, 0.0284, 0.0264, 0.0707, 0.0425, 0.0872, 0.0601], device='cuda:0'), in_proj_covar=tensor([0.0287, 0.0424, 0.0337, 0.0334, 0.0343, 0.0388, 0.0232, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:04:07,577 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:04:08,876 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6469, 4.5517, 4.7029, 4.8579, 5.0102, 4.4780, 4.9869, 5.0022], device='cuda:0'), covar=tensor([0.1646, 0.1046, 0.1359, 0.0643, 0.0474, 0.1023, 0.0535, 0.0555], device='cuda:0'), in_proj_covar=tensor([0.0620, 0.0767, 0.0888, 0.0777, 0.0589, 0.0613, 0.0641, 0.0739], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:05:02,421 INFO [train.py:904] (0/8) Epoch 22, batch 7350, loss[loss=0.2238, simple_loss=0.3027, pruned_loss=0.07247, over 15386.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.287, pruned_loss=0.05704, over 3068947.55 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:05:04,311 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5856, 2.1288, 1.7315, 1.9311, 2.4520, 2.0956, 2.3330, 2.6088], device='cuda:0'), covar=tensor([0.0189, 0.0409, 0.0569, 0.0474, 0.0253, 0.0411, 0.0209, 0.0271], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0230, 0.0223, 0.0222, 0.0231, 0.0230, 0.0230, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:05:05,572 INFO [optim.py:368] (0/8) 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 12:05:17,908 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220513.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:14,243 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-01 12:06:20,489 INFO [train.py:904] (0/8) Epoch 22, batch 7400, loss[loss=0.1967, simple_loss=0.287, pruned_loss=0.05319, over 16705.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2878, pruned_loss=0.0575, over 3078890.55 frames. ], batch size: 76, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:06:24,816 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:35,404 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:06:47,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6648, 3.0636, 3.1529, 1.9333, 2.8382, 2.1028, 3.2624, 3.2744], device='cuda:0'), covar=tensor([0.0286, 0.0841, 0.0626, 0.2142, 0.0877, 0.1102, 0.0647, 0.0979], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0165, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 12:07:04,985 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8654, 2.3348, 1.8622, 2.1373, 2.7022, 2.3437, 2.5790, 2.8636], device='cuda:0'), covar=tensor([0.0190, 0.0432, 0.0618, 0.0459, 0.0247, 0.0393, 0.0205, 0.0262], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0231, 0.0224, 0.0222, 0.0232, 0.0231, 0.0231, 0.0226], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:07:13,082 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220585.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:07:13,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6417, 1.8271, 2.2763, 2.5787, 2.5541, 2.9831, 1.9897, 2.9720], device='cuda:0'), covar=tensor([0.0247, 0.0534, 0.0350, 0.0338, 0.0339, 0.0203, 0.0530, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0192, 0.0178, 0.0182, 0.0196, 0.0152, 0.0195, 0.0150], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:07:41,488 INFO [train.py:904] (0/8) Epoch 22, batch 7450, loss[loss=0.2274, simple_loss=0.3208, pruned_loss=0.067, over 16423.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2892, pruned_loss=0.05874, over 3076400.31 frames. ], batch size: 146, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:07:43,944 INFO [optim.py:368] (0/8) 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 12:07:45,848 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5252, 3.4637, 3.4558, 2.7539, 3.3425, 2.1329, 3.1489, 2.9060], device='cuda:0'), covar=tensor([0.0173, 0.0155, 0.0196, 0.0243, 0.0115, 0.2220, 0.0146, 0.0238], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0154, 0.0197, 0.0177, 0.0173, 0.0207, 0.0185, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:08:01,226 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:08:04,683 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220616.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 12:08:07,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2375, 4.1041, 4.2891, 4.4828, 4.6390, 4.1518, 4.5119, 4.6545], device='cuda:0'), covar=tensor([0.1903, 0.1315, 0.1619, 0.0714, 0.0578, 0.1243, 0.0856, 0.0623], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0766, 0.0888, 0.0776, 0.0590, 0.0614, 0.0643, 0.0739], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:08:21,832 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-01 12:09:00,837 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220651.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:09:03,155 INFO [train.py:904] (0/8) Epoch 22, batch 7500, loss[loss=0.2089, simple_loss=0.2937, pruned_loss=0.062, over 15326.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2897, pruned_loss=0.05799, over 3075430.36 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:20,595 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5935, 3.6713, 2.7929, 2.2582, 2.4409, 2.3544, 3.9131, 3.2848], device='cuda:0'), covar=tensor([0.2954, 0.0654, 0.1825, 0.2739, 0.2730, 0.2090, 0.0450, 0.1282], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0268, 0.0305, 0.0314, 0.0298, 0.0260, 0.0296, 0.0336], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:10:21,155 INFO [train.py:904] (0/8) Epoch 22, batch 7550, loss[loss=0.1891, simple_loss=0.2758, pruned_loss=0.05127, over 16883.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2893, pruned_loss=0.05871, over 3070595.15 frames. ], batch size: 109, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:10:24,494 INFO [optim.py:368] (0/8) 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 12:10:35,500 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220712.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:10:47,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3735, 3.3370, 3.3987, 3.4767, 3.5079, 3.2832, 3.4563, 3.5626], device='cuda:0'), covar=tensor([0.1258, 0.0951, 0.1000, 0.0621, 0.0654, 0.2281, 0.1163, 0.0777], device='cuda:0'), in_proj_covar=tensor([0.0622, 0.0767, 0.0888, 0.0777, 0.0591, 0.0615, 0.0644, 0.0740], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:11:23,173 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220743.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:11:38,546 INFO [train.py:904] (0/8) Epoch 22, batch 7600, loss[loss=0.2051, simple_loss=0.2849, pruned_loss=0.06265, over 15168.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2891, pruned_loss=0.05961, over 3054848.04 frames. ], batch size: 190, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:11:39,353 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-01 12:11:40,742 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:29,245 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 12:12:36,982 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:55,416 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220802.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:12:56,367 INFO [train.py:904] (0/8) Epoch 22, batch 7650, loss[loss=0.184, simple_loss=0.2819, pruned_loss=0.04308, over 16847.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2889, pruned_loss=0.05923, over 3076593.63 frames. ], batch size: 96, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:12:59,172 INFO [optim.py:368] (0/8) 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 12:13:26,585 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220822.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:14:06,909 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8631, 2.7758, 2.4190, 2.7417, 3.2255, 2.8490, 3.4235, 3.4475], device='cuda:0'), covar=tensor([0.0101, 0.0408, 0.0507, 0.0363, 0.0260, 0.0397, 0.0203, 0.0227], device='cuda:0'), in_proj_covar=tensor([0.0207, 0.0233, 0.0225, 0.0223, 0.0234, 0.0231, 0.0232, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:14:11,873 INFO [train.py:904] (0/8) Epoch 22, batch 7700, loss[loss=0.1927, simple_loss=0.2795, pruned_loss=0.05295, over 16385.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2895, pruned_loss=0.05966, over 3088861.94 frames. ], batch size: 165, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:14:26,704 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6588, 1.7532, 1.6218, 1.4860, 1.8783, 1.6310, 1.5655, 1.8969], device='cuda:0'), covar=tensor([0.0182, 0.0274, 0.0374, 0.0301, 0.0193, 0.0245, 0.0179, 0.0190], device='cuda:0'), in_proj_covar=tensor([0.0207, 0.0232, 0.0225, 0.0223, 0.0233, 0.0231, 0.0232, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:14:50,483 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3939, 2.9508, 2.6370, 2.3206, 2.2326, 2.2938, 2.9840, 2.8455], device='cuda:0'), covar=tensor([0.2384, 0.0707, 0.1630, 0.2203, 0.2255, 0.2165, 0.0508, 0.1187], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0268, 0.0306, 0.0315, 0.0299, 0.0260, 0.0296, 0.0337], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:14:58,262 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220883.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:01,341 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220885.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:29,149 INFO [train.py:904] (0/8) Epoch 22, batch 7750, loss[loss=0.2404, simple_loss=0.3211, pruned_loss=0.0799, over 15515.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2897, pruned_loss=0.05942, over 3083200.39 frames. ], batch size: 191, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:15:30,774 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=220904.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:32,048 INFO [optim.py:368] (0/8) 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 12:15:40,105 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=220911.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:15:45,306 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=220914.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:15:53,569 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3393, 3.8645, 3.8445, 2.5584, 3.6216, 3.9239, 3.5795, 1.8541], device='cuda:0'), covar=tensor([0.0635, 0.0088, 0.0090, 0.0503, 0.0122, 0.0186, 0.0123, 0.0678], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0083, 0.0084, 0.0132, 0.0097, 0.0109, 0.0094, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:16:15,125 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220933.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:16:42,235 INFO [train.py:904] (0/8) Epoch 22, batch 7800, loss[loss=0.2468, simple_loss=0.3265, pruned_loss=0.08359, over 16533.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2901, pruned_loss=0.05985, over 3082834.18 frames. ], batch size: 62, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:16:54,501 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 12:16:56,569 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=220962.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:00,289 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=220965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:17:17,084 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9899, 4.0434, 4.3365, 4.3064, 4.3045, 4.0868, 4.0622, 4.0568], device='cuda:0'), covar=tensor([0.0369, 0.0641, 0.0439, 0.0444, 0.0472, 0.0463, 0.0930, 0.0527], device='cuda:0'), in_proj_covar=tensor([0.0410, 0.0456, 0.0439, 0.0408, 0.0487, 0.0464, 0.0549, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 12:17:55,733 INFO [train.py:904] (0/8) Epoch 22, batch 7850, loss[loss=0.2215, simple_loss=0.2917, pruned_loss=0.07568, over 11799.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2907, pruned_loss=0.05947, over 3082880.77 frames. ], batch size: 249, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:17:58,018 INFO [optim.py:368] (0/8) 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 12:18:02,924 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:18:15,564 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0085, 2.1123, 2.2546, 3.4910, 2.0964, 2.4323, 2.2523, 2.2339], device='cuda:0'), covar=tensor([0.1450, 0.3748, 0.2802, 0.0645, 0.4295, 0.2432, 0.3520, 0.3361], device='cuda:0'), in_proj_covar=tensor([0.0396, 0.0445, 0.0363, 0.0324, 0.0434, 0.0511, 0.0417, 0.0518], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:18:33,604 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5970, 1.7977, 2.2534, 2.5358, 2.5571, 2.9511, 1.9195, 2.8295], device='cuda:0'), covar=tensor([0.0223, 0.0523, 0.0338, 0.0330, 0.0301, 0.0175, 0.0568, 0.0157], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0190, 0.0176, 0.0181, 0.0193, 0.0151, 0.0193, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:18:50,591 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5577, 3.4211, 3.8360, 1.8927, 3.9645, 4.0174, 2.9855, 2.9840], device='cuda:0'), covar=tensor([0.0796, 0.0271, 0.0207, 0.1283, 0.0071, 0.0181, 0.0449, 0.0469], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0107, 0.0097, 0.0138, 0.0080, 0.0125, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 12:19:09,104 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 12:19:09,461 INFO [train.py:904] (0/8) Epoch 22, batch 7900, loss[loss=0.2183, simple_loss=0.3105, pruned_loss=0.06303, over 16858.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2901, pruned_loss=0.05903, over 3096052.98 frames. ], batch size: 96, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:20:27,108 INFO [train.py:904] (0/8) Epoch 22, batch 7950, loss[loss=0.2493, simple_loss=0.3162, pruned_loss=0.09116, over 11697.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2908, pruned_loss=0.05993, over 3085595.40 frames. ], batch size: 246, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:20:32,013 INFO [optim.py:368] (0/8) 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 12:21:09,409 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-01 12:21:41,837 INFO [train.py:904] (0/8) Epoch 22, batch 8000, loss[loss=0.1851, simple_loss=0.2736, pruned_loss=0.04831, over 16787.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2904, pruned_loss=0.05973, over 3105184.16 frames. ], batch size: 124, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:22:01,734 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 12:22:18,533 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:22:55,189 INFO [train.py:904] (0/8) Epoch 22, batch 8050, loss[loss=0.2109, simple_loss=0.2924, pruned_loss=0.06466, over 16928.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2906, pruned_loss=0.05978, over 3096250.61 frames. ], batch size: 109, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:23:01,255 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.760e+02 2.816e+02 3.381e+02 4.131e+02 1.220e+03, threshold=6.763e+02, percent-clipped=7.0 2023-05-01 12:23:01,823 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8272, 2.6776, 2.8453, 2.1831, 2.6881, 2.1377, 2.7217, 2.9311], device='cuda:0'), covar=tensor([0.0255, 0.0799, 0.0475, 0.1665, 0.0795, 0.0911, 0.0543, 0.0739], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0152, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 12:23:07,202 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:23:09,472 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 12:23:15,131 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7773, 3.7744, 3.8791, 3.6724, 3.8439, 4.2506, 3.8972, 3.5766], device='cuda:0'), covar=tensor([0.1952, 0.2114, 0.2314, 0.2623, 0.2573, 0.1677, 0.1683, 0.2703], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0590, 0.0654, 0.0491, 0.0653, 0.0683, 0.0515, 0.0659], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:23:20,296 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7851, 2.8657, 2.5464, 4.6870, 3.2792, 4.0893, 1.6369, 2.9564], device='cuda:0'), covar=tensor([0.1452, 0.0886, 0.1358, 0.0225, 0.0442, 0.0475, 0.1853, 0.0932], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0174, 0.0194, 0.0190, 0.0207, 0.0215, 0.0202, 0.0192], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 12:24:08,793 INFO [train.py:904] (0/8) Epoch 22, batch 8100, loss[loss=0.2059, simple_loss=0.2999, pruned_loss=0.0559, over 16814.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2906, pruned_loss=0.05991, over 3066579.53 frames. ], batch size: 102, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:24:17,307 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:24:19,329 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221260.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:25:21,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5375, 2.6043, 2.5490, 4.3367, 2.4487, 2.8651, 2.6133, 2.7445], device='cuda:0'), covar=tensor([0.1209, 0.3372, 0.2710, 0.0461, 0.3837, 0.2428, 0.3228, 0.3014], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0448, 0.0366, 0.0326, 0.0438, 0.0517, 0.0420, 0.0523], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:25:24,337 INFO [train.py:904] (0/8) Epoch 22, batch 8150, loss[loss=0.1773, simple_loss=0.2784, pruned_loss=0.03807, over 16739.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2874, pruned_loss=0.05804, over 3091016.34 frames. ], batch size: 83, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:25:31,011 INFO [optim.py:368] (0/8) 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 12:25:31,405 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221307.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:25:39,033 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0342, 3.0375, 2.6228, 2.9591, 3.4253, 3.0658, 3.6162, 3.6023], device='cuda:0'), covar=tensor([0.0100, 0.0377, 0.0486, 0.0366, 0.0249, 0.0349, 0.0258, 0.0219], device='cuda:0'), in_proj_covar=tensor([0.0206, 0.0230, 0.0223, 0.0222, 0.0231, 0.0229, 0.0231, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:26:41,093 INFO [train.py:904] (0/8) Epoch 22, batch 8200, loss[loss=0.1878, simple_loss=0.2811, pruned_loss=0.0472, over 16125.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2853, pruned_loss=0.05799, over 3072221.93 frames. ], batch size: 165, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:26:44,118 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221355.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:27:33,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9876, 2.7593, 2.8195, 2.1171, 2.5809, 2.0597, 2.8590, 2.9486], device='cuda:0'), covar=tensor([0.0300, 0.0962, 0.0586, 0.2063, 0.1055, 0.1170, 0.0645, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0163, 0.0165, 0.0153, 0.0144, 0.0129, 0.0142, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 12:27:58,682 INFO [train.py:904] (0/8) Epoch 22, batch 8250, loss[loss=0.1737, simple_loss=0.2593, pruned_loss=0.04404, over 12140.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.284, pruned_loss=0.05546, over 3057429.76 frames. ], batch size: 246, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:28:05,607 INFO [optim.py:368] (0/8) 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 12:29:17,818 INFO [train.py:904] (0/8) Epoch 22, batch 8300, loss[loss=0.1819, simple_loss=0.264, pruned_loss=0.04993, over 11816.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.281, pruned_loss=0.05223, over 3030429.76 frames. ], batch size: 248, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:29:55,544 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5005, 3.3521, 2.7545, 2.1578, 2.0975, 2.2973, 3.4836, 3.0239], device='cuda:0'), covar=tensor([0.2841, 0.0688, 0.1717, 0.2974, 0.2983, 0.2278, 0.0431, 0.1344], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0266, 0.0302, 0.0312, 0.0295, 0.0258, 0.0293, 0.0333], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:29:57,526 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221478.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:30:05,700 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-01 12:30:12,941 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4169, 4.3869, 4.7461, 4.7147, 4.7243, 4.5122, 4.4313, 4.3726], device='cuda:0'), covar=tensor([0.0337, 0.0733, 0.0434, 0.0428, 0.0457, 0.0446, 0.0947, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0456, 0.0439, 0.0407, 0.0486, 0.0462, 0.0547, 0.0370], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 12:30:38,306 INFO [train.py:904] (0/8) Epoch 22, batch 8350, loss[loss=0.2171, simple_loss=0.2932, pruned_loss=0.0705, over 11809.00 frames. ], tot_loss[loss=0.1906, simple_loss=0.2804, pruned_loss=0.05038, over 3031794.52 frames. ], batch size: 247, lr: 3.05e-03, grad_scale: 4.0 2023-05-01 12:30:43,701 INFO [optim.py:368] (0/8) 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 12:30:44,238 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:30:57,274 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 12:31:03,771 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 12:31:15,679 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221526.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:31:56,143 INFO [train.py:904] (0/8) Epoch 22, batch 8400, loss[loss=0.1573, simple_loss=0.2542, pruned_loss=0.03019, over 16908.00 frames. ], tot_loss[loss=0.1874, simple_loss=0.2778, pruned_loss=0.04849, over 3011851.13 frames. ], batch size: 90, lr: 3.05e-03, grad_scale: 8.0 2023-05-01 12:32:08,958 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=221560.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:32:21,089 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:33:17,381 INFO [train.py:904] (0/8) Epoch 22, batch 8450, loss[loss=0.1764, simple_loss=0.2638, pruned_loss=0.04452, over 12422.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2762, pruned_loss=0.04657, over 3035904.55 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:33:24,325 INFO [optim.py:368] (0/8) 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 12:33:26,152 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=221608.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:33:36,984 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8082, 3.8183, 3.9897, 3.7291, 3.9258, 4.3315, 3.9214, 3.5994], device='cuda:0'), covar=tensor([0.2185, 0.2363, 0.2102, 0.2421, 0.2524, 0.1411, 0.1657, 0.2589], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0577, 0.0638, 0.0478, 0.0636, 0.0670, 0.0505, 0.0643], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:34:38,843 INFO [train.py:904] (0/8) Epoch 22, batch 8500, loss[loss=0.1648, simple_loss=0.2457, pruned_loss=0.04192, over 12261.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2723, pruned_loss=0.0445, over 3024950.08 frames. ], batch size: 248, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:02,542 INFO [train.py:904] (0/8) Epoch 22, batch 8550, loss[loss=0.1789, simple_loss=0.2659, pruned_loss=0.04591, over 11991.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2707, pruned_loss=0.04386, over 3002159.59 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:36:07,085 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6074, 2.5452, 1.8869, 2.7484, 2.0642, 2.7530, 2.1288, 2.3635], device='cuda:0'), covar=tensor([0.0281, 0.0307, 0.1251, 0.0273, 0.0609, 0.0445, 0.1161, 0.0548], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0170, 0.0188, 0.0156, 0.0171, 0.0209, 0.0197, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-01 12:36:10,061 INFO [optim.py:368] (0/8) 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 12:36:13,618 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-01 12:37:41,433 INFO [train.py:904] (0/8) Epoch 22, batch 8600, loss[loss=0.1755, simple_loss=0.2787, pruned_loss=0.03619, over 17053.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2703, pruned_loss=0.04258, over 3003778.47 frames. ], batch size: 53, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:37:53,154 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221758.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:39:21,621 INFO [train.py:904] (0/8) Epoch 22, batch 8650, loss[loss=0.1686, simple_loss=0.2678, pruned_loss=0.03468, over 16206.00 frames. ], tot_loss[loss=0.1753, simple_loss=0.2685, pruned_loss=0.04104, over 3018900.14 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:39:30,559 INFO [optim.py:368] (0/8) 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 12:39:58,388 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221819.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:40:53,739 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9649, 4.3018, 3.2576, 2.3935, 2.6818, 2.6487, 4.5904, 3.6163], device='cuda:0'), covar=tensor([0.2526, 0.0510, 0.1646, 0.2803, 0.2731, 0.2009, 0.0343, 0.1223], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0261, 0.0298, 0.0307, 0.0289, 0.0255, 0.0289, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:41:06,673 INFO [train.py:904] (0/8) Epoch 22, batch 8700, loss[loss=0.1422, simple_loss=0.2353, pruned_loss=0.02456, over 16689.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2657, pruned_loss=0.03985, over 3029715.65 frames. ], batch size: 83, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:41:27,677 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=221863.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:41:51,232 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2741, 4.2325, 4.0894, 3.3579, 4.1688, 1.6829, 3.9405, 3.7697], device='cuda:0'), covar=tensor([0.0103, 0.0107, 0.0179, 0.0290, 0.0110, 0.2980, 0.0137, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0152, 0.0193, 0.0172, 0.0171, 0.0204, 0.0183, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:42:16,261 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=221890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:42:42,875 INFO [train.py:904] (0/8) Epoch 22, batch 8750, loss[loss=0.1879, simple_loss=0.2847, pruned_loss=0.04549, over 16720.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2662, pruned_loss=0.03958, over 3034195.62 frames. ], batch size: 124, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:42:49,212 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6436, 3.7973, 2.9296, 2.2051, 2.4162, 2.4782, 4.0826, 3.3301], device='cuda:0'), covar=tensor([0.3043, 0.0678, 0.1833, 0.3063, 0.2975, 0.2129, 0.0459, 0.1346], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0260, 0.0298, 0.0306, 0.0288, 0.0254, 0.0289, 0.0327], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 12:42:53,176 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.058e+02 2.561e+02 3.126e+02 5.404e+02, threshold=5.122e+02, percent-clipped=2.0 2023-05-01 12:43:29,312 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 12:44:31,219 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=221951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:44:34,628 INFO [train.py:904] (0/8) Epoch 22, batch 8800, loss[loss=0.1789, simple_loss=0.2801, pruned_loss=0.03891, over 16844.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2646, pruned_loss=0.03852, over 3047151.59 frames. ], batch size: 102, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:45:54,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4933, 3.4480, 3.5553, 3.5995, 3.6433, 3.3537, 3.6519, 3.7109], device='cuda:0'), covar=tensor([0.1221, 0.0928, 0.0921, 0.0550, 0.0576, 0.2209, 0.0707, 0.0658], device='cuda:0'), in_proj_covar=tensor([0.0614, 0.0759, 0.0878, 0.0766, 0.0585, 0.0611, 0.0635, 0.0736], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:46:13,617 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-222000.pt 2023-05-01 12:46:22,305 INFO [train.py:904] (0/8) Epoch 22, batch 8850, loss[loss=0.1729, simple_loss=0.275, pruned_loss=0.03542, over 15361.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2671, pruned_loss=0.03807, over 3034611.83 frames. ], batch size: 192, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:46:28,905 INFO [optim.py:368] (0/8) 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 12:46:35,320 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:48:07,873 INFO [train.py:904] (0/8) Epoch 22, batch 8900, loss[loss=0.1896, simple_loss=0.2838, pruned_loss=0.04771, over 16574.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2669, pruned_loss=0.03723, over 3043757.64 frames. ], batch size: 75, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:48:32,999 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 12:48:42,831 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222070.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:50:11,684 INFO [train.py:904] (0/8) Epoch 22, batch 8950, loss[loss=0.1621, simple_loss=0.2574, pruned_loss=0.03345, over 16250.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2664, pruned_loss=0.03775, over 3046005.62 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:50:23,610 INFO [optim.py:368] (0/8) 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 12:50:37,682 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222114.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:51:38,663 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9759, 2.2756, 2.3038, 2.9220, 1.8376, 3.2509, 1.7579, 2.7547], device='cuda:0'), covar=tensor([0.1184, 0.0694, 0.1077, 0.0155, 0.0079, 0.0355, 0.1513, 0.0658], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0173, 0.0193, 0.0185, 0.0202, 0.0212, 0.0201, 0.0191], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 12:52:03,808 INFO [train.py:904] (0/8) Epoch 22, batch 9000, loss[loss=0.1795, simple_loss=0.2665, pruned_loss=0.04629, over 16792.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2637, pruned_loss=0.03653, over 3061913.57 frames. ], batch size: 124, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:52:03,810 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 12:52:14,711 INFO [train.py:938] (0/8) Epoch 22, validation: loss=0.1453, simple_loss=0.2492, pruned_loss=0.0207, over 944034.00 frames. 2023-05-01 12:52:14,712 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 12:52:36,447 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222163.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:52:48,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0485, 3.2909, 3.4522, 1.6743, 3.6062, 3.7951, 2.9757, 2.6960], device='cuda:0'), covar=tensor([0.1125, 0.0205, 0.0187, 0.1378, 0.0095, 0.0151, 0.0415, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0104, 0.0093, 0.0134, 0.0077, 0.0119, 0.0124, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 12:52:58,410 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1197, 5.1417, 4.9600, 4.5427, 4.6262, 5.0390, 4.9129, 4.7050], device='cuda:0'), covar=tensor([0.0678, 0.0611, 0.0350, 0.0350, 0.1118, 0.0613, 0.0306, 0.0771], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0412, 0.0331, 0.0326, 0.0333, 0.0380, 0.0227, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 12:53:22,214 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222185.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:53:28,185 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.62 vs. limit=2.0 2023-05-01 12:53:58,636 INFO [train.py:904] (0/8) Epoch 22, batch 9050, loss[loss=0.1474, simple_loss=0.2392, pruned_loss=0.02783, over 16925.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2649, pruned_loss=0.03714, over 3080418.80 frames. ], batch size: 102, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:54:03,994 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222205.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:54:09,044 INFO [optim.py:368] (0/8) 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 12:54:17,179 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,443 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:55:31,587 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222246.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 12:55:44,666 INFO [train.py:904] (0/8) Epoch 22, batch 9100, loss[loss=0.1716, simple_loss=0.2752, pruned_loss=0.03404, over 16191.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2647, pruned_loss=0.03766, over 3079944.35 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:56:10,979 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 12:57:42,766 INFO [train.py:904] (0/8) Epoch 22, batch 9150, loss[loss=0.1705, simple_loss=0.2575, pruned_loss=0.0417, over 12017.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2654, pruned_loss=0.03734, over 3082587.08 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 12:57:53,765 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.620e+02 2.177e+02 2.613e+02 3.280e+02 4.905e+02, threshold=5.227e+02, percent-clipped=0.0 2023-05-01 12:59:28,948 INFO [train.py:904] (0/8) Epoch 22, batch 9200, loss[loss=0.1517, simple_loss=0.2518, pruned_loss=0.02583, over 16591.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2611, pruned_loss=0.03636, over 3067014.54 frames. ], batch size: 75, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 12:59:52,293 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222365.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:01:04,229 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6575, 3.7242, 3.4660, 3.1396, 3.3261, 3.6556, 3.3635, 3.4606], device='cuda:0'), covar=tensor([0.0586, 0.0729, 0.0287, 0.0240, 0.0479, 0.0545, 0.1557, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0410, 0.0329, 0.0323, 0.0331, 0.0377, 0.0226, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:01:05,921 INFO [train.py:904] (0/8) Epoch 22, batch 9250, loss[loss=0.1517, simple_loss=0.2369, pruned_loss=0.03324, over 12883.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2608, pruned_loss=0.03675, over 3051370.55 frames. ], batch size: 247, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:01:16,250 INFO [optim.py:368] (0/8) 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 13:01:28,880 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:02:56,977 INFO [train.py:904] (0/8) Epoch 22, batch 9300, loss[loss=0.1479, simple_loss=0.2434, pruned_loss=0.02626, over 16368.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2593, pruned_loss=0.03627, over 3051212.99 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:03:16,992 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:03:40,390 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4901, 4.5709, 4.3776, 4.0611, 4.0904, 4.5004, 4.2298, 4.2027], device='cuda:0'), covar=tensor([0.0569, 0.0599, 0.0335, 0.0310, 0.0852, 0.0531, 0.0536, 0.0640], device='cuda:0'), in_proj_covar=tensor([0.0281, 0.0410, 0.0328, 0.0323, 0.0331, 0.0377, 0.0226, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:04:34,465 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8955, 2.8300, 2.7247, 2.0068, 2.5612, 2.8333, 2.7157, 1.9391], device='cuda:0'), covar=tensor([0.0440, 0.0067, 0.0073, 0.0358, 0.0122, 0.0101, 0.0096, 0.0451], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0081, 0.0082, 0.0131, 0.0095, 0.0106, 0.0092, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 13:04:40,784 INFO [train.py:904] (0/8) Epoch 22, batch 9350, loss[loss=0.1668, simple_loss=0.2584, pruned_loss=0.03756, over 16877.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2588, pruned_loss=0.03601, over 3045425.97 frames. ], batch size: 116, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:04:49,918 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.318e+02 1.994e+02 2.436e+02 3.069e+02 5.740e+02, threshold=4.871e+02, percent-clipped=2.0 2023-05-01 13:05:01,812 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.29 vs. limit=5.0 2023-05-01 13:05:23,321 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:05:57,379 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222541.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:06:08,613 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:20,191 INFO [train.py:904] (0/8) Epoch 22, batch 9400, loss[loss=0.1844, simple_loss=0.2849, pruned_loss=0.04199, over 16893.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2583, pruned_loss=0.03554, over 3038265.39 frames. ], batch size: 116, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:06:38,155 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:06:55,321 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222570.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:07:24,548 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222584.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:07:43,448 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:08:00,591 INFO [train.py:904] (0/8) Epoch 22, batch 9450, loss[loss=0.1625, simple_loss=0.2577, pruned_loss=0.03367, over 16169.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2607, pruned_loss=0.0359, over 3054966.19 frames. ], batch size: 165, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:08:08,419 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 2.238e+02 2.730e+02 3.189e+02 9.782e+02, threshold=5.460e+02, percent-clipped=5.0 2023-05-01 13:08:56,874 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=222631.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:09:40,759 INFO [train.py:904] (0/8) Epoch 22, batch 9500, loss[loss=0.159, simple_loss=0.2528, pruned_loss=0.03255, over 16576.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2601, pruned_loss=0.03537, over 3074152.19 frames. ], batch size: 62, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:10:06,968 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:11:18,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1124, 3.4995, 3.4729, 2.3508, 3.1811, 3.5208, 3.3411, 2.0256], device='cuda:0'), covar=tensor([0.0537, 0.0052, 0.0059, 0.0387, 0.0112, 0.0091, 0.0083, 0.0504], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0080, 0.0081, 0.0130, 0.0095, 0.0106, 0.0091, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 13:11:22,406 INFO [train.py:904] (0/8) Epoch 22, batch 9550, loss[loss=0.1824, simple_loss=0.2788, pruned_loss=0.043, over 16359.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2601, pruned_loss=0.03534, over 3098943.47 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:11:26,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6114, 3.6786, 3.4634, 3.0995, 3.2980, 3.5816, 3.3714, 3.4160], device='cuda:0'), covar=tensor([0.0600, 0.0624, 0.0303, 0.0276, 0.0519, 0.0515, 0.1321, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0279, 0.0406, 0.0326, 0.0321, 0.0330, 0.0375, 0.0224, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002], device='cuda:0') 2023-05-01 13:11:34,511 INFO [optim.py:368] (0/8) 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 13:11:44,580 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222713.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:13:00,677 INFO [train.py:904] (0/8) Epoch 22, batch 9600, loss[loss=0.1794, simple_loss=0.276, pruned_loss=0.04141, over 16790.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2613, pruned_loss=0.03585, over 3098852.55 frames. ], batch size: 83, lr: 3.04e-03, grad_scale: 8.0 2023-05-01 13:13:47,354 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9302, 2.7697, 2.8871, 2.1194, 2.6618, 2.1915, 2.6696, 2.8900], device='cuda:0'), covar=tensor([0.0325, 0.0922, 0.0572, 0.1888, 0.0892, 0.0934, 0.0704, 0.0857], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0158, 0.0163, 0.0151, 0.0142, 0.0127, 0.0139, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 13:14:44,343 INFO [train.py:904] (0/8) Epoch 22, batch 9650, loss[loss=0.1722, simple_loss=0.2748, pruned_loss=0.03477, over 15272.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2635, pruned_loss=0.03641, over 3074878.87 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:14:58,750 INFO [optim.py:368] (0/8) 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 13:16:03,749 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222841.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:16:27,400 INFO [train.py:904] (0/8) Epoch 22, batch 9700, loss[loss=0.1646, simple_loss=0.2493, pruned_loss=0.03992, over 12422.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2621, pruned_loss=0.03622, over 3053520.00 frames. ], batch size: 250, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:16:43,292 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=222861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:17:22,890 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222879.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:17:42,992 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:03,162 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8010, 1.3920, 1.6983, 1.6990, 1.8809, 1.8897, 1.6813, 1.7959], device='cuda:0'), covar=tensor([0.0275, 0.0420, 0.0231, 0.0333, 0.0318, 0.0234, 0.0488, 0.0139], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0187, 0.0175, 0.0178, 0.0192, 0.0148, 0.0191, 0.0146], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:18:08,548 INFO [train.py:904] (0/8) Epoch 22, batch 9750, loss[loss=0.1705, simple_loss=0.2671, pruned_loss=0.03692, over 16400.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2608, pruned_loss=0.03632, over 3046887.93 frames. ], batch size: 146, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:18:18,157 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=222909.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:18:52,121 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=222926.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:06,264 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6737, 4.8651, 4.9808, 4.7570, 4.8180, 5.3569, 4.8554, 4.5559], device='cuda:0'), covar=tensor([0.1079, 0.2057, 0.2429, 0.2078, 0.2360, 0.0893, 0.1605, 0.2364], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0561, 0.0623, 0.0463, 0.0620, 0.0647, 0.0490, 0.0622], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 13:19:24,084 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222940.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:19:45,660 INFO [train.py:904] (0/8) Epoch 22, batch 9800, loss[loss=0.1596, simple_loss=0.2602, pruned_loss=0.02954, over 16810.00 frames. ], tot_loss[loss=0.166, simple_loss=0.261, pruned_loss=0.03551, over 3066953.27 frames. ], batch size: 124, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:20:15,220 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222969.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:20:40,798 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=222984.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:20:43,290 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 13:21:23,030 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223001.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:21:26,500 INFO [train.py:904] (0/8) Epoch 22, batch 9850, loss[loss=0.1697, simple_loss=0.2768, pruned_loss=0.03129, over 15408.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2622, pruned_loss=0.03542, over 3069447.66 frames. ], batch size: 191, lr: 3.04e-03, grad_scale: 4.0 2023-05-01 13:21:28,319 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 13:21:37,450 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.461e+02 2.181e+02 2.565e+02 3.075e+02 7.070e+02, threshold=5.130e+02, percent-clipped=2.0 2023-05-01 13:22:19,845 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223030.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:23:01,463 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223045.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:23:08,460 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 13:23:15,943 INFO [train.py:904] (0/8) Epoch 22, batch 9900, loss[loss=0.1855, simple_loss=0.2842, pruned_loss=0.04347, over 16739.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2625, pruned_loss=0.03527, over 3070076.41 frames. ], batch size: 134, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:23:43,612 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:25:13,521 INFO [train.py:904] (0/8) Epoch 22, batch 9950, loss[loss=0.1805, simple_loss=0.2808, pruned_loss=0.04009, over 15372.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2653, pruned_loss=0.0358, over 3092035.61 frames. ], batch size: 190, lr: 3.03e-03, grad_scale: 4.0 2023-05-01 13:25:27,894 INFO [optim.py:368] (0/8) 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 13:25:39,791 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 13:26:07,003 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223125.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:26:44,831 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0386, 2.1190, 2.2597, 3.4916, 2.1089, 2.3544, 2.2325, 2.2169], device='cuda:0'), covar=tensor([0.1305, 0.3624, 0.2958, 0.0618, 0.4245, 0.2715, 0.3738, 0.3467], device='cuda:0'), in_proj_covar=tensor([0.0391, 0.0439, 0.0362, 0.0318, 0.0430, 0.0502, 0.0410, 0.0511], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:26:59,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1048, 3.6105, 3.5897, 2.3214, 3.2902, 3.5985, 3.3983, 2.1805], device='cuda:0'), covar=tensor([0.0536, 0.0048, 0.0048, 0.0391, 0.0095, 0.0079, 0.0067, 0.0458], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0081, 0.0082, 0.0131, 0.0096, 0.0106, 0.0092, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 13:27:13,212 INFO [train.py:904] (0/8) Epoch 22, batch 10000, loss[loss=0.1832, simple_loss=0.2852, pruned_loss=0.04054, over 16802.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2636, pruned_loss=0.03498, over 3115167.16 frames. ], batch size: 124, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:27:32,114 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4525, 2.0317, 1.7506, 1.8036, 2.3140, 2.0324, 1.8701, 2.3823], device='cuda:0'), covar=tensor([0.0181, 0.0383, 0.0523, 0.0469, 0.0242, 0.0334, 0.0207, 0.0234], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0230, 0.0222, 0.0222, 0.0230, 0.0228, 0.0224, 0.0220], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:27:48,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8689, 4.8427, 4.5792, 4.0235, 4.7211, 1.8702, 4.4515, 4.4016], device='cuda:0'), covar=tensor([0.0074, 0.0069, 0.0190, 0.0291, 0.0092, 0.2714, 0.0127, 0.0240], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0151, 0.0190, 0.0167, 0.0169, 0.0203, 0.0181, 0.0164], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:27:51,714 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223173.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:28:04,773 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223179.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 13:28:54,098 INFO [train.py:904] (0/8) Epoch 22, batch 10050, loss[loss=0.1554, simple_loss=0.2548, pruned_loss=0.02793, over 16536.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2629, pruned_loss=0.03474, over 3098296.84 frames. ], batch size: 68, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:29:04,328 INFO [optim.py:368] (0/8) 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 13:29:36,396 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3711, 4.3526, 4.1461, 3.6314, 4.2669, 1.6641, 4.0441, 3.9484], device='cuda:0'), covar=tensor([0.0113, 0.0128, 0.0219, 0.0287, 0.0134, 0.2692, 0.0168, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0150, 0.0190, 0.0166, 0.0169, 0.0203, 0.0180, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:29:40,720 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:42,096 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223227.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:29:54,214 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223234.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 13:30:27,338 INFO [train.py:904] (0/8) Epoch 22, batch 10100, loss[loss=0.171, simple_loss=0.2629, pruned_loss=0.03955, over 16918.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2635, pruned_loss=0.03517, over 3106030.96 frames. ], batch size: 109, lr: 3.03e-03, grad_scale: 8.0 2023-05-01 13:31:10,812 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223274.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:31:39,305 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223296.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:31:47,510 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-22.pt 2023-05-01 13:32:13,437 INFO [train.py:904] (0/8) Epoch 23, batch 0, loss[loss=0.1712, simple_loss=0.2622, pruned_loss=0.04008, over 15958.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2622, pruned_loss=0.04008, over 15958.00 frames. ], batch size: 35, lr: 2.97e-03, grad_scale: 8.0 2023-05-01 13:32:13,438 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 13:32:20,849 INFO [train.py:938] (0/8) Epoch 23, validation: loss=0.1454, simple_loss=0.2487, pruned_loss=0.02111, over 944034.00 frames. 2023-05-01 13:32:20,850 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 13:32:28,410 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 2.627e+02 3.041e+02 3.739e+02 7.225e+02, threshold=6.083e+02, percent-clipped=7.0 2023-05-01 13:32:49,525 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:33:09,275 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:33:26,117 INFO [train.py:904] (0/8) Epoch 23, batch 50, loss[loss=0.2033, simple_loss=0.2777, pruned_loss=0.06447, over 16779.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2664, pruned_loss=0.04668, over 749944.24 frames. ], batch size: 124, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:33:46,450 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5829, 1.7563, 2.1713, 2.3915, 2.5425, 2.4145, 1.8311, 2.6483], device='cuda:0'), covar=tensor([0.0211, 0.0545, 0.0371, 0.0319, 0.0323, 0.0347, 0.0613, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0189, 0.0176, 0.0179, 0.0193, 0.0150, 0.0193, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:34:09,097 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8926, 4.5711, 3.2428, 2.4128, 2.8057, 2.6970, 4.8815, 3.6660], device='cuda:0'), covar=tensor([0.2916, 0.0552, 0.1794, 0.2914, 0.3103, 0.2154, 0.0336, 0.1451], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0261, 0.0298, 0.0306, 0.0285, 0.0254, 0.0287, 0.0327], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 13:34:32,076 INFO [train.py:904] (0/8) Epoch 23, batch 100, loss[loss=0.136, simple_loss=0.2256, pruned_loss=0.02325, over 17213.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2637, pruned_loss=0.04536, over 1317395.75 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:34:42,064 INFO [optim.py:368] (0/8) 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 13:34:55,420 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:35:38,704 INFO [train.py:904] (0/8) Epoch 23, batch 150, loss[loss=0.1847, simple_loss=0.283, pruned_loss=0.04321, over 17089.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2633, pruned_loss=0.04399, over 1762878.22 frames. ], batch size: 53, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:35:54,501 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4397, 5.3772, 5.2760, 4.7716, 4.8812, 5.3071, 5.2725, 4.8981], device='cuda:0'), covar=tensor([0.0551, 0.0459, 0.0343, 0.0365, 0.1136, 0.0436, 0.0247, 0.0756], device='cuda:0'), in_proj_covar=tensor([0.0283, 0.0414, 0.0331, 0.0326, 0.0336, 0.0381, 0.0225, 0.0394], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:36:34,295 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4572, 2.2031, 2.2930, 4.2143, 2.1922, 2.4953, 2.3164, 2.3380], device='cuda:0'), covar=tensor([0.1343, 0.4386, 0.3288, 0.0564, 0.4926, 0.3130, 0.3904, 0.4267], device='cuda:0'), in_proj_covar=tensor([0.0397, 0.0446, 0.0367, 0.0323, 0.0435, 0.0510, 0.0417, 0.0520], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:36:47,616 INFO [train.py:904] (0/8) Epoch 23, batch 200, loss[loss=0.214, simple_loss=0.2791, pruned_loss=0.07447, over 16888.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2619, pruned_loss=0.04407, over 2108997.32 frames. ], batch size: 116, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:36:51,815 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 13:36:57,908 INFO [optim.py:368] (0/8) 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 13:37:21,176 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223529.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:37:31,654 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 13:37:52,730 INFO [train.py:904] (0/8) Epoch 23, batch 250, loss[loss=0.1401, simple_loss=0.23, pruned_loss=0.02511, over 16812.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2601, pruned_loss=0.04483, over 2390956.27 frames. ], batch size: 42, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:37:53,906 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8634, 4.6335, 4.9148, 5.0701, 5.2894, 4.6580, 5.2957, 5.2838], device='cuda:0'), covar=tensor([0.2163, 0.1566, 0.1961, 0.0900, 0.0609, 0.0925, 0.0608, 0.0687], device='cuda:0'), in_proj_covar=tensor([0.0627, 0.0772, 0.0894, 0.0781, 0.0593, 0.0619, 0.0648, 0.0747], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:38:29,109 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=223578.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:38:51,100 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3782, 3.3960, 3.7155, 2.6279, 3.3814, 3.8128, 3.4648, 2.2604], device='cuda:0'), covar=tensor([0.0516, 0.0209, 0.0061, 0.0362, 0.0118, 0.0095, 0.0117, 0.0474], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0134, 0.0098, 0.0109, 0.0094, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 13:38:54,871 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223596.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:04,051 INFO [train.py:904] (0/8) Epoch 23, batch 300, loss[loss=0.1736, simple_loss=0.2485, pruned_loss=0.04932, over 16741.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2584, pruned_loss=0.04387, over 2598731.98 frames. ], batch size: 124, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:39:14,757 INFO [optim.py:368] (0/8) 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 13:39:18,746 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7447, 3.8469, 2.4518, 4.4347, 3.0039, 4.3660, 2.6984, 3.1515], device='cuda:0'), covar=tensor([0.0355, 0.0461, 0.1667, 0.0326, 0.0826, 0.0545, 0.1391, 0.0812], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0176, 0.0193, 0.0162, 0.0176, 0.0215, 0.0202, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 13:39:34,904 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:39,100 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1011, 4.8838, 5.1390, 5.3010, 5.5276, 4.8748, 5.4768, 5.5055], device='cuda:0'), covar=tensor([0.2163, 0.1335, 0.1882, 0.0815, 0.0527, 0.0776, 0.0517, 0.0566], device='cuda:0'), in_proj_covar=tensor([0.0629, 0.0774, 0.0897, 0.0783, 0.0594, 0.0621, 0.0649, 0.0749], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:39:54,130 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=223639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:39:55,311 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223640.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:01,673 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223644.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:40:13,650 INFO [train.py:904] (0/8) Epoch 23, batch 350, loss[loss=0.209, simple_loss=0.3047, pruned_loss=0.05667, over 17080.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.256, pruned_loss=0.04241, over 2759761.36 frames. ], batch size: 53, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:40:42,286 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223673.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:02,487 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223688.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:41:08,259 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 13:41:22,675 INFO [train.py:904] (0/8) Epoch 23, batch 400, loss[loss=0.1466, simple_loss=0.2326, pruned_loss=0.03027, over 17238.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2548, pruned_loss=0.0417, over 2894278.06 frames. ], batch size: 44, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:41:34,907 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.389e+02 2.103e+02 2.400e+02 2.968e+02 1.328e+03, threshold=4.800e+02, percent-clipped=3.0 2023-05-01 13:41:46,599 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223720.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:42:32,056 INFO [train.py:904] (0/8) Epoch 23, batch 450, loss[loss=0.1644, simple_loss=0.2425, pruned_loss=0.04318, over 16485.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2535, pruned_loss=0.041, over 2993872.00 frames. ], batch size: 146, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:42:52,147 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:43:40,982 INFO [train.py:904] (0/8) Epoch 23, batch 500, loss[loss=0.1719, simple_loss=0.2527, pruned_loss=0.04557, over 16748.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2523, pruned_loss=0.03999, over 3064196.13 frames. ], batch size: 89, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:43:50,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7786, 2.8243, 2.6673, 4.9002, 3.9862, 4.2781, 1.6449, 3.0509], device='cuda:0'), covar=tensor([0.1433, 0.0829, 0.1253, 0.0244, 0.0256, 0.0453, 0.1668, 0.0825], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0174, 0.0194, 0.0188, 0.0201, 0.0215, 0.0203, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 13:43:52,984 INFO [optim.py:368] (0/8) 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 13:44:03,989 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 13:44:15,121 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 13:44:16,831 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8528, 4.9397, 5.3470, 5.3402, 5.3341, 5.0153, 4.9385, 4.7735], device='cuda:0'), covar=tensor([0.0402, 0.0572, 0.0422, 0.0419, 0.0448, 0.0425, 0.0994, 0.0473], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0459, 0.0445, 0.0412, 0.0489, 0.0468, 0.0549, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 13:44:16,855 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=223829.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:44:48,209 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1405, 4.1539, 4.5035, 4.4800, 4.5134, 4.2134, 4.2342, 4.1486], device='cuda:0'), covar=tensor([0.0412, 0.0793, 0.0435, 0.0439, 0.0560, 0.0494, 0.0856, 0.0607], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0460, 0.0447, 0.0413, 0.0490, 0.0469, 0.0549, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 13:44:49,715 INFO [train.py:904] (0/8) Epoch 23, batch 550, loss[loss=0.1908, simple_loss=0.2715, pruned_loss=0.05508, over 16703.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2515, pruned_loss=0.04017, over 3122522.60 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:45:24,267 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=223877.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 13:45:59,118 INFO [train.py:904] (0/8) Epoch 23, batch 600, loss[loss=0.1564, simple_loss=0.2498, pruned_loss=0.03147, over 17231.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2515, pruned_loss=0.04048, over 3170173.16 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:46:10,994 INFO [optim.py:368] (0/8) 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 13:46:27,577 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 13:46:34,411 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8446, 5.1760, 4.9825, 4.9336, 4.6728, 4.7258, 4.6804, 5.2964], device='cuda:0'), covar=tensor([0.1276, 0.1100, 0.1128, 0.0939, 0.0965, 0.0999, 0.1252, 0.0996], device='cuda:0'), in_proj_covar=tensor([0.0678, 0.0826, 0.0678, 0.0625, 0.0519, 0.0532, 0.0695, 0.0644], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 13:46:41,952 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=223934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:46:54,630 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8888, 4.0039, 2.9897, 2.3550, 2.5334, 2.4843, 4.0760, 3.3588], device='cuda:0'), covar=tensor([0.2564, 0.0547, 0.1726, 0.2919, 0.2752, 0.2101, 0.0467, 0.1620], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0268, 0.0305, 0.0313, 0.0294, 0.0261, 0.0295, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 13:47:08,249 INFO [train.py:904] (0/8) Epoch 23, batch 650, loss[loss=0.1546, simple_loss=0.2363, pruned_loss=0.0365, over 15998.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2496, pruned_loss=0.04003, over 3204415.25 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:47:39,199 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8506, 4.4138, 4.4225, 3.2044, 3.6750, 4.4033, 3.9753, 2.6467], device='cuda:0'), covar=tensor([0.0493, 0.0062, 0.0041, 0.0344, 0.0137, 0.0087, 0.0085, 0.0451], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0084, 0.0134, 0.0099, 0.0109, 0.0095, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 13:47:52,630 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8634, 4.3405, 4.3612, 3.0959, 3.6086, 4.3299, 3.9229, 2.6312], device='cuda:0'), covar=tensor([0.0480, 0.0073, 0.0051, 0.0348, 0.0144, 0.0094, 0.0092, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0085, 0.0134, 0.0099, 0.0109, 0.0095, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 13:48:14,715 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-224000.pt 2023-05-01 13:48:22,028 INFO [train.py:904] (0/8) Epoch 23, batch 700, loss[loss=0.179, simple_loss=0.2681, pruned_loss=0.04492, over 17017.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2496, pruned_loss=0.03941, over 3236295.49 frames. ], batch size: 50, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:48:35,478 INFO [optim.py:368] (0/8) 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 13:49:33,340 INFO [train.py:904] (0/8) Epoch 23, batch 750, loss[loss=0.1672, simple_loss=0.2621, pruned_loss=0.03613, over 17015.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2505, pruned_loss=0.03929, over 3267201.17 frames. ], batch size: 53, lr: 2.96e-03, grad_scale: 2.0 2023-05-01 13:50:39,416 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9231, 2.7279, 2.8710, 2.1806, 2.6866, 2.1511, 2.7589, 2.8610], device='cuda:0'), covar=tensor([0.0315, 0.0862, 0.0489, 0.1794, 0.0801, 0.0901, 0.0608, 0.0858], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0154, 0.0145, 0.0130, 0.0143, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 13:50:41,992 INFO [train.py:904] (0/8) Epoch 23, batch 800, loss[loss=0.1378, simple_loss=0.2252, pruned_loss=0.02524, over 16801.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2503, pruned_loss=0.03887, over 3273816.99 frames. ], batch size: 39, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:50:54,819 INFO [optim.py:368] (0/8) 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 13:51:04,688 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0672, 2.4830, 2.6462, 1.8846, 2.7567, 2.7953, 2.4713, 2.4020], device='cuda:0'), covar=tensor([0.0769, 0.0276, 0.0265, 0.1101, 0.0153, 0.0285, 0.0474, 0.0499], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0108, 0.0098, 0.0140, 0.0081, 0.0125, 0.0128, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 13:51:51,767 INFO [train.py:904] (0/8) Epoch 23, batch 850, loss[loss=0.1516, simple_loss=0.2488, pruned_loss=0.02715, over 17185.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2492, pruned_loss=0.03832, over 3288128.90 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:00,738 INFO [train.py:904] (0/8) Epoch 23, batch 900, loss[loss=0.1607, simple_loss=0.2519, pruned_loss=0.03474, over 17253.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2484, pruned_loss=0.03759, over 3298301.59 frames. ], batch size: 52, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:53:14,895 INFO [optim.py:368] (0/8) 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 13:53:15,225 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224212.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:53:46,105 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224234.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:10,201 INFO [train.py:904] (0/8) Epoch 23, batch 950, loss[loss=0.1514, simple_loss=0.2285, pruned_loss=0.03717, over 15969.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.248, pruned_loss=0.0376, over 3308263.18 frames. ], batch size: 35, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:54:38,597 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224273.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:54:50,481 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 13:55:20,391 INFO [train.py:904] (0/8) Epoch 23, batch 1000, loss[loss=0.1584, simple_loss=0.233, pruned_loss=0.04186, over 16248.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2479, pruned_loss=0.03769, over 3315180.74 frames. ], batch size: 165, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:55:33,540 INFO [optim.py:368] (0/8) 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 13:55:34,068 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7407, 3.9076, 2.5596, 4.5438, 2.9601, 4.4963, 2.6342, 3.1504], device='cuda:0'), covar=tensor([0.0329, 0.0453, 0.1592, 0.0382, 0.0936, 0.0587, 0.1532, 0.0811], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0180, 0.0196, 0.0168, 0.0179, 0.0220, 0.0206, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 13:56:31,386 INFO [train.py:904] (0/8) Epoch 23, batch 1050, loss[loss=0.1477, simple_loss=0.2342, pruned_loss=0.03063, over 17212.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2474, pruned_loss=0.03742, over 3325168.35 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:22,259 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6159, 3.6758, 4.1924, 2.4778, 3.2605, 2.7164, 4.0705, 3.8846], device='cuda:0'), covar=tensor([0.0262, 0.0925, 0.0488, 0.1820, 0.0834, 0.0949, 0.0548, 0.1171], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0166, 0.0153, 0.0145, 0.0129, 0.0142, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 13:57:42,159 INFO [train.py:904] (0/8) Epoch 23, batch 1100, loss[loss=0.1678, simple_loss=0.2498, pruned_loss=0.04287, over 17218.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2469, pruned_loss=0.03759, over 3332996.52 frames. ], batch size: 45, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:57:54,082 INFO [optim.py:368] (0/8) 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 13:58:12,345 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-01 13:58:20,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7662, 4.8108, 5.0902, 5.1163, 5.1467, 4.8467, 4.7983, 4.6331], device='cuda:0'), covar=tensor([0.0336, 0.0654, 0.0532, 0.0457, 0.0473, 0.0452, 0.0855, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0466, 0.0454, 0.0419, 0.0498, 0.0475, 0.0557, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 13:58:51,572 INFO [train.py:904] (0/8) Epoch 23, batch 1150, loss[loss=0.1493, simple_loss=0.2438, pruned_loss=0.02739, over 17170.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2469, pruned_loss=0.03752, over 3326861.83 frames. ], batch size: 46, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 13:59:32,949 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-01 14:00:00,971 INFO [train.py:904] (0/8) Epoch 23, batch 1200, loss[loss=0.1544, simple_loss=0.2326, pruned_loss=0.03808, over 16762.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2462, pruned_loss=0.03711, over 3319086.55 frames. ], batch size: 134, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:00:14,525 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.279e+02 2.643e+02 3.361e+02 1.197e+03, threshold=5.285e+02, percent-clipped=8.0 2023-05-01 14:00:37,430 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9451, 3.7651, 4.2899, 2.1335, 4.4687, 4.5159, 3.2932, 3.4620], device='cuda:0'), covar=tensor([0.0746, 0.0242, 0.0204, 0.1208, 0.0084, 0.0196, 0.0447, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0108, 0.0098, 0.0140, 0.0082, 0.0126, 0.0129, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:01:10,503 INFO [train.py:904] (0/8) Epoch 23, batch 1250, loss[loss=0.1513, simple_loss=0.2519, pruned_loss=0.02537, over 17255.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2467, pruned_loss=0.03787, over 3317003.40 frames. ], batch size: 52, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:01:22,904 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224562.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:01:31,201 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224568.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:02:20,428 INFO [train.py:904] (0/8) Epoch 23, batch 1300, loss[loss=0.1573, simple_loss=0.2441, pruned_loss=0.03525, over 16508.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.247, pruned_loss=0.03801, over 3315479.86 frames. ], batch size: 68, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:02:33,663 INFO [optim.py:368] (0/8) 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 14:02:34,504 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 14:02:49,067 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224623.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:03:29,365 INFO [train.py:904] (0/8) Epoch 23, batch 1350, loss[loss=0.1659, simple_loss=0.2624, pruned_loss=0.0347, over 17115.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2472, pruned_loss=0.03792, over 3316358.68 frames. ], batch size: 49, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:22,857 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-01 14:04:38,993 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 14:04:40,095 INFO [train.py:904] (0/8) Epoch 23, batch 1400, loss[loss=0.1651, simple_loss=0.2602, pruned_loss=0.03493, over 16580.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2471, pruned_loss=0.03808, over 3318529.07 frames. ], batch size: 62, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:04:52,699 INFO [optim.py:368] (0/8) 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 14:05:07,948 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6640, 3.7933, 2.9266, 2.2281, 2.3871, 2.3458, 3.8877, 3.2236], device='cuda:0'), covar=tensor([0.2722, 0.0552, 0.1687, 0.3096, 0.2884, 0.2207, 0.0462, 0.1565], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0270, 0.0308, 0.0316, 0.0298, 0.0264, 0.0299, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 14:05:08,259 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 14:05:14,005 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=224726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:05:33,042 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 23, batch 1450, loss[loss=0.1433, simple_loss=0.2297, pruned_loss=0.02845, over 16838.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2464, pruned_loss=0.03761, over 3322574.55 frames. ], batch size: 42, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:06:23,779 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9082, 4.2821, 4.3479, 2.9174, 3.6820, 4.2773, 3.8744, 2.6378], device='cuda:0'), covar=tensor([0.0490, 0.0091, 0.0052, 0.0409, 0.0132, 0.0105, 0.0100, 0.0482], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0085, 0.0086, 0.0136, 0.0100, 0.0111, 0.0097, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 14:06:31,489 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0984, 2.4440, 2.6433, 1.8855, 2.7939, 2.7821, 2.4629, 2.3774], device='cuda:0'), covar=tensor([0.0818, 0.0297, 0.0278, 0.1144, 0.0149, 0.0322, 0.0537, 0.0522], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0109, 0.0099, 0.0141, 0.0082, 0.0127, 0.0129, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:06:39,652 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:06:59,967 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=224801.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:07:01,748 INFO [train.py:904] (0/8) Epoch 23, batch 1500, loss[loss=0.1543, simple_loss=0.2304, pruned_loss=0.03907, over 16231.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2463, pruned_loss=0.03794, over 3320761.08 frames. ], batch size: 165, lr: 2.96e-03, grad_scale: 8.0 2023-05-01 14:07:16,529 INFO [optim.py:368] (0/8) 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 14:07:55,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6167, 6.0352, 5.7797, 5.7722, 5.3959, 5.4436, 5.4003, 6.1530], device='cuda:0'), covar=tensor([0.1530, 0.0932, 0.1125, 0.0927, 0.0939, 0.0726, 0.1255, 0.0929], device='cuda:0'), in_proj_covar=tensor([0.0697, 0.0849, 0.0696, 0.0642, 0.0534, 0.0543, 0.0712, 0.0662], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:08:14,261 INFO [train.py:904] (0/8) Epoch 23, batch 1550, loss[loss=0.1824, simple_loss=0.2683, pruned_loss=0.04827, over 16648.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2477, pruned_loss=0.03907, over 3322667.62 frames. ], batch size: 62, lr: 2.96e-03, grad_scale: 4.0 2023-05-01 14:08:34,622 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=224868.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:23,930 INFO [train.py:904] (0/8) Epoch 23, batch 1600, loss[loss=0.1863, simple_loss=0.2581, pruned_loss=0.05718, over 16916.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2499, pruned_loss=0.04006, over 3325851.35 frames. ], batch size: 109, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:09:36,820 INFO [optim.py:368] (0/8) 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 14:09:41,253 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=224916.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:09:43,706 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=224918.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:10:32,920 INFO [train.py:904] (0/8) Epoch 23, batch 1650, loss[loss=0.1706, simple_loss=0.2618, pruned_loss=0.03973, over 16643.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2505, pruned_loss=0.04015, over 3332405.89 frames. ], batch size: 62, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:00,561 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-01 14:11:41,683 INFO [train.py:904] (0/8) Epoch 23, batch 1700, loss[loss=0.1741, simple_loss=0.2549, pruned_loss=0.04663, over 16881.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2528, pruned_loss=0.04117, over 3328884.79 frames. ], batch size: 116, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:11:56,162 INFO [optim.py:368] (0/8) 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 14:12:05,953 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6245, 4.5753, 4.5249, 4.0577, 4.5612, 1.8454, 4.3254, 4.1670], device='cuda:0'), covar=tensor([0.0148, 0.0121, 0.0196, 0.0328, 0.0130, 0.2797, 0.0169, 0.0249], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0163, 0.0206, 0.0183, 0.0184, 0.0216, 0.0196, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:12:52,539 INFO [train.py:904] (0/8) Epoch 23, batch 1750, loss[loss=0.1739, simple_loss=0.2653, pruned_loss=0.04119, over 17080.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2532, pruned_loss=0.04111, over 3324584.15 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:13:20,607 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0494, 4.8596, 5.1065, 5.2699, 5.5119, 4.8273, 5.4744, 5.4898], device='cuda:0'), covar=tensor([0.1921, 0.1359, 0.1801, 0.0858, 0.0557, 0.1060, 0.0590, 0.0658], device='cuda:0'), in_proj_covar=tensor([0.0674, 0.0831, 0.0962, 0.0841, 0.0635, 0.0667, 0.0691, 0.0801], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:13:32,967 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:13:52,569 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225096.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:14:01,749 INFO [train.py:904] (0/8) Epoch 23, batch 1800, loss[loss=0.1646, simple_loss=0.2514, pruned_loss=0.03888, over 16998.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2547, pruned_loss=0.04095, over 3331181.23 frames. ], batch size: 41, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:14:02,312 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225103.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:14:07,378 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 14:14:15,808 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.405e+02 2.184e+02 2.489e+02 3.045e+02 6.303e+02, threshold=4.978e+02, percent-clipped=2.0 2023-05-01 14:14:57,247 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225142.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:15:12,128 INFO [train.py:904] (0/8) Epoch 23, batch 1850, loss[loss=0.1978, simple_loss=0.2808, pruned_loss=0.05737, over 12227.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2552, pruned_loss=0.0407, over 3322091.24 frames. ], batch size: 247, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:15:27,756 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:15:44,149 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.81 vs. limit=5.0 2023-05-01 14:15:56,970 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3724, 5.3609, 5.1335, 4.5891, 5.1742, 2.2078, 4.9788, 5.1074], device='cuda:0'), covar=tensor([0.0095, 0.0094, 0.0217, 0.0451, 0.0110, 0.2639, 0.0160, 0.0208], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0165, 0.0207, 0.0184, 0.0185, 0.0218, 0.0197, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:16:22,255 INFO [train.py:904] (0/8) Epoch 23, batch 1900, loss[loss=0.1711, simple_loss=0.2487, pruned_loss=0.04674, over 16765.00 frames. ], tot_loss[loss=0.1685, simple_loss=0.2555, pruned_loss=0.04072, over 3322201.64 frames. ], batch size: 124, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:16:22,760 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225203.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:16:33,317 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 14:16:36,684 INFO [optim.py:368] (0/8) 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 14:16:44,053 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:17:22,569 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1081, 4.8700, 5.1450, 5.3014, 5.5836, 4.8202, 5.5293, 5.5266], device='cuda:0'), covar=tensor([0.2106, 0.1460, 0.2044, 0.0962, 0.0600, 0.0968, 0.0569, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0678, 0.0838, 0.0971, 0.0847, 0.0640, 0.0672, 0.0696, 0.0806], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:17:31,821 INFO [train.py:904] (0/8) Epoch 23, batch 1950, loss[loss=0.1836, simple_loss=0.2703, pruned_loss=0.04841, over 17062.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2554, pruned_loss=0.04023, over 3319668.78 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:17:50,057 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225266.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:18:42,386 INFO [train.py:904] (0/8) Epoch 23, batch 2000, loss[loss=0.1271, simple_loss=0.2201, pruned_loss=0.0171, over 16821.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2539, pruned_loss=0.03975, over 3315910.04 frames. ], batch size: 39, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:18:56,052 INFO [optim.py:368] (0/8) 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 14:19:25,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7668, 3.9560, 2.4787, 4.5860, 3.1486, 4.5199, 2.6025, 3.2735], device='cuda:0'), covar=tensor([0.0366, 0.0460, 0.1645, 0.0295, 0.0791, 0.0605, 0.1519, 0.0742], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0169, 0.0179, 0.0221, 0.0205, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:19:52,862 INFO [train.py:904] (0/8) Epoch 23, batch 2050, loss[loss=0.1707, simple_loss=0.2482, pruned_loss=0.04664, over 16774.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2542, pruned_loss=0.03981, over 3318193.58 frames. ], batch size: 124, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:20:15,492 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 14:20:27,345 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8878, 4.8037, 4.6339, 3.4984, 4.7439, 1.7656, 4.3228, 4.2904], device='cuda:0'), covar=tensor([0.0206, 0.0182, 0.0294, 0.0803, 0.0186, 0.3597, 0.0254, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0164, 0.0207, 0.0184, 0.0185, 0.0217, 0.0197, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:20:35,638 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:20:54,846 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225396.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:04,284 INFO [train.py:904] (0/8) Epoch 23, batch 2100, loss[loss=0.1626, simple_loss=0.243, pruned_loss=0.04112, over 16807.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2546, pruned_loss=0.04018, over 3318290.55 frames. ], batch size: 102, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:21:18,936 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.459e+02 2.181e+02 2.818e+02 3.430e+02 7.396e+02, threshold=5.635e+02, percent-clipped=7.0 2023-05-01 14:21:44,084 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:21:56,535 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225444.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:15,773 INFO [train.py:904] (0/8) Epoch 23, batch 2150, loss[loss=0.2171, simple_loss=0.2938, pruned_loss=0.07022, over 12168.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2551, pruned_loss=0.04021, over 3325485.99 frames. ], batch size: 247, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:22:25,211 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:22:29,447 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225462.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:18,673 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225498.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:22,043 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:26,036 INFO [train.py:904] (0/8) Epoch 23, batch 2200, loss[loss=0.1636, simple_loss=0.2427, pruned_loss=0.04222, over 16679.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2556, pruned_loss=0.04059, over 3326374.47 frames. ], batch size: 134, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:23:29,384 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225505.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:40,048 INFO [optim.py:368] (0/8) 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 14:23:52,812 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225522.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:23:54,648 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225523.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:24:36,368 INFO [train.py:904] (0/8) Epoch 23, batch 2250, loss[loss=0.1528, simple_loss=0.2355, pruned_loss=0.03511, over 16499.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2562, pruned_loss=0.04099, over 3330521.90 frames. ], batch size: 75, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:24:41,602 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-01 14:24:45,182 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225559.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 14:24:54,925 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:15,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8054, 2.9503, 2.7735, 4.8955, 4.0564, 4.3505, 1.6387, 3.2549], device='cuda:0'), covar=tensor([0.1379, 0.0736, 0.1145, 0.0209, 0.0226, 0.0390, 0.1598, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0176, 0.0196, 0.0193, 0.0205, 0.0219, 0.0205, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:25:18,174 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225583.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:25:46,980 INFO [train.py:904] (0/8) Epoch 23, batch 2300, loss[loss=0.1528, simple_loss=0.2524, pruned_loss=0.02657, over 17240.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2576, pruned_loss=0.0417, over 3318775.09 frames. ], batch size: 52, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:26:01,561 INFO [optim.py:368] (0/8) 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 14:26:07,678 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8718, 3.9721, 3.0280, 2.3949, 2.6881, 2.5628, 4.1990, 3.4578], device='cuda:0'), covar=tensor([0.2559, 0.0549, 0.1642, 0.2765, 0.2646, 0.1999, 0.0484, 0.1392], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0270, 0.0308, 0.0317, 0.0300, 0.0264, 0.0300, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 14:26:11,588 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225620.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:26:58,889 INFO [train.py:904] (0/8) Epoch 23, batch 2350, loss[loss=0.1501, simple_loss=0.2492, pruned_loss=0.02546, over 17026.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2579, pruned_loss=0.04216, over 3320758.80 frames. ], batch size: 50, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:27:19,965 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3783, 4.2203, 4.4711, 4.5810, 4.6608, 4.2756, 4.4950, 4.6785], device='cuda:0'), covar=tensor([0.1634, 0.1180, 0.1237, 0.0715, 0.0620, 0.1168, 0.2872, 0.0732], device='cuda:0'), in_proj_covar=tensor([0.0674, 0.0836, 0.0968, 0.0844, 0.0640, 0.0671, 0.0694, 0.0803], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:27:35,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9016, 2.8086, 2.5177, 2.7116, 3.1574, 2.9370, 3.6091, 3.3117], device='cuda:0'), covar=tensor([0.0138, 0.0424, 0.0477, 0.0408, 0.0299, 0.0363, 0.0213, 0.0283], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0243, 0.0233, 0.0234, 0.0244, 0.0242, 0.0244, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:27:58,257 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8239, 4.2683, 2.9629, 2.3615, 2.7055, 2.6480, 4.6590, 3.5611], device='cuda:0'), covar=tensor([0.2953, 0.0544, 0.1881, 0.2960, 0.2925, 0.2019, 0.0376, 0.1438], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0271, 0.0309, 0.0317, 0.0300, 0.0264, 0.0300, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 14:28:08,055 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 14:28:10,319 INFO [train.py:904] (0/8) Epoch 23, batch 2400, loss[loss=0.1772, simple_loss=0.2599, pruned_loss=0.04731, over 16880.00 frames. ], tot_loss[loss=0.1729, simple_loss=0.2597, pruned_loss=0.04303, over 3307852.75 frames. ], batch size: 116, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:28:23,216 INFO [optim.py:368] (0/8) 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 14:28:30,218 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 14:29:17,844 INFO [train.py:904] (0/8) Epoch 23, batch 2450, loss[loss=0.1838, simple_loss=0.2639, pruned_loss=0.05186, over 16859.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2589, pruned_loss=0.04227, over 3316041.18 frames. ], batch size: 116, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:29:26,316 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225759.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:17,169 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225795.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:22,269 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=225798.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:30:29,833 INFO [train.py:904] (0/8) Epoch 23, batch 2500, loss[loss=0.1792, simple_loss=0.2561, pruned_loss=0.05111, over 16690.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.259, pruned_loss=0.04223, over 3310883.36 frames. ], batch size: 124, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:30:36,107 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 14:30:47,119 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-01 14:30:50,745 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225818.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:04,193 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6091, 4.5178, 4.5177, 4.2118, 4.3052, 4.5404, 4.2534, 4.2904], device='cuda:0'), covar=tensor([0.0595, 0.0811, 0.0269, 0.0306, 0.0694, 0.0478, 0.0560, 0.0661], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0454, 0.0361, 0.0360, 0.0365, 0.0419, 0.0245, 0.0432], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 14:31:28,976 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=225846.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:38,345 INFO [train.py:904] (0/8) Epoch 23, batch 2550, loss[loss=0.1825, simple_loss=0.2704, pruned_loss=0.04731, over 15609.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2585, pruned_loss=0.04171, over 3320557.48 frames. ], batch size: 191, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:31:49,569 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225861.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:31:50,944 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225862.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:13,464 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225878.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:16,573 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0924, 2.6070, 2.1152, 2.4682, 2.9598, 2.7710, 3.0504, 3.0565], device='cuda:0'), covar=tensor([0.0274, 0.0417, 0.0572, 0.0477, 0.0303, 0.0384, 0.0275, 0.0312], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0242, 0.0231, 0.0233, 0.0242, 0.0241, 0.0243, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:32:38,038 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=225895.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:32:49,288 INFO [train.py:904] (0/8) Epoch 23, batch 2600, loss[loss=0.1646, simple_loss=0.257, pruned_loss=0.03611, over 17112.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2588, pruned_loss=0.04167, over 3319949.99 frames. ], batch size: 47, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:32:51,038 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 14:33:03,049 INFO [optim.py:368] (0/8) 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 14:33:06,087 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=225915.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 14:33:16,598 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225923.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:33:29,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6658, 3.3565, 3.0408, 5.1941, 4.4994, 4.4672, 1.7157, 3.5401], device='cuda:0'), covar=tensor([0.1469, 0.0636, 0.1109, 0.0170, 0.0213, 0.0370, 0.1583, 0.0700], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0194, 0.0205, 0.0218, 0.0205, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:33:56,025 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6721, 3.5276, 3.9275, 2.0776, 3.9913, 3.9833, 3.2229, 2.9269], device='cuda:0'), covar=tensor([0.0769, 0.0229, 0.0151, 0.1174, 0.0104, 0.0206, 0.0372, 0.0473], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0110, 0.0100, 0.0141, 0.0082, 0.0128, 0.0129, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:33:58,802 INFO [train.py:904] (0/8) Epoch 23, batch 2650, loss[loss=0.1642, simple_loss=0.26, pruned_loss=0.03424, over 16711.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.2587, pruned_loss=0.04102, over 3324734.12 frames. ], batch size: 76, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:34:04,299 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=225956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:34:46,587 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7548, 2.6420, 2.2679, 2.4827, 3.0036, 2.7869, 3.2621, 3.2465], device='cuda:0'), covar=tensor([0.0171, 0.0477, 0.0583, 0.0526, 0.0332, 0.0422, 0.0365, 0.0308], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0243, 0.0231, 0.0233, 0.0243, 0.0241, 0.0244, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:34:51,887 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-05-01 14:35:05,057 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-226000.pt 2023-05-01 14:35:12,277 INFO [train.py:904] (0/8) Epoch 23, batch 2700, loss[loss=0.1767, simple_loss=0.2594, pruned_loss=0.047, over 16817.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2586, pruned_loss=0.041, over 3322811.60 frames. ], batch size: 102, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:35:25,001 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0480, 4.8008, 5.0495, 5.2422, 5.4685, 4.7794, 5.4496, 5.4592], device='cuda:0'), covar=tensor([0.1956, 0.1447, 0.1874, 0.0815, 0.0534, 0.0928, 0.0575, 0.0616], device='cuda:0'), in_proj_covar=tensor([0.0672, 0.0835, 0.0965, 0.0842, 0.0640, 0.0668, 0.0692, 0.0803], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:35:25,732 INFO [optim.py:368] (0/8) 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 14:36:06,285 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226041.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:36:13,625 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7773, 4.7268, 4.6420, 4.0718, 4.6975, 1.7724, 4.4599, 4.3786], device='cuda:0'), covar=tensor([0.0129, 0.0106, 0.0184, 0.0339, 0.0107, 0.2893, 0.0147, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0165, 0.0207, 0.0184, 0.0185, 0.0216, 0.0197, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:36:23,070 INFO [train.py:904] (0/8) Epoch 23, batch 2750, loss[loss=0.1624, simple_loss=0.2654, pruned_loss=0.0297, over 17068.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2587, pruned_loss=0.04019, over 3324409.54 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:21,846 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226095.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:25,071 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6818, 1.8166, 2.2673, 2.4738, 2.6449, 2.5405, 1.9377, 2.8072], device='cuda:0'), covar=tensor([0.0188, 0.0511, 0.0341, 0.0297, 0.0312, 0.0352, 0.0567, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0197, 0.0184, 0.0190, 0.0201, 0.0160, 0.0200, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:37:31,349 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226102.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:32,649 INFO [train.py:904] (0/8) Epoch 23, batch 2800, loss[loss=0.1825, simple_loss=0.2591, pruned_loss=0.05295, over 16909.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2589, pruned_loss=0.04088, over 3326342.79 frames. ], batch size: 109, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:37:47,351 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226118.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:37:59,205 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8375, 1.9352, 2.4502, 2.7600, 2.7584, 3.2203, 2.1508, 3.2467], device='cuda:0'), covar=tensor([0.0285, 0.0611, 0.0394, 0.0391, 0.0368, 0.0253, 0.0608, 0.0171], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0197, 0.0184, 0.0190, 0.0202, 0.0160, 0.0201, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:38:29,331 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226143.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:38:42,396 INFO [train.py:904] (0/8) Epoch 23, batch 2850, loss[loss=0.1679, simple_loss=0.2587, pruned_loss=0.0385, over 17208.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2574, pruned_loss=0.04049, over 3328985.74 frames. ], batch size: 46, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:38:54,346 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226161.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:00,761 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226166.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:18,094 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226178.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:39:51,840 INFO [train.py:904] (0/8) Epoch 23, batch 2900, loss[loss=0.1597, simple_loss=0.2494, pruned_loss=0.03497, over 17100.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2564, pruned_loss=0.04046, over 3325849.25 frames. ], batch size: 53, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:39:59,570 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-01 14:40:00,218 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 14:40:09,312 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226215.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 14:40:13,519 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226218.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:24,112 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:33,451 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:40:58,444 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:41:00,400 INFO [train.py:904] (0/8) Epoch 23, batch 2950, loss[loss=0.1741, simple_loss=0.2649, pruned_loss=0.04166, over 16716.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.256, pruned_loss=0.04112, over 3326410.81 frames. ], batch size: 57, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:41:14,092 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226263.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 14:41:56,429 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=226294.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:42:08,781 INFO [train.py:904] (0/8) Epoch 23, batch 3000, loss[loss=0.157, simple_loss=0.2425, pruned_loss=0.0358, over 16770.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2569, pruned_loss=0.04162, over 3325109.43 frames. ], batch size: 39, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:42:08,782 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 14:42:17,864 INFO [train.py:938] (0/8) Epoch 23, validation: loss=0.1344, simple_loss=0.2397, pruned_loss=0.01456, over 944034.00 frames. 2023-05-01 14:42:17,864 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 14:42:30,995 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.698e+02 2.313e+02 2.743e+02 3.282e+02 6.136e+02, threshold=5.486e+02, percent-clipped=4.0 2023-05-01 14:43:24,627 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 14:43:26,889 INFO [train.py:904] (0/8) Epoch 23, batch 3050, loss[loss=0.1763, simple_loss=0.2562, pruned_loss=0.04825, over 12565.00 frames. ], tot_loss[loss=0.17, simple_loss=0.257, pruned_loss=0.04152, over 3327057.09 frames. ], batch size: 246, lr: 2.95e-03, grad_scale: 8.0 2023-05-01 14:44:29,427 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:44:37,980 INFO [train.py:904] (0/8) Epoch 23, batch 3100, loss[loss=0.1547, simple_loss=0.2392, pruned_loss=0.0351, over 15640.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2567, pruned_loss=0.04189, over 3326929.41 frames. ], batch size: 191, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:44:51,601 INFO [optim.py:368] (0/8) 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 14:45:18,540 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0011, 5.1114, 5.4988, 5.4848, 5.5187, 5.1765, 5.1014, 4.9052], device='cuda:0'), covar=tensor([0.0330, 0.0514, 0.0399, 0.0411, 0.0428, 0.0373, 0.0905, 0.0443], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0479, 0.0465, 0.0431, 0.0510, 0.0487, 0.0574, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 14:45:47,159 INFO [train.py:904] (0/8) Epoch 23, batch 3150, loss[loss=0.166, simple_loss=0.2438, pruned_loss=0.04406, over 16890.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.256, pruned_loss=0.0416, over 3330873.36 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:46:54,863 INFO [train.py:904] (0/8) Epoch 23, batch 3200, loss[loss=0.172, simple_loss=0.2479, pruned_loss=0.04806, over 16925.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2554, pruned_loss=0.04155, over 3328707.45 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:47:09,850 INFO [optim.py:368] (0/8) 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 14:47:15,802 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:01,443 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226551.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:04,033 INFO [train.py:904] (0/8) Epoch 23, batch 3250, loss[loss=0.1581, simple_loss=0.2488, pruned_loss=0.03371, over 17251.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2554, pruned_loss=0.04143, over 3326658.96 frames. ], batch size: 45, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:48:22,284 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226566.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:54,457 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=226589.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:48:54,667 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6674, 2.6574, 2.2408, 2.5070, 2.9951, 2.6808, 3.2798, 3.1896], device='cuda:0'), covar=tensor([0.0209, 0.0487, 0.0619, 0.0502, 0.0293, 0.0465, 0.0270, 0.0311], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0246, 0.0234, 0.0235, 0.0246, 0.0244, 0.0247, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:49:08,848 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226599.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:49:14,406 INFO [train.py:904] (0/8) Epoch 23, batch 3300, loss[loss=0.1733, simple_loss=0.2654, pruned_loss=0.04058, over 16747.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2555, pruned_loss=0.04126, over 3336113.97 frames. ], batch size: 57, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:49:27,929 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.422e+02 2.105e+02 2.619e+02 3.003e+02 4.974e+02, threshold=5.238e+02, percent-clipped=1.0 2023-05-01 14:49:38,671 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9253, 3.1462, 3.2476, 2.1991, 2.8395, 2.2548, 3.5272, 3.4253], device='cuda:0'), covar=tensor([0.0254, 0.0821, 0.0641, 0.1800, 0.0886, 0.1012, 0.0516, 0.0851], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-01 14:50:07,282 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9266, 1.3634, 1.6928, 1.7137, 1.8139, 1.9953, 1.6378, 1.8319], device='cuda:0'), covar=tensor([0.0273, 0.0436, 0.0236, 0.0318, 0.0301, 0.0222, 0.0491, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0198, 0.0185, 0.0191, 0.0204, 0.0161, 0.0202, 0.0158], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:50:23,129 INFO [train.py:904] (0/8) Epoch 23, batch 3350, loss[loss=0.1837, simple_loss=0.2615, pruned_loss=0.05291, over 16882.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2564, pruned_loss=0.04142, over 3329404.37 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:17,156 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5241, 5.8835, 5.6551, 5.7240, 5.3226, 5.2630, 5.2620, 6.0623], device='cuda:0'), covar=tensor([0.1465, 0.0989, 0.1079, 0.0926, 0.0946, 0.0806, 0.1262, 0.0902], device='cuda:0'), in_proj_covar=tensor([0.0707, 0.0862, 0.0708, 0.0654, 0.0543, 0.0550, 0.0722, 0.0673], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:51:25,707 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226697.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:51:34,610 INFO [train.py:904] (0/8) Epoch 23, batch 3400, loss[loss=0.1567, simple_loss=0.2536, pruned_loss=0.02988, over 17103.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2562, pruned_loss=0.04114, over 3330716.54 frames. ], batch size: 49, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:51:47,771 INFO [optim.py:368] (0/8) 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 14:52:33,243 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226745.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:52:43,441 INFO [train.py:904] (0/8) Epoch 23, batch 3450, loss[loss=0.1506, simple_loss=0.2331, pruned_loss=0.03406, over 16478.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2541, pruned_loss=0.04018, over 3335065.93 frames. ], batch size: 68, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:52:51,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9670, 3.1478, 2.9120, 5.1855, 4.1704, 4.5375, 1.8610, 3.2482], device='cuda:0'), covar=tensor([0.1377, 0.0750, 0.1170, 0.0206, 0.0213, 0.0357, 0.1606, 0.0778], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0176, 0.0197, 0.0196, 0.0207, 0.0219, 0.0205, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:53:52,963 INFO [train.py:904] (0/8) Epoch 23, batch 3500, loss[loss=0.1558, simple_loss=0.2504, pruned_loss=0.0306, over 17250.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2533, pruned_loss=0.0398, over 3334804.23 frames. ], batch size: 52, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:54:07,207 INFO [optim.py:368] (0/8) 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 14:55:03,888 INFO [train.py:904] (0/8) Epoch 23, batch 3550, loss[loss=0.1802, simple_loss=0.2491, pruned_loss=0.05569, over 16931.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2526, pruned_loss=0.0397, over 3328732.31 frames. ], batch size: 96, lr: 2.94e-03, grad_scale: 16.0 2023-05-01 14:55:32,128 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3414, 4.2946, 4.2759, 3.7190, 4.2631, 1.8550, 4.0755, 3.7725], device='cuda:0'), covar=tensor([0.0137, 0.0125, 0.0187, 0.0259, 0.0106, 0.2812, 0.0141, 0.0259], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0167, 0.0208, 0.0185, 0.0186, 0.0215, 0.0197, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:55:52,870 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=226889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:56:12,763 INFO [train.py:904] (0/8) Epoch 23, batch 3600, loss[loss=0.1522, simple_loss=0.2551, pruned_loss=0.02468, over 17124.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2509, pruned_loss=0.03917, over 3335434.91 frames. ], batch size: 48, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:56:28,219 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.466e+02 2.070e+02 2.545e+02 2.955e+02 4.911e+02, threshold=5.089e+02, percent-clipped=1.0 2023-05-01 14:57:02,357 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=226937.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:57:24,673 INFO [train.py:904] (0/8) Epoch 23, batch 3650, loss[loss=0.152, simple_loss=0.2236, pruned_loss=0.04025, over 16889.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2505, pruned_loss=0.03956, over 3325499.79 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:19,547 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=226989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:58:27,863 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9588, 5.1878, 5.3739, 5.1540, 5.2018, 5.7721, 5.2628, 5.0130], device='cuda:0'), covar=tensor([0.1172, 0.1992, 0.1956, 0.2036, 0.2497, 0.0960, 0.1619, 0.2249], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0617, 0.0681, 0.0508, 0.0680, 0.0703, 0.0534, 0.0681], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 14:58:40,560 INFO [train.py:904] (0/8) Epoch 23, batch 3700, loss[loss=0.1608, simple_loss=0.2411, pruned_loss=0.04022, over 16318.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2496, pruned_loss=0.04142, over 3297614.24 frames. ], batch size: 165, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 14:58:56,578 INFO [optim.py:368] (0/8) 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 14:59:31,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7753, 4.0104, 2.5705, 4.6855, 3.1599, 4.6370, 2.6539, 3.1695], device='cuda:0'), covar=tensor([0.0285, 0.0334, 0.1493, 0.0145, 0.0717, 0.0356, 0.1375, 0.0710], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0180, 0.0195, 0.0170, 0.0178, 0.0223, 0.0205, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 14:59:39,684 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6749, 4.6471, 4.5957, 4.0239, 4.6394, 1.8851, 4.4566, 4.2621], device='cuda:0'), covar=tensor([0.0135, 0.0105, 0.0173, 0.0296, 0.0102, 0.2700, 0.0136, 0.0221], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0165, 0.0206, 0.0182, 0.0184, 0.0212, 0.0195, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 14:59:50,715 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227050.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 14:59:53,704 INFO [train.py:904] (0/8) Epoch 23, batch 3750, loss[loss=0.1717, simple_loss=0.2519, pruned_loss=0.04576, over 16220.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2498, pruned_loss=0.04255, over 3275672.54 frames. ], batch size: 165, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:00:55,185 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 15:01:07,857 INFO [train.py:904] (0/8) Epoch 23, batch 3800, loss[loss=0.1888, simple_loss=0.2541, pruned_loss=0.06176, over 16843.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2512, pruned_loss=0.04398, over 3276263.79 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:01:25,259 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.697e+02 2.146e+02 2.630e+02 3.388e+02 6.503e+02, threshold=5.260e+02, percent-clipped=3.0 2023-05-01 15:01:33,034 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-01 15:01:45,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3563, 3.5740, 3.6704, 3.6371, 3.6518, 3.5186, 3.5339, 3.5498], device='cuda:0'), covar=tensor([0.0406, 0.0571, 0.0450, 0.0473, 0.0566, 0.0459, 0.0729, 0.0506], device='cuda:0'), in_proj_covar=tensor([0.0429, 0.0480, 0.0463, 0.0430, 0.0511, 0.0485, 0.0572, 0.0389], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 15:02:21,632 INFO [train.py:904] (0/8) Epoch 23, batch 3850, loss[loss=0.1726, simple_loss=0.2475, pruned_loss=0.04889, over 16858.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2516, pruned_loss=0.04469, over 3271221.94 frames. ], batch size: 109, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:19,668 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-01 15:03:34,969 INFO [train.py:904] (0/8) Epoch 23, batch 3900, loss[loss=0.2021, simple_loss=0.2822, pruned_loss=0.06095, over 12352.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2519, pruned_loss=0.04527, over 3270623.09 frames. ], batch size: 248, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:03:51,651 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.638e+02 2.230e+02 2.530e+02 3.049e+02 6.179e+02, threshold=5.060e+02, percent-clipped=1.0 2023-05-01 15:04:47,752 INFO [train.py:904] (0/8) Epoch 23, batch 3950, loss[loss=0.1886, simple_loss=0.2596, pruned_loss=0.05883, over 16707.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2513, pruned_loss=0.04593, over 3273026.11 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:00,630 INFO [train.py:904] (0/8) Epoch 23, batch 4000, loss[loss=0.1693, simple_loss=0.2468, pruned_loss=0.04591, over 17071.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2512, pruned_loss=0.04581, over 3275460.26 frames. ], batch size: 55, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:06:17,148 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.172e+02 2.478e+02 2.977e+02 5.073e+02, threshold=4.957e+02, percent-clipped=1.0 2023-05-01 15:07:01,810 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227345.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:07:13,394 INFO [train.py:904] (0/8) Epoch 23, batch 4050, loss[loss=0.1625, simple_loss=0.2475, pruned_loss=0.03874, over 16722.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2517, pruned_loss=0.0453, over 3269268.33 frames. ], batch size: 89, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:15,539 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0397, 3.5802, 3.4737, 2.1717, 3.2700, 3.5724, 3.2251, 1.9429], device='cuda:0'), covar=tensor([0.0524, 0.0047, 0.0058, 0.0432, 0.0100, 0.0104, 0.0111, 0.0471], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0084, 0.0086, 0.0133, 0.0099, 0.0111, 0.0096, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 15:08:27,054 INFO [train.py:904] (0/8) Epoch 23, batch 4100, loss[loss=0.1768, simple_loss=0.2693, pruned_loss=0.04219, over 16357.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.253, pruned_loss=0.04469, over 3274701.15 frames. ], batch size: 146, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:08:34,590 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7360, 1.9103, 2.3987, 2.7225, 2.6919, 3.0972, 1.9919, 2.9921], device='cuda:0'), covar=tensor([0.0252, 0.0578, 0.0356, 0.0364, 0.0325, 0.0202, 0.0583, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0196, 0.0184, 0.0191, 0.0204, 0.0160, 0.0201, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:08:39,200 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227411.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:08:42,426 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-01 15:08:42,770 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.244e+02 1.752e+02 1.921e+02 2.258e+02 5.560e+02, threshold=3.841e+02, percent-clipped=1.0 2023-05-01 15:09:45,401 INFO [train.py:904] (0/8) Epoch 23, batch 4150, loss[loss=0.2007, simple_loss=0.2984, pruned_loss=0.05148, over 16665.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2607, pruned_loss=0.04704, over 3237569.15 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:09:57,509 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2046, 4.0969, 4.2115, 4.3956, 4.5185, 4.1752, 4.4921, 4.5471], device='cuda:0'), covar=tensor([0.1873, 0.1313, 0.1835, 0.0858, 0.0711, 0.1263, 0.0819, 0.0831], device='cuda:0'), in_proj_covar=tensor([0.0664, 0.0827, 0.0954, 0.0834, 0.0634, 0.0663, 0.0686, 0.0797], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:10:15,624 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227472.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:10:36,123 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5362, 3.4178, 2.5985, 2.2635, 2.3384, 2.3394, 3.6166, 3.1520], device='cuda:0'), covar=tensor([0.3002, 0.0725, 0.1903, 0.2708, 0.2568, 0.2087, 0.0517, 0.1292], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0273, 0.0310, 0.0319, 0.0304, 0.0266, 0.0301, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 15:11:03,946 INFO [train.py:904] (0/8) Epoch 23, batch 4200, loss[loss=0.221, simple_loss=0.3098, pruned_loss=0.06606, over 15243.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2681, pruned_loss=0.04888, over 3222661.68 frames. ], batch size: 190, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:11:20,463 INFO [optim.py:368] (0/8) 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 15:12:13,307 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0144, 2.5040, 2.6341, 1.9218, 2.7984, 2.8314, 2.4744, 2.4242], device='cuda:0'), covar=tensor([0.0764, 0.0248, 0.0266, 0.0973, 0.0131, 0.0244, 0.0472, 0.0446], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0109, 0.0099, 0.0138, 0.0082, 0.0127, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 15:12:19,945 INFO [train.py:904] (0/8) Epoch 23, batch 4250, loss[loss=0.1688, simple_loss=0.2662, pruned_loss=0.03565, over 16888.00 frames. ], tot_loss[loss=0.1844, simple_loss=0.2716, pruned_loss=0.04858, over 3210113.30 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:12:32,653 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6024, 4.6906, 4.9632, 4.9146, 4.9546, 4.7040, 4.5909, 4.4538], device='cuda:0'), covar=tensor([0.0323, 0.0465, 0.0426, 0.0474, 0.0464, 0.0389, 0.1045, 0.0539], device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0469, 0.0453, 0.0420, 0.0500, 0.0475, 0.0562, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 15:13:36,403 INFO [train.py:904] (0/8) Epoch 23, batch 4300, loss[loss=0.1831, simple_loss=0.287, pruned_loss=0.03957, over 16805.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.272, pruned_loss=0.04753, over 3196425.88 frames. ], batch size: 102, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:13:55,090 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.568e+02 2.112e+02 2.458e+02 3.022e+02 6.068e+02, threshold=4.917e+02, percent-clipped=1.0 2023-05-01 15:13:56,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3709, 4.4931, 4.2874, 4.0139, 3.9888, 4.4156, 4.0718, 4.1069], device='cuda:0'), covar=tensor([0.0628, 0.0405, 0.0307, 0.0321, 0.0877, 0.0375, 0.0640, 0.0583], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0449, 0.0358, 0.0356, 0.0362, 0.0413, 0.0243, 0.0426], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:14:41,972 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=227645.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:14:53,546 INFO [train.py:904] (0/8) Epoch 23, batch 4350, loss[loss=0.2168, simple_loss=0.3031, pruned_loss=0.06524, over 16829.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2753, pruned_loss=0.04851, over 3206387.36 frames. ], batch size: 116, lr: 2.94e-03, grad_scale: 4.0 2023-05-01 15:15:45,309 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-01 15:15:55,223 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:15:55,352 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227693.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:16:09,827 INFO [train.py:904] (0/8) Epoch 23, batch 4400, loss[loss=0.176, simple_loss=0.2675, pruned_loss=0.0423, over 16231.00 frames. ], tot_loss[loss=0.1882, simple_loss=0.2774, pruned_loss=0.04951, over 3203435.01 frames. ], batch size: 165, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:16:27,163 INFO [optim.py:368] (0/8) 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 15:16:41,858 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1597, 2.4085, 2.2832, 3.9589, 2.0890, 2.7315, 2.4174, 2.4242], device='cuda:0'), covar=tensor([0.1266, 0.2966, 0.2729, 0.0489, 0.4104, 0.2177, 0.2867, 0.3351], device='cuda:0'), in_proj_covar=tensor([0.0407, 0.0457, 0.0372, 0.0332, 0.0439, 0.0527, 0.0426, 0.0534], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:17:22,050 INFO [train.py:904] (0/8) Epoch 23, batch 4450, loss[loss=0.2008, simple_loss=0.2909, pruned_loss=0.05539, over 16744.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2809, pruned_loss=0.05078, over 3218523.57 frames. ], batch size: 124, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:17:23,574 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227754.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:17:41,690 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=227767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:18:23,613 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 15:18:35,094 INFO [train.py:904] (0/8) Epoch 23, batch 4500, loss[loss=0.1729, simple_loss=0.2669, pruned_loss=0.03952, over 16690.00 frames. ], tot_loss[loss=0.1917, simple_loss=0.2809, pruned_loss=0.05122, over 3225347.98 frames. ], batch size: 134, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:18:52,359 INFO [optim.py:368] (0/8) 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 15:19:11,667 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=227827.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:19:48,161 INFO [train.py:904] (0/8) Epoch 23, batch 4550, loss[loss=0.1963, simple_loss=0.2694, pruned_loss=0.06156, over 11963.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2821, pruned_loss=0.05213, over 3244304.21 frames. ], batch size: 247, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:20:39,349 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=227888.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:21:00,779 INFO [train.py:904] (0/8) Epoch 23, batch 4600, loss[loss=0.1867, simple_loss=0.275, pruned_loss=0.04926, over 16855.00 frames. ], tot_loss[loss=0.1939, simple_loss=0.283, pruned_loss=0.05238, over 3237745.83 frames. ], batch size: 83, lr: 2.94e-03, grad_scale: 8.0 2023-05-01 15:21:18,012 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8685, 2.7904, 2.6197, 1.9060, 2.6545, 2.8044, 2.5862, 1.9449], device='cuda:0'), covar=tensor([0.0459, 0.0079, 0.0080, 0.0405, 0.0114, 0.0120, 0.0124, 0.0407], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 15:21:18,784 INFO [optim.py:368] (0/8) 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 15:22:14,196 INFO [train.py:904] (0/8) Epoch 23, batch 4650, loss[loss=0.1837, simple_loss=0.2684, pruned_loss=0.04952, over 17165.00 frames. ], tot_loss[loss=0.1932, simple_loss=0.2818, pruned_loss=0.05231, over 3242885.75 frames. ], batch size: 46, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:20,835 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-228000.pt 2023-05-01 15:23:28,877 INFO [train.py:904] (0/8) Epoch 23, batch 4700, loss[loss=0.1943, simple_loss=0.2837, pruned_loss=0.05246, over 16454.00 frames. ], tot_loss[loss=0.1913, simple_loss=0.2796, pruned_loss=0.0515, over 3229484.97 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:23:34,944 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228007.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:23:45,414 INFO [optim.py:368] (0/8) 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 15:24:34,816 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 23, batch 4750, loss[loss=0.1443, simple_loss=0.2396, pruned_loss=0.0245, over 16865.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2753, pruned_loss=0.04951, over 3225998.91 frames. ], batch size: 96, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:24:45,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9882, 4.9763, 4.8809, 3.5189, 4.9636, 1.6552, 4.4924, 4.5293], device='cuda:0'), covar=tensor([0.0189, 0.0169, 0.0227, 0.0867, 0.0155, 0.3580, 0.0211, 0.0382], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0162, 0.0203, 0.0180, 0.0180, 0.0209, 0.0191, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:25:02,243 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228067.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:03,396 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228068.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:19,517 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228080.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:25:34,848 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5969, 3.6542, 4.2485, 1.8468, 4.4505, 4.3984, 3.1425, 3.1585], device='cuda:0'), covar=tensor([0.0842, 0.0262, 0.0151, 0.1350, 0.0052, 0.0114, 0.0413, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0108, 0.0099, 0.0139, 0.0081, 0.0126, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 15:25:54,069 INFO [train.py:904] (0/8) Epoch 23, batch 4800, loss[loss=0.1744, simple_loss=0.2725, pruned_loss=0.03816, over 15268.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2714, pruned_loss=0.04764, over 3225308.93 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:26:10,774 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228115.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:26:13,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7328, 4.9019, 5.0482, 4.8562, 4.8566, 5.4611, 4.9356, 4.6229], device='cuda:0'), covar=tensor([0.1164, 0.1666, 0.1856, 0.1843, 0.2476, 0.0944, 0.1460, 0.2429], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0601, 0.0659, 0.0494, 0.0658, 0.0689, 0.0515, 0.0663], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 15:26:49,978 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228141.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:27:07,826 INFO [train.py:904] (0/8) Epoch 23, batch 4850, loss[loss=0.1621, simple_loss=0.2602, pruned_loss=0.03205, over 16397.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2727, pruned_loss=0.047, over 3215298.29 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:27:54,522 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228183.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:28:23,604 INFO [train.py:904] (0/8) Epoch 23, batch 4900, loss[loss=0.1943, simple_loss=0.2739, pruned_loss=0.05733, over 12121.00 frames. ], tot_loss[loss=0.1821, simple_loss=0.2723, pruned_loss=0.04592, over 3196797.69 frames. ], batch size: 247, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:28:27,544 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-01 15:28:42,126 INFO [optim.py:368] (0/8) 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 15:29:36,714 INFO [train.py:904] (0/8) Epoch 23, batch 4950, loss[loss=0.18, simple_loss=0.277, pruned_loss=0.04144, over 16679.00 frames. ], tot_loss[loss=0.1808, simple_loss=0.2715, pruned_loss=0.04508, over 3213089.38 frames. ], batch size: 76, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:29:50,521 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4066, 3.4725, 2.6943, 2.1806, 2.2830, 2.3857, 3.6112, 3.1929], device='cuda:0'), covar=tensor([0.3268, 0.0731, 0.1885, 0.2945, 0.2519, 0.1926, 0.0665, 0.1280], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0271, 0.0308, 0.0317, 0.0301, 0.0264, 0.0300, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 15:30:51,758 INFO [train.py:904] (0/8) Epoch 23, batch 5000, loss[loss=0.1885, simple_loss=0.2855, pruned_loss=0.04573, over 16841.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2729, pruned_loss=0.04489, over 3216142.12 frames. ], batch size: 83, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:31:10,001 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.477e+02 1.949e+02 2.305e+02 2.743e+02 3.734e+02, threshold=4.611e+02, percent-clipped=0.0 2023-05-01 15:31:30,421 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228329.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:36,885 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:31:58,780 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228349.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:04,606 INFO [train.py:904] (0/8) Epoch 23, batch 5050, loss[loss=0.187, simple_loss=0.285, pruned_loss=0.04453, over 15411.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.2734, pruned_loss=0.04485, over 3218314.26 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:32:06,169 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3185, 4.2136, 4.4099, 4.5534, 4.7263, 4.3115, 4.6948, 4.7431], device='cuda:0'), covar=tensor([0.1821, 0.1273, 0.1523, 0.0681, 0.0483, 0.0979, 0.0660, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0797, 0.0921, 0.0803, 0.0613, 0.0639, 0.0661, 0.0766], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:32:19,797 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228363.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:32:27,572 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3966, 3.3547, 3.4458, 3.5067, 3.5682, 3.3062, 3.5509, 3.6165], device='cuda:0'), covar=tensor([0.1198, 0.0920, 0.1019, 0.0609, 0.0578, 0.2198, 0.0956, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0643, 0.0797, 0.0920, 0.0802, 0.0612, 0.0638, 0.0661, 0.0766], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:32:40,872 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-01 15:32:58,456 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228390.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:04,514 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228394.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:09,586 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228397.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:33:17,248 INFO [train.py:904] (0/8) Epoch 23, batch 5100, loss[loss=0.1762, simple_loss=0.2655, pruned_loss=0.04345, over 16434.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2714, pruned_loss=0.04416, over 3218131.69 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:33:34,759 INFO [optim.py:368] (0/8) 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 15:33:55,492 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228429.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:34:05,816 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228436.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:34:30,933 INFO [train.py:904] (0/8) Epoch 23, batch 5150, loss[loss=0.1712, simple_loss=0.2648, pruned_loss=0.03884, over 16505.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2705, pruned_loss=0.0436, over 3207150.98 frames. ], batch size: 75, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:34:54,570 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6133, 2.7546, 2.2275, 2.5416, 3.1309, 2.7655, 3.1870, 3.3073], device='cuda:0'), covar=tensor([0.0120, 0.0464, 0.0567, 0.0461, 0.0241, 0.0396, 0.0225, 0.0275], device='cuda:0'), in_proj_covar=tensor([0.0217, 0.0238, 0.0228, 0.0231, 0.0239, 0.0238, 0.0239, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:34:57,782 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3724, 4.2002, 4.0258, 2.7088, 3.6745, 4.1800, 3.5735, 1.9019], device='cuda:0'), covar=tensor([0.0730, 0.0085, 0.0088, 0.0539, 0.0146, 0.0177, 0.0233, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0085, 0.0087, 0.0135, 0.0100, 0.0112, 0.0096, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 15:35:15,457 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:35:26,002 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228490.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:35:38,115 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6364, 2.5325, 2.3339, 3.8061, 2.6166, 3.8645, 1.5033, 2.9649], device='cuda:0'), covar=tensor([0.1412, 0.0808, 0.1339, 0.0139, 0.0189, 0.0385, 0.1731, 0.0731], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0176, 0.0196, 0.0192, 0.0205, 0.0215, 0.0204, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 15:35:44,370 INFO [train.py:904] (0/8) Epoch 23, batch 5200, loss[loss=0.1599, simple_loss=0.2428, pruned_loss=0.03847, over 16630.00 frames. ], tot_loss[loss=0.178, simple_loss=0.2694, pruned_loss=0.04331, over 3216380.81 frames. ], batch size: 62, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:36:01,317 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228531.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:36:57,637 INFO [train.py:904] (0/8) Epoch 23, batch 5250, loss[loss=0.1701, simple_loss=0.266, pruned_loss=0.0371, over 15401.00 frames. ], tot_loss[loss=0.177, simple_loss=0.2674, pruned_loss=0.04332, over 3202943.31 frames. ], batch size: 191, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:37:59,460 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4339, 3.4199, 2.7116, 2.1990, 2.2601, 2.3595, 3.5108, 3.0749], device='cuda:0'), covar=tensor([0.3173, 0.0675, 0.1933, 0.3052, 0.2637, 0.2148, 0.0590, 0.1429], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0271, 0.0308, 0.0316, 0.0300, 0.0263, 0.0300, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 15:38:10,580 INFO [train.py:904] (0/8) Epoch 23, batch 5300, loss[loss=0.158, simple_loss=0.2523, pruned_loss=0.03183, over 15308.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.264, pruned_loss=0.04215, over 3197887.50 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:38:28,440 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 1.965e+02 2.226e+02 2.604e+02 5.338e+02, threshold=4.452e+02, percent-clipped=3.0 2023-05-01 15:38:28,861 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5677, 3.5191, 3.5202, 2.8870, 3.4136, 1.9876, 3.2637, 2.9211], device='cuda:0'), covar=tensor([0.0154, 0.0148, 0.0179, 0.0336, 0.0128, 0.2485, 0.0162, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0160, 0.0199, 0.0177, 0.0178, 0.0206, 0.0188, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:39:23,402 INFO [train.py:904] (0/8) Epoch 23, batch 5350, loss[loss=0.1589, simple_loss=0.2567, pruned_loss=0.03054, over 16561.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2625, pruned_loss=0.0416, over 3195687.70 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:39:38,458 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:39:48,368 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2696, 4.3416, 4.5043, 4.2812, 4.3394, 4.8341, 4.3544, 4.0201], device='cuda:0'), covar=tensor([0.1726, 0.1862, 0.1924, 0.1975, 0.2706, 0.1059, 0.1587, 0.2761], device='cuda:0'), in_proj_covar=tensor([0.0410, 0.0590, 0.0647, 0.0485, 0.0649, 0.0678, 0.0508, 0.0655], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 15:40:10,176 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228685.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:16,741 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228689.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:36,803 INFO [train.py:904] (0/8) Epoch 23, batch 5400, loss[loss=0.2247, simple_loss=0.3065, pruned_loss=0.07145, over 12184.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2655, pruned_loss=0.04232, over 3196595.61 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:40:48,795 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228711.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:40:54,348 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.419e+02 1.938e+02 2.190e+02 2.538e+02 4.586e+02, threshold=4.379e+02, percent-clipped=1.0 2023-05-01 15:41:26,160 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6312, 4.3650, 4.2796, 2.9329, 3.7097, 4.3057, 3.7320, 2.4069], device='cuda:0'), covar=tensor([0.0493, 0.0036, 0.0040, 0.0352, 0.0103, 0.0069, 0.0095, 0.0427], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0084, 0.0086, 0.0134, 0.0099, 0.0110, 0.0096, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 15:41:27,307 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228736.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:41:28,736 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7172, 1.8298, 1.6667, 1.4857, 1.9729, 1.6375, 1.6262, 1.9458], device='cuda:0'), covar=tensor([0.0169, 0.0312, 0.0409, 0.0364, 0.0229, 0.0265, 0.0175, 0.0222], device='cuda:0'), in_proj_covar=tensor([0.0217, 0.0239, 0.0229, 0.0231, 0.0239, 0.0239, 0.0238, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:41:54,054 INFO [train.py:904] (0/8) Epoch 23, batch 5450, loss[loss=0.1977, simple_loss=0.2936, pruned_loss=0.05086, over 16765.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2683, pruned_loss=0.04367, over 3184256.30 frames. ], batch size: 89, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:41:57,264 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-01 15:42:42,360 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=228784.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:42:44,361 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=228785.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 15:43:12,597 INFO [train.py:904] (0/8) Epoch 23, batch 5500, loss[loss=0.2271, simple_loss=0.3079, pruned_loss=0.07312, over 16286.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2754, pruned_loss=0.0482, over 3154252.91 frames. ], batch size: 165, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:43:32,427 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.602e+02 3.171e+02 3.709e+02 7.104e+02, threshold=6.343e+02, percent-clipped=12.0 2023-05-01 15:44:13,412 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4116, 4.3743, 4.3099, 3.5767, 4.3431, 1.7457, 4.1347, 3.9865], device='cuda:0'), covar=tensor([0.0114, 0.0106, 0.0195, 0.0354, 0.0111, 0.2780, 0.0149, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0177, 0.0206, 0.0189, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:44:31,570 INFO [train.py:904] (0/8) Epoch 23, batch 5550, loss[loss=0.1826, simple_loss=0.2697, pruned_loss=0.04772, over 16267.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2826, pruned_loss=0.05316, over 3129921.67 frames. ], batch size: 35, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:45:52,985 INFO [train.py:904] (0/8) Epoch 23, batch 5600, loss[loss=0.2811, simple_loss=0.3321, pruned_loss=0.1151, over 11108.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2878, pruned_loss=0.05752, over 3095038.21 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:46:10,854 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.072e+02 3.210e+02 3.734e+02 4.915e+02 1.270e+03, threshold=7.469e+02, percent-clipped=7.0 2023-05-01 15:46:31,346 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-01 15:46:45,080 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=228935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:47:13,536 INFO [train.py:904] (0/8) Epoch 23, batch 5650, loss[loss=0.1773, simple_loss=0.2678, pruned_loss=0.04334, over 16858.00 frames. ], tot_loss[loss=0.2071, simple_loss=0.292, pruned_loss=0.06108, over 3056708.18 frames. ], batch size: 42, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:47:58,457 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.39 vs. limit=5.0 2023-05-01 15:48:04,706 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228985.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:10,655 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=228989.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:11,135 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-01 15:48:21,865 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=228996.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:48:32,730 INFO [train.py:904] (0/8) Epoch 23, batch 5700, loss[loss=0.2675, simple_loss=0.3302, pruned_loss=0.1024, over 11118.00 frames. ], tot_loss[loss=0.2116, simple_loss=0.2949, pruned_loss=0.06419, over 3021169.76 frames. ], batch size: 246, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:48:50,701 INFO [optim.py:368] (0/8) 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 15:49:09,438 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.06 vs. limit=5.0 2023-05-01 15:49:18,828 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229033.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:49:24,808 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:49:50,905 INFO [train.py:904] (0/8) Epoch 23, batch 5750, loss[loss=0.2094, simple_loss=0.2813, pruned_loss=0.06877, over 11117.00 frames. ], tot_loss[loss=0.2144, simple_loss=0.298, pruned_loss=0.06538, over 3030819.64 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:50:44,199 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229085.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 15:51:13,983 INFO [train.py:904] (0/8) Epoch 23, batch 5800, loss[loss=0.2103, simple_loss=0.301, pruned_loss=0.05983, over 15351.00 frames. ], tot_loss[loss=0.2117, simple_loss=0.2972, pruned_loss=0.06314, over 3057218.67 frames. ], batch size: 191, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:51:17,213 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9802, 2.4231, 2.1673, 2.2834, 2.7475, 2.3799, 2.6734, 2.8906], device='cuda:0'), covar=tensor([0.0223, 0.0451, 0.0526, 0.0464, 0.0283, 0.0431, 0.0267, 0.0297], device='cuda:0'), in_proj_covar=tensor([0.0215, 0.0236, 0.0226, 0.0229, 0.0237, 0.0236, 0.0236, 0.0234], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:51:32,216 INFO [optim.py:368] (0/8) 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 15:52:01,461 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229133.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 15:52:31,404 INFO [train.py:904] (0/8) Epoch 23, batch 5850, loss[loss=0.2122, simple_loss=0.2921, pruned_loss=0.0661, over 11730.00 frames. ], tot_loss[loss=0.2084, simple_loss=0.2944, pruned_loss=0.06118, over 3057845.96 frames. ], batch size: 248, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:52:44,236 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-01 15:53:50,454 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7810, 2.8816, 2.7783, 4.7634, 3.5585, 4.1943, 1.7237, 3.2818], device='cuda:0'), covar=tensor([0.1360, 0.0731, 0.1093, 0.0134, 0.0300, 0.0342, 0.1602, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0175, 0.0196, 0.0193, 0.0206, 0.0216, 0.0204, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 15:53:53,144 INFO [train.py:904] (0/8) Epoch 23, batch 5900, loss[loss=0.185, simple_loss=0.281, pruned_loss=0.04448, over 16895.00 frames. ], tot_loss[loss=0.207, simple_loss=0.2933, pruned_loss=0.06038, over 3068046.24 frames. ], batch size: 96, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:54:16,067 INFO [optim.py:368] (0/8) 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 15:55:17,176 INFO [train.py:904] (0/8) Epoch 23, batch 5950, loss[loss=0.206, simple_loss=0.2931, pruned_loss=0.05945, over 16454.00 frames. ], tot_loss[loss=0.2063, simple_loss=0.2939, pruned_loss=0.05936, over 3067978.02 frames. ], batch size: 146, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:55:55,709 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1243, 2.2437, 2.2620, 3.7827, 2.1469, 2.5307, 2.3224, 2.3972], device='cuda:0'), covar=tensor([0.1404, 0.3454, 0.2875, 0.0575, 0.4103, 0.2526, 0.3416, 0.3175], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0449, 0.0367, 0.0325, 0.0433, 0.0516, 0.0420, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 15:56:18,209 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229291.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:33,820 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229300.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:56:37,759 INFO [train.py:904] (0/8) Epoch 23, batch 6000, loss[loss=0.1943, simple_loss=0.2807, pruned_loss=0.05395, over 16839.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2926, pruned_loss=0.05926, over 3062031.16 frames. ], batch size: 102, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:56:37,760 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 15:56:49,494 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 15:57:01,397 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7553, 2.4287, 2.3212, 3.2665, 2.3321, 3.6068, 1.5200, 2.7452], device='cuda:0'), covar=tensor([0.1364, 0.0814, 0.1287, 0.0203, 0.0179, 0.0371, 0.1768, 0.0805], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0176, 0.0197, 0.0193, 0.0207, 0.0216, 0.0204, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 15:57:07,756 INFO [optim.py:368] (0/8) 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 15:58:06,106 INFO [train.py:904] (0/8) Epoch 23, batch 6050, loss[loss=0.1983, simple_loss=0.2974, pruned_loss=0.04956, over 16645.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.2919, pruned_loss=0.05925, over 3058215.96 frames. ], batch size: 89, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:58:20,413 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229361.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 15:59:05,209 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.78 vs. limit=2.0 2023-05-01 15:59:19,853 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 15:59:21,996 INFO [train.py:904] (0/8) Epoch 23, batch 6100, loss[loss=0.184, simple_loss=0.2777, pruned_loss=0.04518, over 16434.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.292, pruned_loss=0.05864, over 3059168.06 frames. ], batch size: 68, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 15:59:40,512 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.924e+02 2.664e+02 3.486e+02 4.048e+02 8.551e+02, threshold=6.973e+02, percent-clipped=3.0 2023-05-01 16:00:22,340 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229443.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:00:36,343 INFO [train.py:904] (0/8) Epoch 23, batch 6150, loss[loss=0.2182, simple_loss=0.3014, pruned_loss=0.06747, over 15449.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2898, pruned_loss=0.05813, over 3059694.85 frames. ], batch size: 190, lr: 2.93e-03, grad_scale: 8.0 2023-05-01 16:01:15,181 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229477.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:01:53,549 INFO [train.py:904] (0/8) Epoch 23, batch 6200, loss[loss=0.1619, simple_loss=0.2586, pruned_loss=0.0326, over 16661.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2874, pruned_loss=0.05749, over 3051571.37 frames. ], batch size: 89, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:01:55,806 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.788e+02 3.260e+02 3.956e+02 6.763e+02, threshold=6.520e+02, percent-clipped=0.0 2023-05-01 16:02:49,418 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:03:11,724 INFO [train.py:904] (0/8) Epoch 23, batch 6250, loss[loss=0.1979, simple_loss=0.2945, pruned_loss=0.0506, over 16613.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2864, pruned_loss=0.05682, over 3068567.15 frames. ], batch size: 68, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:03:17,468 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0431, 3.0448, 1.8706, 3.2872, 2.3162, 3.2920, 2.0834, 2.5281], device='cuda:0'), covar=tensor([0.0291, 0.0425, 0.1617, 0.0258, 0.0872, 0.0644, 0.1493, 0.0781], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0165, 0.0177, 0.0217, 0.0202, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:03:22,983 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3145, 3.5681, 3.8021, 2.2702, 3.2828, 2.6482, 3.8147, 3.8393], device='cuda:0'), covar=tensor([0.0207, 0.0767, 0.0538, 0.1927, 0.0743, 0.0906, 0.0515, 0.0815], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0154, 0.0147, 0.0131, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:03:48,902 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:09,699 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:04:28,582 INFO [train.py:904] (0/8) Epoch 23, batch 6300, loss[loss=0.1895, simple_loss=0.2796, pruned_loss=0.0497, over 16812.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2859, pruned_loss=0.05569, over 3090385.17 frames. ], batch size: 116, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:04:29,013 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4478, 4.4301, 4.3137, 3.5211, 4.3454, 1.6965, 4.1184, 3.9129], device='cuda:0'), covar=tensor([0.0119, 0.0106, 0.0202, 0.0369, 0.0103, 0.2843, 0.0139, 0.0264], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0160, 0.0201, 0.0179, 0.0178, 0.0208, 0.0189, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:04:50,640 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.872e+02 2.619e+02 3.224e+02 4.167e+02 1.378e+03, threshold=6.449e+02, percent-clipped=8.0 2023-05-01 16:05:27,303 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:28,547 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=229639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:41,504 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3516, 3.2496, 2.6566, 2.1133, 2.2108, 2.1946, 3.3613, 2.9798], device='cuda:0'), covar=tensor([0.3156, 0.0707, 0.1803, 0.2795, 0.2655, 0.2279, 0.0575, 0.1389], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0270, 0.0306, 0.0316, 0.0298, 0.0262, 0.0299, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 16:05:48,017 INFO [train.py:904] (0/8) Epoch 23, batch 6350, loss[loss=0.2032, simple_loss=0.2845, pruned_loss=0.06089, over 16688.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2874, pruned_loss=0.05701, over 3081758.80 frames. ], batch size: 62, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:05:48,993 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 16:05:49,834 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3633, 3.3672, 3.4082, 3.5019, 3.5180, 3.2973, 3.5024, 3.5724], device='cuda:0'), covar=tensor([0.1405, 0.1032, 0.1114, 0.0758, 0.0840, 0.2113, 0.1208, 0.0973], device='cuda:0'), in_proj_covar=tensor([0.0635, 0.0787, 0.0906, 0.0792, 0.0604, 0.0631, 0.0653, 0.0758], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:05:53,060 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229656.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:05:57,430 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2637, 3.9891, 3.9637, 2.5153, 3.6260, 4.0112, 3.5926, 2.2889], device='cuda:0'), covar=tensor([0.0607, 0.0051, 0.0054, 0.0445, 0.0093, 0.0103, 0.0105, 0.0460], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0084, 0.0085, 0.0134, 0.0098, 0.0111, 0.0095, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 16:05:57,447 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229659.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:07:04,890 INFO [train.py:904] (0/8) Epoch 23, batch 6400, loss[loss=0.1971, simple_loss=0.2836, pruned_loss=0.05534, over 16695.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2875, pruned_loss=0.05798, over 3080710.52 frames. ], batch size: 134, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:07:24,872 INFO [optim.py:368] (0/8) 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 16:07:32,449 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229720.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 16:08:21,149 INFO [train.py:904] (0/8) Epoch 23, batch 6450, loss[loss=0.1925, simple_loss=0.2983, pruned_loss=0.04332, over 16687.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2878, pruned_loss=0.05784, over 3079508.96 frames. ], batch size: 89, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:08:36,207 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-01 16:09:08,435 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-01 16:09:34,335 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229799.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:09:40,125 INFO [train.py:904] (0/8) Epoch 23, batch 6500, loss[loss=0.1842, simple_loss=0.271, pruned_loss=0.04872, over 16729.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2864, pruned_loss=0.05736, over 3080538.02 frames. ], batch size: 89, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:09:59,296 INFO [optim.py:368] (0/8) 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 16:10:25,445 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=229833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:10:57,910 INFO [train.py:904] (0/8) Epoch 23, batch 6550, loss[loss=0.193, simple_loss=0.3002, pruned_loss=0.04289, over 16786.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.289, pruned_loss=0.05766, over 3098075.27 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:01,909 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229894.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:12:15,584 INFO [train.py:904] (0/8) Epoch 23, batch 6600, loss[loss=0.258, simple_loss=0.3187, pruned_loss=0.09865, over 11369.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2915, pruned_loss=0.05859, over 3091069.18 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:12:35,470 INFO [optim.py:368] (0/8) 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 16:12:43,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0473, 2.1727, 2.2518, 3.7197, 2.0844, 2.5577, 2.3197, 2.2825], device='cuda:0'), covar=tensor([0.1481, 0.3694, 0.3025, 0.0617, 0.4398, 0.2480, 0.3489, 0.3468], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0452, 0.0369, 0.0327, 0.0437, 0.0520, 0.0424, 0.0529], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:13:01,782 INFO [zipformer.py:625] (0/8) 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] (0/8) attn_weights_entropy = tensor([4.8373, 5.1276, 5.2666, 5.0743, 5.1020, 5.6446, 5.1064, 4.8996], device='cuda:0'), covar=tensor([0.1040, 0.1773, 0.1985, 0.1823, 0.2257, 0.0875, 0.1596, 0.2295], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0603, 0.0664, 0.0496, 0.0660, 0.0692, 0.0519, 0.0668], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 16:13:33,873 INFO [train.py:904] (0/8) Epoch 23, batch 6650, loss[loss=0.1918, simple_loss=0.281, pruned_loss=0.05128, over 16329.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2912, pruned_loss=0.05875, over 3095921.51 frames. ], batch size: 146, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:13:37,581 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=229955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:13:38,770 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=229956.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:14,203 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4326, 3.3330, 2.5545, 2.1212, 2.3348, 2.1628, 3.4040, 3.0494], device='cuda:0'), covar=tensor([0.3288, 0.0831, 0.2071, 0.2814, 0.2703, 0.2382, 0.0629, 0.1396], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0269, 0.0305, 0.0315, 0.0297, 0.0262, 0.0299, 0.0340], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 16:14:26,053 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=229987.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:14:45,669 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-230000.pt 2023-05-01 16:14:52,377 INFO [train.py:904] (0/8) Epoch 23, batch 6700, loss[loss=0.2124, simple_loss=0.2902, pruned_loss=0.06727, over 16312.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2904, pruned_loss=0.05939, over 3082041.15 frames. ], batch size: 35, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:14:54,796 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230004.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:15:11,265 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230015.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:15:11,773 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 16:15:12,083 INFO [optim.py:368] (0/8) 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 16:15:23,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4089, 2.9171, 3.1010, 1.9838, 2.7659, 2.0421, 3.0565, 3.1179], device='cuda:0'), covar=tensor([0.0305, 0.0809, 0.0635, 0.2097, 0.0901, 0.1093, 0.0722, 0.0909], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0167, 0.0169, 0.0154, 0.0147, 0.0131, 0.0143, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:16:02,769 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230048.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:16:09,695 INFO [train.py:904] (0/8) Epoch 23, batch 6750, loss[loss=0.1917, simple_loss=0.2792, pruned_loss=0.05207, over 16724.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2893, pruned_loss=0.05935, over 3083193.65 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:16:22,501 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-01 16:16:39,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5735, 3.5415, 3.5194, 2.7751, 3.3969, 2.0499, 3.2230, 2.9558], device='cuda:0'), covar=tensor([0.0152, 0.0139, 0.0201, 0.0242, 0.0102, 0.2364, 0.0134, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0160, 0.0200, 0.0178, 0.0176, 0.0207, 0.0189, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:17:00,542 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8345, 4.6578, 4.9204, 5.0631, 5.2136, 4.6464, 5.2153, 5.2626], device='cuda:0'), covar=tensor([0.2031, 0.1402, 0.1509, 0.0695, 0.0582, 0.0922, 0.0586, 0.0617], device='cuda:0'), in_proj_covar=tensor([0.0628, 0.0781, 0.0898, 0.0789, 0.0599, 0.0624, 0.0649, 0.0754], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:17:19,192 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230099.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:17:24,164 INFO [train.py:904] (0/8) Epoch 23, batch 6800, loss[loss=0.2156, simple_loss=0.3047, pruned_loss=0.06328, over 17056.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2886, pruned_loss=0.05856, over 3120077.73 frames. ], batch size: 55, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:17:43,641 INFO [optim.py:368] (0/8) 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 16:18:11,084 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230133.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:31,930 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:36,419 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230150.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:18:40,529 INFO [train.py:904] (0/8) Epoch 23, batch 6850, loss[loss=0.2538, simple_loss=0.315, pruned_loss=0.09625, over 11691.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2893, pruned_loss=0.05881, over 3111687.75 frames. ], batch size: 246, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:18:51,520 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7821, 3.7427, 3.9298, 3.7541, 3.9051, 4.2545, 3.8901, 3.6086], device='cuda:0'), covar=tensor([0.2102, 0.2444, 0.2739, 0.2350, 0.2748, 0.1876, 0.1858, 0.2911], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0598, 0.0662, 0.0495, 0.0658, 0.0689, 0.0517, 0.0664], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 16:19:23,355 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:19:56,864 INFO [train.py:904] (0/8) Epoch 23, batch 6900, loss[loss=0.1779, simple_loss=0.2745, pruned_loss=0.04068, over 16825.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2914, pruned_loss=0.05871, over 3092204.22 frames. ], batch size: 102, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:20:10,766 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:12,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9750, 4.6427, 4.5857, 3.2619, 4.0260, 4.6052, 4.1295, 2.5447], device='cuda:0'), covar=tensor([0.0467, 0.0055, 0.0052, 0.0347, 0.0100, 0.0126, 0.0094, 0.0428], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0086, 0.0087, 0.0136, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 16:20:20,077 INFO [optim.py:368] (0/8) 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 16:20:45,539 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230233.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:20:58,977 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230241.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:13,084 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:21:17,998 INFO [train.py:904] (0/8) Epoch 23, batch 6950, loss[loss=0.1991, simple_loss=0.2896, pruned_loss=0.05429, over 16498.00 frames. ], tot_loss[loss=0.2073, simple_loss=0.2935, pruned_loss=0.06052, over 3079134.98 frames. ], batch size: 75, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:01,069 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230281.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,184 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230302.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:22:34,888 INFO [train.py:904] (0/8) Epoch 23, batch 7000, loss[loss=0.1952, simple_loss=0.2883, pruned_loss=0.05107, over 16468.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2936, pruned_loss=0.05969, over 3078617.07 frames. ], batch size: 35, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:22:52,335 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1721, 3.2869, 3.0181, 5.3428, 4.0359, 4.5436, 2.0248, 3.5067], device='cuda:0'), covar=tensor([0.1246, 0.0685, 0.1094, 0.0171, 0.0400, 0.0372, 0.1469, 0.0735], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0177, 0.0198, 0.0193, 0.0208, 0.0217, 0.0205, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:22:53,563 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230315.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:22:56,099 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.703e+02 3.306e+02 4.193e+02 6.685e+02, threshold=6.612e+02, percent-clipped=2.0 2023-05-01 16:23:17,272 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 16:23:34,504 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-01 16:23:35,267 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:23:50,814 INFO [train.py:904] (0/8) Epoch 23, batch 7050, loss[loss=0.2201, simple_loss=0.2929, pruned_loss=0.0737, over 11768.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2942, pruned_loss=0.05896, over 3107900.49 frames. ], batch size: 247, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:24:05,845 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230363.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:24:12,811 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 16:25:07,486 INFO [train.py:904] (0/8) Epoch 23, batch 7100, loss[loss=0.2058, simple_loss=0.2909, pruned_loss=0.06033, over 16674.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2928, pruned_loss=0.05876, over 3101477.49 frames. ], batch size: 62, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:25:30,797 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.944e+02 2.925e+02 3.270e+02 3.973e+02 7.550e+02, threshold=6.541e+02, percent-clipped=2.0 2023-05-01 16:26:08,159 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9300, 2.7761, 2.8143, 2.1282, 2.6861, 2.2260, 2.7557, 2.9616], device='cuda:0'), covar=tensor([0.0298, 0.0743, 0.0472, 0.1764, 0.0768, 0.0844, 0.0580, 0.0756], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0166, 0.0169, 0.0155, 0.0147, 0.0131, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:26:24,968 INFO [train.py:904] (0/8) Epoch 23, batch 7150, loss[loss=0.2278, simple_loss=0.3122, pruned_loss=0.07166, over 16960.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2914, pruned_loss=0.05909, over 3097201.65 frames. ], batch size: 41, lr: 2.92e-03, grad_scale: 4.0 2023-05-01 16:26:48,741 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230468.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:27:35,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7212, 4.7456, 4.5589, 3.2977, 4.6388, 1.6052, 4.2914, 4.2374], device='cuda:0'), covar=tensor([0.0168, 0.0144, 0.0273, 0.0810, 0.0159, 0.3772, 0.0226, 0.0395], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0159, 0.0199, 0.0176, 0.0175, 0.0206, 0.0187, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:27:42,216 INFO [train.py:904] (0/8) Epoch 23, batch 7200, loss[loss=0.1923, simple_loss=0.2827, pruned_loss=0.0509, over 17114.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2893, pruned_loss=0.05757, over 3095031.60 frames. ], batch size: 49, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:27:47,316 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 16:28:16,793 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:16,922 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0407, 3.2068, 3.4462, 1.9605, 2.9848, 2.1655, 3.4334, 3.5153], device='cuda:0'), covar=tensor([0.0234, 0.0825, 0.0544, 0.2145, 0.0814, 0.1017, 0.0629, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0130, 0.0144, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:28:22,556 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230529.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:28:29,595 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230533.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:28:58,903 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230550.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:02,534 INFO [train.py:904] (0/8) Epoch 23, batch 7250, loss[loss=0.183, simple_loss=0.2734, pruned_loss=0.0463, over 16667.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2869, pruned_loss=0.05625, over 3090753.63 frames. ], batch size: 76, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:29:51,477 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:29:58,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2301, 5.2400, 5.6663, 5.6442, 5.6706, 5.3458, 5.2498, 4.9925], device='cuda:0'), covar=tensor([0.0355, 0.0549, 0.0386, 0.0388, 0.0547, 0.0379, 0.0948, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0462, 0.0448, 0.0416, 0.0493, 0.0470, 0.0557, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 16:30:03,527 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:08,125 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230597.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:10,083 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230598.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:30:18,075 INFO [train.py:904] (0/8) Epoch 23, batch 7300, loss[loss=0.1843, simple_loss=0.2799, pruned_loss=0.0443, over 16706.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2855, pruned_loss=0.05571, over 3097976.60 frames. ], batch size: 134, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:30:21,122 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 16:30:39,632 INFO [optim.py:368] (0/8) 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 16:30:57,515 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4814, 3.3984, 2.6970, 2.1357, 2.2537, 2.3145, 3.7331, 3.0679], device='cuda:0'), covar=tensor([0.3075, 0.0733, 0.1844, 0.2763, 0.2963, 0.2148, 0.0520, 0.1444], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0271, 0.0308, 0.0318, 0.0300, 0.0263, 0.0300, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 16:30:59,190 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 16:31:19,788 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230643.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:31:33,671 INFO [train.py:904] (0/8) Epoch 23, batch 7350, loss[loss=0.1847, simple_loss=0.2738, pruned_loss=0.04782, over 17274.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2865, pruned_loss=0.05667, over 3086019.20 frames. ], batch size: 52, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:31:34,906 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6251, 1.8547, 2.2906, 2.5671, 2.5520, 3.0446, 2.0272, 2.9056], device='cuda:0'), covar=tensor([0.0248, 0.0541, 0.0321, 0.0361, 0.0354, 0.0168, 0.0537, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0190, 0.0193, 0.0179, 0.0185, 0.0199, 0.0155, 0.0197, 0.0154], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:32:33,457 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230691.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:32:53,501 INFO [train.py:904] (0/8) Epoch 23, batch 7400, loss[loss=0.2083, simple_loss=0.295, pruned_loss=0.06077, over 16260.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2881, pruned_loss=0.05751, over 3084118.33 frames. ], batch size: 35, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:33:16,063 INFO [optim.py:368] (0/8) 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 16:34:13,435 INFO [train.py:904] (0/8) Epoch 23, batch 7450, loss[loss=0.1883, simple_loss=0.283, pruned_loss=0.04684, over 16885.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2891, pruned_loss=0.05835, over 3079369.11 frames. ], batch size: 116, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:34:59,218 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1062, 3.4244, 3.4867, 2.3003, 3.2158, 3.5319, 3.2772, 1.9636], device='cuda:0'), covar=tensor([0.0591, 0.0089, 0.0076, 0.0432, 0.0120, 0.0125, 0.0107, 0.0533], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0086, 0.0086, 0.0135, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 16:35:18,604 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:35,344 INFO [train.py:904] (0/8) Epoch 23, batch 7500, loss[loss=0.1724, simple_loss=0.2663, pruned_loss=0.03927, over 16779.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2891, pruned_loss=0.05748, over 3075050.66 frames. ], batch size: 83, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:35:39,165 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0645, 3.1144, 2.7613, 2.9888, 3.4950, 3.0572, 3.6228, 3.6229], device='cuda:0'), covar=tensor([0.0107, 0.0406, 0.0457, 0.0379, 0.0259, 0.0379, 0.0252, 0.0227], device='cuda:0'), in_proj_covar=tensor([0.0211, 0.0233, 0.0224, 0.0226, 0.0235, 0.0232, 0.0232, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:35:41,005 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230806.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:35:58,126 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.010e+02 2.839e+02 3.579e+02 4.414e+02 1.002e+03, threshold=7.158e+02, percent-clipped=2.0 2023-05-01 16:36:09,661 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230824.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 16:36:52,035 INFO [train.py:904] (0/8) Epoch 23, batch 7550, loss[loss=0.188, simple_loss=0.2804, pruned_loss=0.04774, over 16758.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2886, pruned_loss=0.05794, over 3062531.05 frames. ], batch size: 89, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:36:52,479 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:36:53,596 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230854.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:01,932 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4440, 3.4848, 2.0484, 3.9558, 2.6723, 3.9197, 2.3401, 2.8140], device='cuda:0'), covar=tensor([0.0319, 0.0419, 0.1807, 0.0242, 0.0895, 0.0633, 0.1544, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0176, 0.0194, 0.0164, 0.0176, 0.0217, 0.0203, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:37:33,413 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230881.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:45,191 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=230889.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:47,080 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=230890.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:37:58,364 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=230897.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:38:06,501 INFO [train.py:904] (0/8) Epoch 23, batch 7600, loss[loss=0.1856, simple_loss=0.2785, pruned_loss=0.04639, over 16745.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2875, pruned_loss=0.05766, over 3071578.90 frames. ], batch size: 89, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:38:27,920 INFO [optim.py:368] (0/8) 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 16:39:10,332 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=230945.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:19,811 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=230951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:39:21,943 INFO [train.py:904] (0/8) Epoch 23, batch 7650, loss[loss=0.1873, simple_loss=0.2715, pruned_loss=0.05159, over 17268.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2877, pruned_loss=0.05833, over 3074622.14 frames. ], batch size: 52, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:39:26,293 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 16:39:53,070 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 16:40:12,986 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9245, 2.1146, 2.2880, 3.4429, 2.0319, 2.3786, 2.2361, 2.2443], device='cuda:0'), covar=tensor([0.1473, 0.3730, 0.2895, 0.0675, 0.4326, 0.2554, 0.3522, 0.3366], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0449, 0.0367, 0.0324, 0.0433, 0.0515, 0.0420, 0.0524], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:40:35,937 INFO [train.py:904] (0/8) Epoch 23, batch 7700, loss[loss=0.1863, simple_loss=0.2831, pruned_loss=0.04477, over 16849.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2884, pruned_loss=0.05946, over 3053151.27 frames. ], batch size: 102, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:40:43,305 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8483, 2.9045, 2.6187, 4.7475, 3.5917, 4.2428, 1.7071, 3.1851], device='cuda:0'), covar=tensor([0.1353, 0.0769, 0.1252, 0.0177, 0.0295, 0.0366, 0.1662, 0.0778], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0178, 0.0199, 0.0194, 0.0209, 0.0218, 0.0206, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:40:57,693 INFO [optim.py:368] (0/8) 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 16:41:53,924 INFO [train.py:904] (0/8) Epoch 23, batch 7750, loss[loss=0.2163, simple_loss=0.2988, pruned_loss=0.06692, over 16752.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2883, pruned_loss=0.0593, over 3034652.34 frames. ], batch size: 124, lr: 2.92e-03, grad_scale: 8.0 2023-05-01 16:43:09,657 INFO [train.py:904] (0/8) Epoch 23, batch 7800, loss[loss=0.1875, simple_loss=0.2762, pruned_loss=0.04937, over 16433.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2885, pruned_loss=0.05918, over 3062373.97 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:43:30,322 INFO [optim.py:368] (0/8) 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 16:43:41,695 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231124.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 16:43:57,922 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3731, 4.6718, 4.4634, 4.4770, 4.1833, 4.1491, 4.1935, 4.7088], device='cuda:0'), covar=tensor([0.1193, 0.0873, 0.0990, 0.0947, 0.0896, 0.1668, 0.1164, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0675, 0.0817, 0.0680, 0.0629, 0.0520, 0.0532, 0.0691, 0.0647], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:44:17,276 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:24,905 INFO [train.py:904] (0/8) Epoch 23, batch 7850, loss[loss=0.2027, simple_loss=0.2924, pruned_loss=0.05646, over 16873.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2895, pruned_loss=0.05905, over 3076343.03 frames. ], batch size: 116, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:44:52,238 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:44:52,331 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0642, 4.1012, 4.4008, 4.3751, 4.4103, 4.1256, 4.1288, 4.0562], device='cuda:0'), covar=tensor([0.0383, 0.0607, 0.0465, 0.0456, 0.0491, 0.0458, 0.1005, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0459, 0.0445, 0.0414, 0.0490, 0.0468, 0.0553, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 16:45:05,351 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231181.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:16,866 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231189.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:45:37,209 INFO [train.py:904] (0/8) Epoch 23, batch 7900, loss[loss=0.203, simple_loss=0.2892, pruned_loss=0.05839, over 17143.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2884, pruned_loss=0.05804, over 3087355.38 frames. ], batch size: 47, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:45:57,448 INFO [optim.py:368] (0/8) 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 16:45:58,562 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.77 vs. limit=5.0 2023-05-01 16:46:16,874 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231229.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:31,296 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231237.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:44,142 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231246.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:52,152 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231251.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:46:54,846 INFO [train.py:904] (0/8) Epoch 23, batch 7950, loss[loss=0.1847, simple_loss=0.2747, pruned_loss=0.04739, over 16713.00 frames. ], tot_loss[loss=0.2032, simple_loss=0.289, pruned_loss=0.05866, over 3085005.79 frames. ], batch size: 89, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:11,522 INFO [train.py:904] (0/8) Epoch 23, batch 8000, loss[loss=0.2205, simple_loss=0.3057, pruned_loss=0.06766, over 15485.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2896, pruned_loss=0.05909, over 3090865.40 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:48:15,236 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1408, 2.5024, 2.0595, 2.2837, 2.8383, 2.4616, 2.7915, 3.0006], device='cuda:0'), covar=tensor([0.0162, 0.0415, 0.0552, 0.0475, 0.0271, 0.0377, 0.0262, 0.0253], device='cuda:0'), in_proj_covar=tensor([0.0211, 0.0233, 0.0224, 0.0227, 0.0234, 0.0231, 0.0232, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:48:25,848 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231312.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:48:33,454 INFO [optim.py:368] (0/8) 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 16:48:50,549 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8690, 2.6603, 2.4161, 4.1606, 2.7228, 3.8350, 1.6635, 2.8667], device='cuda:0'), covar=tensor([0.1390, 0.0922, 0.1441, 0.0228, 0.0346, 0.0533, 0.1808, 0.0982], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0175, 0.0196, 0.0191, 0.0207, 0.0215, 0.0204, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 16:49:22,934 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 16:49:26,919 INFO [train.py:904] (0/8) Epoch 23, batch 8050, loss[loss=0.2013, simple_loss=0.2921, pruned_loss=0.05518, over 15344.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2898, pruned_loss=0.05881, over 3093734.33 frames. ], batch size: 190, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:49:38,369 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9287, 2.7730, 2.6743, 2.0109, 2.6461, 2.7610, 2.6266, 1.9117], device='cuda:0'), covar=tensor([0.0406, 0.0094, 0.0082, 0.0337, 0.0126, 0.0127, 0.0124, 0.0414], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0085, 0.0085, 0.0134, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 16:50:14,160 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 16:50:31,572 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-01 16:50:42,760 INFO [train.py:904] (0/8) Epoch 23, batch 8100, loss[loss=0.1853, simple_loss=0.271, pruned_loss=0.04976, over 17028.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2896, pruned_loss=0.05848, over 3092913.36 frames. ], batch size: 55, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:51:04,322 INFO [optim.py:368] (0/8) 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 16:51:51,133 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:51:59,311 INFO [train.py:904] (0/8) Epoch 23, batch 8150, loss[loss=0.2149, simple_loss=0.2862, pruned_loss=0.07173, over 11504.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.287, pruned_loss=0.05717, over 3104785.19 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:05,172 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:53:14,396 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4929, 3.5426, 2.7943, 2.1587, 2.3101, 2.3551, 3.7379, 3.1710], device='cuda:0'), covar=tensor([0.3091, 0.0604, 0.1763, 0.2969, 0.2689, 0.2187, 0.0484, 0.1325], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0270, 0.0309, 0.0318, 0.0301, 0.0264, 0.0299, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 16:53:14,922 INFO [train.py:904] (0/8) Epoch 23, batch 8200, loss[loss=0.186, simple_loss=0.2814, pruned_loss=0.04529, over 16384.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2844, pruned_loss=0.05695, over 3100706.89 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:53:38,116 INFO [optim.py:368] (0/8) 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 16:54:26,265 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:54:37,504 INFO [train.py:904] (0/8) Epoch 23, batch 8250, loss[loss=0.1629, simple_loss=0.2519, pruned_loss=0.03692, over 12229.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2837, pruned_loss=0.05473, over 3073500.37 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:55:42,907 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231594.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:55:57,542 INFO [train.py:904] (0/8) Epoch 23, batch 8300, loss[loss=0.1767, simple_loss=0.261, pruned_loss=0.04621, over 11936.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2808, pruned_loss=0.0517, over 3067520.10 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:56:04,249 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=231607.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:56:17,506 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3669, 3.2406, 3.2549, 3.4910, 3.5122, 3.2755, 3.4468, 3.5620], device='cuda:0'), covar=tensor([0.1597, 0.1494, 0.1787, 0.1008, 0.1087, 0.3190, 0.1536, 0.1296], device='cuda:0'), in_proj_covar=tensor([0.0627, 0.0782, 0.0897, 0.0787, 0.0602, 0.0626, 0.0655, 0.0762], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:56:22,159 INFO [optim.py:368] (0/8) 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 16:56:49,678 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0063, 1.8285, 1.6709, 1.4614, 1.9812, 1.6408, 1.6090, 1.8905], device='cuda:0'), covar=tensor([0.0196, 0.0297, 0.0416, 0.0363, 0.0229, 0.0283, 0.0162, 0.0200], device='cuda:0'), in_proj_covar=tensor([0.0208, 0.0230, 0.0222, 0.0223, 0.0231, 0.0229, 0.0229, 0.0227], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 16:57:11,283 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=231647.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 16:57:19,288 INFO [train.py:904] (0/8) Epoch 23, batch 8350, loss[loss=0.2004, simple_loss=0.2834, pruned_loss=0.0587, over 11975.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2799, pruned_loss=0.04991, over 3045592.12 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 16:58:41,948 INFO [train.py:904] (0/8) Epoch 23, batch 8400, loss[loss=0.1746, simple_loss=0.2749, pruned_loss=0.03718, over 16897.00 frames. ], tot_loss[loss=0.1865, simple_loss=0.2775, pruned_loss=0.04775, over 3042157.03 frames. ], batch size: 116, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 16:58:46,560 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 16:58:50,846 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=231708.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 16:59:06,595 INFO [optim.py:368] (0/8) 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 16:59:44,191 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2250, 4.1801, 4.0828, 3.3319, 4.1339, 1.6766, 3.9401, 3.8491], device='cuda:0'), covar=tensor([0.0120, 0.0155, 0.0212, 0.0372, 0.0130, 0.3099, 0.0167, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0159, 0.0200, 0.0177, 0.0176, 0.0208, 0.0188, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:00:04,867 INFO [train.py:904] (0/8) Epoch 23, batch 8450, loss[loss=0.1659, simple_loss=0.269, pruned_loss=0.03137, over 16864.00 frames. ], tot_loss[loss=0.184, simple_loss=0.2758, pruned_loss=0.04609, over 3050806.66 frames. ], batch size: 96, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:24,909 INFO [train.py:904] (0/8) Epoch 23, batch 8500, loss[loss=0.1644, simple_loss=0.2567, pruned_loss=0.03611, over 16435.00 frames. ], tot_loss[loss=0.1799, simple_loss=0.2721, pruned_loss=0.04384, over 3047774.93 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:01:48,516 INFO [optim.py:368] (0/8) 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 17:02:48,231 INFO [train.py:904] (0/8) Epoch 23, batch 8550, loss[loss=0.1718, simple_loss=0.2708, pruned_loss=0.03643, over 15208.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2694, pruned_loss=0.04296, over 3023532.19 frames. ], batch size: 190, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:02:58,122 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0541, 3.0751, 1.8719, 3.3187, 2.2538, 3.3182, 2.1559, 2.6435], device='cuda:0'), covar=tensor([0.0327, 0.0380, 0.1668, 0.0283, 0.0900, 0.0535, 0.1470, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0173, 0.0192, 0.0161, 0.0175, 0.0213, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 17:04:28,273 INFO [train.py:904] (0/8) Epoch 23, batch 8600, loss[loss=0.1588, simple_loss=0.2539, pruned_loss=0.03186, over 16853.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2695, pruned_loss=0.04164, over 3033832.75 frames. ], batch size: 42, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:04:36,652 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=231907.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:05:00,383 INFO [optim.py:368] (0/8) 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 17:05:06,684 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.95 vs. limit=2.0 2023-05-01 17:06:03,603 INFO [train.py:904] (0/8) Epoch 23, batch 8650, loss[loss=0.1575, simple_loss=0.256, pruned_loss=0.02949, over 15289.00 frames. ], tot_loss[loss=0.1744, simple_loss=0.2678, pruned_loss=0.04043, over 3021167.98 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:06:09,851 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=231955.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:07:40,221 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9670, 2.7492, 2.9307, 2.1260, 2.7565, 2.2063, 2.7528, 2.9275], device='cuda:0'), covar=tensor([0.0256, 0.0836, 0.0440, 0.1715, 0.0716, 0.0904, 0.0548, 0.0717], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0160, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 17:07:43,535 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-232000.pt 2023-05-01 17:07:53,209 INFO [train.py:904] (0/8) Epoch 23, batch 8700, loss[loss=0.1724, simple_loss=0.2652, pruned_loss=0.03977, over 16958.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2656, pruned_loss=0.03923, over 3041169.15 frames. ], batch size: 109, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:07:54,279 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232003.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 17:08:23,063 INFO [optim.py:368] (0/8) 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 17:09:27,793 INFO [train.py:904] (0/8) Epoch 23, batch 8750, loss[loss=0.1629, simple_loss=0.2525, pruned_loss=0.03667, over 12415.00 frames. ], tot_loss[loss=0.171, simple_loss=0.2653, pruned_loss=0.03836, over 3060319.62 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:09:42,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5156, 3.5958, 2.7767, 2.1543, 2.3485, 2.3848, 3.8313, 3.1308], device='cuda:0'), covar=tensor([0.3141, 0.0691, 0.1889, 0.3016, 0.2732, 0.2295, 0.0473, 0.1595], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0263, 0.0302, 0.0311, 0.0292, 0.0259, 0.0293, 0.0333], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:11:15,146 INFO [train.py:904] (0/8) Epoch 23, batch 8800, loss[loss=0.1823, simple_loss=0.2704, pruned_loss=0.04712, over 12655.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2637, pruned_loss=0.03721, over 3077306.91 frames. ], batch size: 248, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:11:29,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9989, 4.2773, 4.1270, 4.1284, 3.8288, 3.8055, 3.9061, 4.2862], device='cuda:0'), covar=tensor([0.1046, 0.0904, 0.0911, 0.0782, 0.0790, 0.1863, 0.1036, 0.0892], device='cuda:0'), in_proj_covar=tensor([0.0666, 0.0810, 0.0670, 0.0620, 0.0513, 0.0527, 0.0681, 0.0636], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:11:46,654 INFO [optim.py:368] (0/8) 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 17:12:57,923 INFO [train.py:904] (0/8) Epoch 23, batch 8850, loss[loss=0.17, simple_loss=0.2752, pruned_loss=0.03245, over 15240.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2661, pruned_loss=0.03662, over 3076587.10 frames. ], batch size: 191, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:14:42,077 INFO [train.py:904] (0/8) Epoch 23, batch 8900, loss[loss=0.1787, simple_loss=0.2798, pruned_loss=0.03885, over 16947.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2669, pruned_loss=0.03647, over 3080946.82 frames. ], batch size: 96, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:15:12,042 INFO [optim.py:368] (0/8) 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 17:16:33,781 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:16:45,335 INFO [train.py:904] (0/8) Epoch 23, batch 8950, loss[loss=0.2026, simple_loss=0.2942, pruned_loss=0.05547, over 12810.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2668, pruned_loss=0.03681, over 3093588.87 frames. ], batch size: 247, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:17:42,294 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.83 vs. limit=2.0 2023-05-01 17:18:31,855 INFO [train.py:904] (0/8) Epoch 23, batch 9000, loss[loss=0.1575, simple_loss=0.2518, pruned_loss=0.03161, over 16371.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2634, pruned_loss=0.03556, over 3088792.79 frames. ], batch size: 146, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:18:31,856 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 17:18:42,674 INFO [train.py:938] (0/8) Epoch 23, validation: loss=0.1452, simple_loss=0.249, pruned_loss=0.02066, over 944034.00 frames. 2023-05-01 17:18:42,675 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 17:18:44,548 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232303.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 17:18:54,208 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232308.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:19:18,068 INFO [optim.py:368] (0/8) 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 17:19:33,768 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7897, 3.1611, 3.3681, 1.9420, 2.8818, 2.1851, 3.3154, 3.3974], device='cuda:0'), covar=tensor([0.0296, 0.0819, 0.0557, 0.2287, 0.0880, 0.1126, 0.0650, 0.1017], device='cuda:0'), in_proj_covar=tensor([0.0152, 0.0160, 0.0164, 0.0151, 0.0142, 0.0127, 0.0140, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 17:19:44,743 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8986, 3.8824, 4.1628, 4.1449, 4.1423, 3.9537, 3.9694, 3.9910], device='cuda:0'), covar=tensor([0.0317, 0.0728, 0.0443, 0.0438, 0.0437, 0.0461, 0.0671, 0.0397], device='cuda:0'), in_proj_covar=tensor([0.0403, 0.0445, 0.0437, 0.0403, 0.0478, 0.0453, 0.0534, 0.0365], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 17:20:23,946 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232351.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:20:28,125 INFO [train.py:904] (0/8) Epoch 23, batch 9050, loss[loss=0.1676, simple_loss=0.2586, pruned_loss=0.03825, over 12484.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2648, pruned_loss=0.03623, over 3093746.52 frames. ], batch size: 246, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:21:23,474 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:22:12,823 INFO [train.py:904] (0/8) Epoch 23, batch 9100, loss[loss=0.1802, simple_loss=0.2797, pruned_loss=0.04033, over 16675.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2639, pruned_loss=0.03677, over 3093463.94 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:22:46,303 INFO [optim.py:368] (0/8) 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 17:22:58,560 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4781, 3.5956, 2.0648, 4.0136, 2.5168, 3.9175, 2.2827, 2.8739], device='cuda:0'), covar=tensor([0.0330, 0.0378, 0.1799, 0.0211, 0.0934, 0.0526, 0.1567, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0170, 0.0189, 0.0158, 0.0171, 0.0208, 0.0198, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 17:23:46,398 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232442.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:24:07,394 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6699, 2.6557, 1.8487, 2.8078, 2.0356, 2.8121, 2.0998, 2.3579], device='cuda:0'), covar=tensor([0.0332, 0.0395, 0.1395, 0.0234, 0.0823, 0.0469, 0.1330, 0.0699], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0171, 0.0189, 0.0158, 0.0171, 0.0209, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 17:24:09,380 INFO [train.py:904] (0/8) Epoch 23, batch 9150, loss[loss=0.1819, simple_loss=0.2732, pruned_loss=0.04536, over 16237.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2645, pruned_loss=0.03689, over 3069847.28 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 4.0 2023-05-01 17:25:00,465 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:25:52,936 INFO [train.py:904] (0/8) Epoch 23, batch 9200, loss[loss=0.182, simple_loss=0.2691, pruned_loss=0.04747, over 16671.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2604, pruned_loss=0.03601, over 3060687.18 frames. ], batch size: 134, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:26:17,512 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2221, 2.1691, 2.2254, 3.8788, 2.1071, 2.4988, 2.2856, 2.2735], device='cuda:0'), covar=tensor([0.1278, 0.3781, 0.3271, 0.0530, 0.4579, 0.2701, 0.3633, 0.3863], device='cuda:0'), in_proj_covar=tensor([0.0394, 0.0443, 0.0364, 0.0318, 0.0428, 0.0506, 0.0414, 0.0515], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:26:21,345 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232518.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:26:24,119 INFO [optim.py:368] (0/8) 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 17:26:43,882 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5216, 4.6225, 4.7782, 4.5733, 4.6836, 5.1642, 4.7285, 4.4138], device='cuda:0'), covar=tensor([0.1314, 0.1892, 0.2259, 0.2071, 0.2657, 0.1004, 0.1557, 0.2486], device='cuda:0'), in_proj_covar=tensor([0.0395, 0.0576, 0.0637, 0.0476, 0.0634, 0.0664, 0.0497, 0.0636], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:26:57,041 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232537.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:27:13,318 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 17:27:27,737 INFO [train.py:904] (0/8) Epoch 23, batch 9250, loss[loss=0.1523, simple_loss=0.2533, pruned_loss=0.02565, over 16200.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2602, pruned_loss=0.03607, over 3043167.71 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:28:01,783 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-01 17:28:23,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3903, 1.8237, 2.1766, 2.3686, 2.4631, 2.6054, 1.9441, 2.5854], device='cuda:0'), covar=tensor([0.0247, 0.0548, 0.0313, 0.0355, 0.0334, 0.0227, 0.0519, 0.0165], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0188, 0.0175, 0.0178, 0.0193, 0.0151, 0.0193, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:28:23,261 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=232579.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:17,467 INFO [train.py:904] (0/8) Epoch 23, batch 9300, loss[loss=0.1699, simple_loss=0.2611, pruned_loss=0.03941, over 16878.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2586, pruned_loss=0.03529, over 3044150.76 frames. ], batch size: 124, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:29:18,381 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232603.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:29:46,520 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.95 vs. limit=5.0 2023-05-01 17:29:58,639 INFO [optim.py:368] (0/8) 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 17:30:01,879 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3239, 3.4176, 2.0120, 3.7042, 2.5077, 3.6873, 2.2055, 2.8125], device='cuda:0'), covar=tensor([0.0311, 0.0391, 0.1671, 0.0344, 0.0885, 0.0606, 0.1543, 0.0777], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0171, 0.0189, 0.0158, 0.0171, 0.0208, 0.0199, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 17:30:20,333 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5693, 4.8762, 4.6799, 4.7032, 4.4096, 4.3657, 4.3456, 4.9278], device='cuda:0'), covar=tensor([0.1025, 0.0828, 0.0919, 0.0772, 0.0743, 0.1334, 0.1065, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0660, 0.0801, 0.0662, 0.0614, 0.0509, 0.0522, 0.0674, 0.0633], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:31:04,798 INFO [train.py:904] (0/8) Epoch 23, batch 9350, loss[loss=0.1811, simple_loss=0.2758, pruned_loss=0.04319, over 16161.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2583, pruned_loss=0.03519, over 3050081.50 frames. ], batch size: 165, lr: 2.91e-03, grad_scale: 8.0 2023-05-01 17:32:47,941 INFO [train.py:904] (0/8) Epoch 23, batch 9400, loss[loss=0.1747, simple_loss=0.2728, pruned_loss=0.03824, over 16116.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2581, pruned_loss=0.03493, over 3043717.20 frames. ], batch size: 165, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:33:21,843 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.567e+02 2.048e+02 2.308e+02 2.875e+02 5.038e+02, threshold=4.615e+02, percent-clipped=1.0 2023-05-01 17:33:57,953 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232737.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 17:34:29,655 INFO [train.py:904] (0/8) Epoch 23, batch 9450, loss[loss=0.1665, simple_loss=0.2644, pruned_loss=0.03435, over 15374.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2595, pruned_loss=0.03511, over 3023374.92 frames. ], batch size: 191, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:09,225 INFO [train.py:904] (0/8) Epoch 23, batch 9500, loss[loss=0.1647, simple_loss=0.2613, pruned_loss=0.03401, over 15452.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2588, pruned_loss=0.03485, over 3024023.62 frames. ], batch size: 192, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:36:44,545 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.549e+02 2.194e+02 2.525e+02 3.286e+02 6.197e+02, threshold=5.051e+02, percent-clipped=6.0 2023-05-01 17:37:07,740 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232832.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:37:11,987 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3650, 3.7603, 3.7955, 2.5908, 3.4422, 3.7779, 3.5604, 2.1245], device='cuda:0'), covar=tensor([0.0548, 0.0054, 0.0061, 0.0412, 0.0116, 0.0119, 0.0092, 0.0564], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0083, 0.0083, 0.0131, 0.0096, 0.0107, 0.0093, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:37:39,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2991, 3.6744, 3.7609, 2.5335, 3.3893, 3.7421, 3.5541, 2.2659], device='cuda:0'), covar=tensor([0.0521, 0.0058, 0.0050, 0.0396, 0.0108, 0.0100, 0.0083, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0082, 0.0083, 0.0131, 0.0096, 0.0107, 0.0092, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:37:47,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7275, 3.9829, 2.8777, 2.2422, 2.4774, 2.4531, 4.2392, 3.3631], device='cuda:0'), covar=tensor([0.3030, 0.0594, 0.1894, 0.3177, 0.3210, 0.2171, 0.0403, 0.1503], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0261, 0.0299, 0.0307, 0.0285, 0.0256, 0.0288, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:37:54,052 INFO [train.py:904] (0/8) Epoch 23, batch 9550, loss[loss=0.2074, simple_loss=0.2956, pruned_loss=0.05961, over 16852.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2586, pruned_loss=0.03482, over 3055620.68 frames. ], batch size: 116, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:38:37,462 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=232874.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:39:33,908 INFO [train.py:904] (0/8) Epoch 23, batch 9600, loss[loss=0.1519, simple_loss=0.2519, pruned_loss=0.02594, over 16512.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2595, pruned_loss=0.03538, over 3059733.98 frames. ], batch size: 68, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:39:34,660 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=232903.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:40:05,775 INFO [optim.py:368] (0/8) 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 17:40:06,982 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6426, 2.6599, 1.8199, 2.8161, 2.0746, 2.8134, 2.1332, 2.3838], device='cuda:0'), covar=tensor([0.0308, 0.0355, 0.1339, 0.0269, 0.0708, 0.0468, 0.1249, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0169, 0.0187, 0.0156, 0.0170, 0.0205, 0.0197, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-01 17:41:17,235 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=232951.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:41:20,705 INFO [train.py:904] (0/8) Epoch 23, batch 9650, loss[loss=0.1666, simple_loss=0.254, pruned_loss=0.03964, over 12354.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2613, pruned_loss=0.0358, over 3037991.49 frames. ], batch size: 248, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:41:55,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8040, 2.0168, 2.4286, 2.7661, 2.6783, 3.2103, 2.1493, 3.1410], device='cuda:0'), covar=tensor([0.0252, 0.0530, 0.0372, 0.0322, 0.0336, 0.0190, 0.0536, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0188, 0.0174, 0.0177, 0.0193, 0.0150, 0.0192, 0.0147], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:42:41,589 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=232990.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:09,194 INFO [train.py:904] (0/8) Epoch 23, batch 9700, loss[loss=0.1488, simple_loss=0.2487, pruned_loss=0.02443, over 16545.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.26, pruned_loss=0.0355, over 3038212.61 frames. ], batch size: 75, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:43:40,114 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:43:43,787 INFO [optim.py:368] (0/8) 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 17:43:59,477 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 17:44:22,000 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:44:50,787 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233051.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 17:44:53,553 INFO [train.py:904] (0/8) Epoch 23, batch 9750, loss[loss=0.1788, simple_loss=0.273, pruned_loss=0.04228, over 16221.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2593, pruned_loss=0.03586, over 3029483.34 frames. ], batch size: 146, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:45:43,932 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:45:59,984 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233085.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:46:21,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3710, 4.2130, 4.4556, 4.5455, 4.7471, 4.3039, 4.7499, 4.7691], device='cuda:0'), covar=tensor([0.1827, 0.1364, 0.1547, 0.0818, 0.0492, 0.1106, 0.0524, 0.0680], device='cuda:0'), in_proj_covar=tensor([0.0614, 0.0758, 0.0872, 0.0769, 0.0584, 0.0608, 0.0636, 0.0741], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:46:31,585 INFO [train.py:904] (0/8) Epoch 23, batch 9800, loss[loss=0.144, simple_loss=0.2353, pruned_loss=0.0264, over 12139.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2595, pruned_loss=0.03479, over 3060975.52 frames. ], batch size: 249, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:47:03,695 INFO [optim.py:368] (0/8) 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 17:47:26,805 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233132.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:47:37,983 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9224, 2.0615, 2.4341, 2.9193, 2.7309, 3.3144, 2.3301, 3.3036], device='cuda:0'), covar=tensor([0.0250, 0.0529, 0.0398, 0.0314, 0.0333, 0.0206, 0.0481, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0188, 0.0174, 0.0178, 0.0193, 0.0150, 0.0192, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:47:43,084 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.30 vs. limit=5.0 2023-05-01 17:48:16,813 INFO [train.py:904] (0/8) Epoch 23, batch 9850, loss[loss=0.1805, simple_loss=0.278, pruned_loss=0.04154, over 16651.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2612, pruned_loss=0.03481, over 3060041.49 frames. ], batch size: 134, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:49:01,105 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233174.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:49:13,856 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233180.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:49:29,075 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9395, 2.1298, 2.4227, 3.2247, 2.1905, 2.3314, 2.3332, 2.1917], device='cuda:0'), covar=tensor([0.1407, 0.3714, 0.2760, 0.0705, 0.4402, 0.2798, 0.3326, 0.3998], device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0441, 0.0364, 0.0317, 0.0428, 0.0504, 0.0414, 0.0513], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:49:48,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8755, 3.9102, 4.0694, 3.8601, 4.0132, 4.3726, 3.9904, 3.6207], device='cuda:0'), covar=tensor([0.2037, 0.2331, 0.1937, 0.2361, 0.2537, 0.1383, 0.1528, 0.2810], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0568, 0.0630, 0.0470, 0.0626, 0.0655, 0.0490, 0.0625], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:50:06,301 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8060, 1.4080, 1.6723, 1.6963, 1.8590, 1.8732, 1.6071, 1.8555], device='cuda:0'), covar=tensor([0.0287, 0.0446, 0.0254, 0.0336, 0.0325, 0.0245, 0.0475, 0.0152], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0188, 0.0174, 0.0177, 0.0193, 0.0150, 0.0192, 0.0148], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:50:08,899 INFO [train.py:904] (0/8) Epoch 23, batch 9900, loss[loss=0.1937, simple_loss=0.2965, pruned_loss=0.04545, over 16382.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2614, pruned_loss=0.03495, over 3035526.62 frames. ], batch size: 146, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:50:14,349 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1727, 3.9401, 4.1693, 4.3793, 4.4982, 4.1607, 4.5763, 4.5646], device='cuda:0'), covar=tensor([0.1875, 0.1601, 0.1993, 0.0990, 0.0741, 0.1265, 0.0847, 0.0817], device='cuda:0'), in_proj_covar=tensor([0.0614, 0.0757, 0.0871, 0.0768, 0.0583, 0.0607, 0.0635, 0.0739], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:50:46,349 INFO [optim.py:368] (0/8) 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 17:50:52,786 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233222.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:52:06,043 INFO [train.py:904] (0/8) Epoch 23, batch 9950, loss[loss=0.161, simple_loss=0.2575, pruned_loss=0.03221, over 16535.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2629, pruned_loss=0.03495, over 3035600.80 frames. ], batch size: 68, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:52:20,250 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:54:07,461 INFO [train.py:904] (0/8) Epoch 23, batch 10000, loss[loss=0.1675, simple_loss=0.2644, pruned_loss=0.03532, over 16559.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2616, pruned_loss=0.03455, over 3063355.11 frames. ], batch size: 62, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:54:28,930 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.61 vs. limit=2.0 2023-05-01 17:54:40,347 INFO [optim.py:368] (0/8) 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 17:54:41,019 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233320.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:55:18,388 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-01 17:55:36,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233346.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 17:55:47,981 INFO [train.py:904] (0/8) Epoch 23, batch 10050, loss[loss=0.1468, simple_loss=0.2483, pruned_loss=0.02265, over 16869.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2614, pruned_loss=0.03418, over 3069786.30 frames. ], batch size: 96, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:56:30,643 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 17:56:50,385 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5272, 3.6239, 2.7850, 2.2131, 2.2490, 2.4607, 3.8622, 3.1008], device='cuda:0'), covar=tensor([0.3114, 0.0597, 0.1749, 0.2992, 0.2908, 0.2108, 0.0371, 0.1496], device='cuda:0'), in_proj_covar=tensor([0.0320, 0.0260, 0.0297, 0.0307, 0.0284, 0.0255, 0.0287, 0.0327], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:57:20,831 INFO [train.py:904] (0/8) Epoch 23, batch 10100, loss[loss=0.1474, simple_loss=0.2436, pruned_loss=0.02564, over 16892.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2619, pruned_loss=0.03434, over 3091750.36 frames. ], batch size: 96, lr: 2.90e-03, grad_scale: 8.0 2023-05-01 17:57:41,366 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4371, 4.4119, 4.2243, 3.7114, 4.3380, 1.6233, 4.0981, 4.0018], device='cuda:0'), covar=tensor([0.0117, 0.0121, 0.0220, 0.0286, 0.0116, 0.2815, 0.0144, 0.0256], device='cuda:0'), in_proj_covar=tensor([0.0164, 0.0156, 0.0194, 0.0170, 0.0172, 0.0204, 0.0184, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 17:57:53,984 INFO [optim.py:368] (0/8) 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 17:58:34,887 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5684, 3.6340, 2.7423, 2.2081, 2.2198, 2.3990, 3.9130, 3.1215], device='cuda:0'), covar=tensor([0.3144, 0.0633, 0.1833, 0.3033, 0.3026, 0.2179, 0.0402, 0.1534], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0261, 0.0299, 0.0308, 0.0284, 0.0256, 0.0288, 0.0328], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 17:58:40,643 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-23.pt 2023-05-01 17:59:06,343 INFO [train.py:904] (0/8) Epoch 24, batch 0, loss[loss=0.1627, simple_loss=0.2462, pruned_loss=0.03959, over 16023.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2462, pruned_loss=0.03959, over 16023.00 frames. ], batch size: 35, lr: 2.84e-03, grad_scale: 8.0 2023-05-01 17:59:06,344 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 17:59:14,241 INFO [train.py:938] (0/8) Epoch 24, validation: loss=0.145, simple_loss=0.2483, pruned_loss=0.02085, over 944034.00 frames. 2023-05-01 17:59:14,242 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 17:59:59,975 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0534, 2.0283, 2.6245, 2.9840, 2.8410, 3.4514, 1.9534, 3.5847], device='cuda:0'), covar=tensor([0.0257, 0.0682, 0.0354, 0.0361, 0.0379, 0.0225, 0.0767, 0.0186], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0190, 0.0176, 0.0179, 0.0195, 0.0151, 0.0194, 0.0149], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:00:06,457 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-01 18:00:23,702 INFO [train.py:904] (0/8) Epoch 24, batch 50, loss[loss=0.165, simple_loss=0.2529, pruned_loss=0.03858, over 17215.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2669, pruned_loss=0.04572, over 761038.69 frames. ], batch size: 44, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:00:44,693 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6718, 2.5618, 2.2804, 2.5035, 2.9464, 2.6719, 3.1846, 3.1140], device='cuda:0'), covar=tensor([0.0159, 0.0521, 0.0572, 0.0496, 0.0339, 0.0471, 0.0315, 0.0337], device='cuda:0'), in_proj_covar=tensor([0.0208, 0.0232, 0.0223, 0.0224, 0.0231, 0.0232, 0.0226, 0.0225], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:00:52,610 INFO [optim.py:368] (0/8) 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 18:01:32,993 INFO [train.py:904] (0/8) Epoch 24, batch 100, loss[loss=0.1594, simple_loss=0.2559, pruned_loss=0.03146, over 17072.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.261, pruned_loss=0.04266, over 1335635.10 frames. ], batch size: 55, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:01:53,736 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4242, 3.5042, 2.1324, 3.7242, 2.7741, 3.6065, 2.2373, 2.8419], device='cuda:0'), covar=tensor([0.0288, 0.0387, 0.1603, 0.0303, 0.0776, 0.0883, 0.1405, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0166, 0.0171, 0.0189, 0.0159, 0.0172, 0.0210, 0.0199, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 18:02:40,671 INFO [train.py:904] (0/8) Epoch 24, batch 150, loss[loss=0.158, simple_loss=0.2379, pruned_loss=0.039, over 17221.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2607, pruned_loss=0.04332, over 1783896.88 frames. ], batch size: 44, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:02:56,021 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:08,351 INFO [optim.py:368] (0/8) 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 18:03:30,592 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233639.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:03:32,088 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-01 18:03:39,289 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233646.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:03:40,823 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 18:03:48,640 INFO [train.py:904] (0/8) Epoch 24, batch 200, loss[loss=0.1808, simple_loss=0.2719, pruned_loss=0.04482, over 16666.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2624, pruned_loss=0.04445, over 2120715.48 frames. ], batch size: 62, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:03:56,147 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.97 vs. limit=2.0 2023-05-01 18:04:11,939 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.49 vs. limit=2.0 2023-05-01 18:04:15,546 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3890, 4.3898, 4.5722, 4.3594, 4.4507, 4.9875, 4.5036, 4.1893], device='cuda:0'), covar=tensor([0.1697, 0.2474, 0.2996, 0.2414, 0.3042, 0.1324, 0.1895, 0.2685], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0589, 0.0652, 0.0487, 0.0649, 0.0679, 0.0507, 0.0649], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 18:04:17,913 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:45,851 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233694.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:04:54,287 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233700.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:04:58,754 INFO [train.py:904] (0/8) Epoch 24, batch 250, loss[loss=0.1863, simple_loss=0.2677, pruned_loss=0.05251, over 16488.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2596, pruned_loss=0.04355, over 2379325.62 frames. ], batch size: 75, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:05:07,769 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=233710.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:05:25,968 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.709e+02 2.284e+02 2.752e+02 3.292e+02 5.130e+02, threshold=5.503e+02, percent-clipped=0.0 2023-05-01 18:06:08,093 INFO [train.py:904] (0/8) Epoch 24, batch 300, loss[loss=0.1421, simple_loss=0.2253, pruned_loss=0.02943, over 16246.00 frames. ], tot_loss[loss=0.1711, simple_loss=0.2572, pruned_loss=0.04249, over 2579401.75 frames. ], batch size: 165, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:06:33,489 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=233771.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:06:54,279 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2396, 4.9790, 5.2985, 5.4537, 5.6355, 5.0256, 5.5823, 5.6084], device='cuda:0'), covar=tensor([0.1899, 0.1266, 0.1623, 0.0781, 0.0557, 0.0800, 0.0564, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0785, 0.0905, 0.0795, 0.0604, 0.0627, 0.0662, 0.0765], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:07:16,412 INFO [train.py:904] (0/8) Epoch 24, batch 350, loss[loss=0.166, simple_loss=0.239, pruned_loss=0.04653, over 16347.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2546, pruned_loss=0.0418, over 2743292.69 frames. ], batch size: 146, lr: 2.84e-03, grad_scale: 1.0 2023-05-01 18:07:43,748 INFO [optim.py:368] (0/8) 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 18:08:25,455 INFO [train.py:904] (0/8) Epoch 24, batch 400, loss[loss=0.1503, simple_loss=0.2567, pruned_loss=0.02192, over 17047.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2529, pruned_loss=0.04135, over 2871048.82 frames. ], batch size: 50, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:08,206 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4625, 5.8777, 5.6400, 5.6743, 5.2711, 5.3855, 5.3090, 6.0127], device='cuda:0'), covar=tensor([0.1578, 0.1163, 0.1111, 0.0909, 0.1044, 0.0780, 0.1348, 0.1132], device='cuda:0'), in_proj_covar=tensor([0.0681, 0.0825, 0.0680, 0.0633, 0.0524, 0.0532, 0.0698, 0.0650], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:09:32,985 INFO [train.py:904] (0/8) Epoch 24, batch 450, loss[loss=0.1549, simple_loss=0.2528, pruned_loss=0.02849, over 17230.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2524, pruned_loss=0.0412, over 2972112.17 frames. ], batch size: 52, lr: 2.84e-03, grad_scale: 2.0 2023-05-01 18:09:47,237 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4679, 3.1151, 3.3696, 1.9832, 3.4504, 3.4369, 2.9239, 2.6587], device='cuda:0'), covar=tensor([0.0757, 0.0249, 0.0215, 0.1127, 0.0138, 0.0240, 0.0450, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0107, 0.0097, 0.0139, 0.0081, 0.0125, 0.0126, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 18:09:49,534 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=233915.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:09:59,640 INFO [optim.py:368] (0/8) 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 18:10:39,009 INFO [train.py:904] (0/8) Epoch 24, batch 500, loss[loss=0.1847, simple_loss=0.2562, pruned_loss=0.05661, over 16842.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2523, pruned_loss=0.04097, over 3047443.63 frames. ], batch size: 116, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:10:48,759 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-01 18:10:53,524 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=233963.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:11:20,677 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 18:11:22,625 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6684, 6.0744, 5.5602, 5.9270, 5.5292, 5.2807, 5.6004, 6.0825], device='cuda:0'), covar=tensor([0.2616, 0.1564, 0.2973, 0.1478, 0.1564, 0.1308, 0.2390, 0.2193], device='cuda:0'), in_proj_covar=tensor([0.0688, 0.0833, 0.0687, 0.0639, 0.0529, 0.0537, 0.0705, 0.0657], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:11:37,672 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=233995.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:11:44,290 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-234000.pt 2023-05-01 18:11:50,842 INFO [train.py:904] (0/8) Epoch 24, batch 550, loss[loss=0.1803, simple_loss=0.2747, pruned_loss=0.04298, over 16677.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2515, pruned_loss=0.04031, over 3107621.78 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:12:17,100 INFO [optim.py:368] (0/8) 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 18:12:20,381 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234025.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:12:35,931 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1260, 3.1722, 3.3557, 2.2945, 3.0193, 2.4356, 3.5592, 3.5025], device='cuda:0'), covar=tensor([0.0242, 0.0941, 0.0572, 0.1867, 0.0830, 0.0980, 0.0494, 0.0949], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0167, 0.0154, 0.0146, 0.0129, 0.0143, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 18:12:57,932 INFO [train.py:904] (0/8) Epoch 24, batch 600, loss[loss=0.1446, simple_loss=0.2318, pruned_loss=0.02875, over 16768.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2509, pruned_loss=0.04018, over 3159927.17 frames. ], batch size: 39, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:13:17,440 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234066.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:13:44,754 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234086.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:14:05,471 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 18:14:08,039 INFO [train.py:904] (0/8) Epoch 24, batch 650, loss[loss=0.1596, simple_loss=0.2326, pruned_loss=0.04332, over 16844.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2491, pruned_loss=0.0395, over 3193306.74 frames. ], batch size: 90, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:14:10,995 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-01 18:14:36,192 INFO [optim.py:368] (0/8) 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 18:15:16,422 INFO [train.py:904] (0/8) Epoch 24, batch 700, loss[loss=0.1759, simple_loss=0.273, pruned_loss=0.03936, over 17052.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.249, pruned_loss=0.03919, over 3232224.23 frames. ], batch size: 50, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:15:24,278 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8477, 5.2039, 4.9374, 4.9476, 4.6254, 4.6681, 4.5920, 5.3010], device='cuda:0'), covar=tensor([0.1414, 0.0930, 0.1133, 0.0893, 0.0995, 0.1092, 0.1284, 0.0965], device='cuda:0'), in_proj_covar=tensor([0.0685, 0.0833, 0.0686, 0.0638, 0.0529, 0.0536, 0.0705, 0.0655], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:16:23,776 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234202.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:16:24,511 INFO [train.py:904] (0/8) Epoch 24, batch 750, loss[loss=0.1418, simple_loss=0.2237, pruned_loss=0.02998, over 15911.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.249, pruned_loss=0.03921, over 3250981.61 frames. ], batch size: 35, lr: 2.83e-03, grad_scale: 2.0 2023-05-01 18:16:44,348 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 18:16:45,096 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9462, 4.3378, 4.3354, 3.1261, 3.5858, 4.2969, 3.8696, 2.4790], device='cuda:0'), covar=tensor([0.0437, 0.0083, 0.0060, 0.0363, 0.0158, 0.0116, 0.0104, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0086, 0.0087, 0.0135, 0.0099, 0.0111, 0.0096, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 18:16:52,435 INFO [optim.py:368] (0/8) 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 18:17:04,935 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234231.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:17:33,989 INFO [train.py:904] (0/8) Epoch 24, batch 800, loss[loss=0.1625, simple_loss=0.2554, pruned_loss=0.03476, over 17152.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2496, pruned_loss=0.03958, over 3270145.34 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:17:48,277 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234263.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 18:17:59,646 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-01 18:18:28,486 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234292.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:18:32,364 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234295.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:18:43,656 INFO [train.py:904] (0/8) Epoch 24, batch 850, loss[loss=0.1521, simple_loss=0.2282, pruned_loss=0.03803, over 16458.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2494, pruned_loss=0.03937, over 3279504.35 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:19:11,884 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.464e+02 2.010e+02 2.479e+02 2.918e+02 4.237e+02, threshold=4.958e+02, percent-clipped=0.0 2023-05-01 18:19:40,590 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234343.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:19:48,870 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4332, 5.4236, 5.2888, 4.7666, 4.9068, 5.3037, 5.2671, 4.9495], device='cuda:0'), covar=tensor([0.0580, 0.0530, 0.0311, 0.0329, 0.1099, 0.0527, 0.0280, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0444, 0.0351, 0.0350, 0.0353, 0.0407, 0.0238, 0.0420], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:19:52,511 INFO [train.py:904] (0/8) Epoch 24, batch 900, loss[loss=0.1789, simple_loss=0.2649, pruned_loss=0.0464, over 16638.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2489, pruned_loss=0.03854, over 3293213.02 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:20:11,491 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234366.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:20:33,662 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:02,828 INFO [train.py:904] (0/8) Epoch 24, batch 950, loss[loss=0.1894, simple_loss=0.2734, pruned_loss=0.05269, over 16634.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2495, pruned_loss=0.039, over 3297348.07 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:21:17,245 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:21:30,311 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.144e+02 2.493e+02 3.109e+02 1.071e+03, threshold=4.986e+02, percent-clipped=5.0 2023-05-01 18:21:37,404 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3305, 2.5331, 2.8881, 3.2263, 3.0938, 3.6946, 2.7527, 3.6835], device='cuda:0'), covar=tensor([0.0218, 0.0467, 0.0307, 0.0307, 0.0305, 0.0176, 0.0418, 0.0160], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0195, 0.0181, 0.0186, 0.0200, 0.0158, 0.0198, 0.0155], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:21:38,410 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4025, 4.4283, 4.7472, 4.7368, 4.7758, 4.4570, 4.4890, 4.3476], device='cuda:0'), covar=tensor([0.0441, 0.0807, 0.0472, 0.0499, 0.0573, 0.0541, 0.0924, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0427, 0.0473, 0.0462, 0.0424, 0.0503, 0.0483, 0.0565, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 18:22:10,614 INFO [train.py:904] (0/8) Epoch 24, batch 1000, loss[loss=0.1482, simple_loss=0.2375, pruned_loss=0.02945, over 16561.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2479, pruned_loss=0.03877, over 3295927.28 frames. ], batch size: 68, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:22:50,096 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8224, 4.0213, 2.6867, 4.6579, 3.1994, 4.5432, 2.8880, 3.2775], device='cuda:0'), covar=tensor([0.0369, 0.0408, 0.1587, 0.0319, 0.0883, 0.0567, 0.1468, 0.0785], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0179, 0.0196, 0.0169, 0.0178, 0.0220, 0.0205, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 18:23:22,529 INFO [train.py:904] (0/8) Epoch 24, batch 1050, loss[loss=0.1593, simple_loss=0.2397, pruned_loss=0.03945, over 16817.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2475, pruned_loss=0.038, over 3296920.85 frames. ], batch size: 96, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:23:28,493 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.98 vs. limit=2.0 2023-05-01 18:23:50,908 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.145e+02 2.609e+02 3.101e+02 1.491e+03, threshold=5.219e+02, percent-clipped=5.0 2023-05-01 18:24:30,882 INFO [train.py:904] (0/8) Epoch 24, batch 1100, loss[loss=0.1667, simple_loss=0.2443, pruned_loss=0.0445, over 16500.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2469, pruned_loss=0.03765, over 3308204.31 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:24:37,572 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234558.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:25:17,482 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=234587.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:25:38,674 INFO [train.py:904] (0/8) Epoch 24, batch 1150, loss[loss=0.1539, simple_loss=0.2422, pruned_loss=0.03279, over 16405.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2468, pruned_loss=0.03728, over 3312991.07 frames. ], batch size: 75, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:25:53,400 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4710, 3.8263, 4.1438, 2.3203, 3.3678, 2.4728, 3.9770, 4.0059], device='cuda:0'), covar=tensor([0.0282, 0.0830, 0.0481, 0.2081, 0.0786, 0.1095, 0.0541, 0.0959], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0165, 0.0168, 0.0155, 0.0146, 0.0130, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 18:26:06,112 INFO [optim.py:368] (0/8) 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 18:26:33,553 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 18:26:46,787 INFO [train.py:904] (0/8) Epoch 24, batch 1200, loss[loss=0.1492, simple_loss=0.2509, pruned_loss=0.02375, over 17089.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2461, pruned_loss=0.03671, over 3315419.23 frames. ], batch size: 50, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:27:09,100 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234669.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:25,040 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:27:53,734 INFO [train.py:904] (0/8) Epoch 24, batch 1250, loss[loss=0.1574, simple_loss=0.2576, pruned_loss=0.0286, over 17268.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2459, pruned_loss=0.03696, over 3318710.98 frames. ], batch size: 52, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:28:21,127 INFO [optim.py:368] (0/8) 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 18:28:25,312 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234726.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:30,424 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234729.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:28:31,728 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234730.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:28:35,898 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0372, 3.0932, 3.4274, 2.0986, 3.0001, 2.3503, 3.4972, 3.4163], device='cuda:0'), covar=tensor([0.0250, 0.0935, 0.0551, 0.2055, 0.0853, 0.1008, 0.0581, 0.0973], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 18:28:37,209 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-01 18:28:39,547 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.39 vs. limit=5.0 2023-05-01 18:29:02,404 INFO [train.py:904] (0/8) Epoch 24, batch 1300, loss[loss=0.1647, simple_loss=0.2483, pruned_loss=0.04055, over 16419.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2458, pruned_loss=0.03729, over 3308124.49 frames. ], batch size: 68, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:29:27,404 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8301, 5.0571, 5.2086, 5.0169, 5.0217, 5.6293, 5.1497, 4.8154], device='cuda:0'), covar=tensor([0.1484, 0.2113, 0.2772, 0.2327, 0.2821, 0.1138, 0.1739, 0.2743], device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0611, 0.0675, 0.0508, 0.0673, 0.0705, 0.0526, 0.0674], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 18:29:41,666 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:29:47,329 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234787.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:30:10,009 INFO [train.py:904] (0/8) Epoch 24, batch 1350, loss[loss=0.1652, simple_loss=0.2443, pruned_loss=0.04309, over 15650.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2462, pruned_loss=0.03763, over 3312679.12 frames. ], batch size: 191, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:30:38,935 INFO [optim.py:368] (0/8) 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 18:30:41,857 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9300, 3.6150, 3.9192, 2.2751, 3.9858, 3.9981, 3.2465, 3.0071], device='cuda:0'), covar=tensor([0.0664, 0.0231, 0.0181, 0.1124, 0.0096, 0.0203, 0.0360, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0109, 0.0100, 0.0141, 0.0083, 0.0129, 0.0129, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 18:30:45,857 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-01 18:31:07,084 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234843.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:19,627 INFO [train.py:904] (0/8) Epoch 24, batch 1400, loss[loss=0.1873, simple_loss=0.2599, pruned_loss=0.05738, over 16860.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2458, pruned_loss=0.03758, over 3303861.73 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:31:27,974 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234858.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:31:59,511 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.80 vs. limit=5.0 2023-05-01 18:32:00,158 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:07,333 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=234887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:29,526 INFO [train.py:904] (0/8) Epoch 24, batch 1450, loss[loss=0.1678, simple_loss=0.2446, pruned_loss=0.04548, over 16790.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2454, pruned_loss=0.03814, over 3315869.04 frames. ], batch size: 102, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:32:34,694 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234906.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:32:37,535 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.24 vs. limit=5.0 2023-05-01 18:32:50,989 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-01 18:32:58,964 INFO [optim.py:368] (0/8) 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 18:33:14,530 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=234935.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:26,740 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=234943.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:38,969 INFO [train.py:904] (0/8) Epoch 24, batch 1500, loss[loss=0.1625, simple_loss=0.2548, pruned_loss=0.03511, over 17101.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2455, pruned_loss=0.03821, over 3311337.36 frames. ], batch size: 47, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:33:43,580 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4737, 2.3998, 2.4120, 4.1546, 2.2922, 2.7571, 2.4634, 2.5522], device='cuda:0'), covar=tensor([0.1363, 0.3557, 0.3158, 0.0593, 0.4198, 0.2658, 0.3479, 0.3682], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0459, 0.0378, 0.0333, 0.0442, 0.0526, 0.0431, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:33:45,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=234958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:33:49,595 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 24, batch 1550, loss[loss=0.1943, simple_loss=0.2835, pruned_loss=0.05249, over 16681.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.248, pruned_loss=0.03926, over 3311412.62 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:35:03,989 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4921, 5.8753, 5.6668, 5.6488, 5.3245, 5.2930, 5.2688, 6.0331], device='cuda:0'), covar=tensor([0.1603, 0.1037, 0.1103, 0.0967, 0.0856, 0.0790, 0.1319, 0.0893], device='cuda:0'), in_proj_covar=tensor([0.0708, 0.0863, 0.0707, 0.0661, 0.0547, 0.0552, 0.0726, 0.0674], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:35:12,908 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235019.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:15,752 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235021.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:35:18,860 INFO [optim.py:368] (0/8) 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 18:35:20,358 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235025.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:35:58,108 INFO [train.py:904] (0/8) Epoch 24, batch 1600, loss[loss=0.145, simple_loss=0.2315, pruned_loss=0.02926, over 16836.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2497, pruned_loss=0.03987, over 3312537.11 frames. ], batch size: 42, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:36:25,436 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-05-01 18:36:37,998 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235082.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:37:06,939 INFO [train.py:904] (0/8) Epoch 24, batch 1650, loss[loss=0.1745, simple_loss=0.2682, pruned_loss=0.04039, over 16641.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2514, pruned_loss=0.04036, over 3323743.80 frames. ], batch size: 57, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:37:35,282 INFO [optim.py:368] (0/8) 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 18:37:36,134 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-01 18:37:54,830 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235138.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:38:16,061 INFO [train.py:904] (0/8) Epoch 24, batch 1700, loss[loss=0.1945, simple_loss=0.2717, pruned_loss=0.05863, over 16879.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2533, pruned_loss=0.04097, over 3324309.80 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:24,684 INFO [train.py:904] (0/8) Epoch 24, batch 1750, loss[loss=0.1732, simple_loss=0.2647, pruned_loss=0.04081, over 17036.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2545, pruned_loss=0.0408, over 3324041.10 frames. ], batch size: 50, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:39:52,421 INFO [optim.py:368] (0/8) 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 18:40:12,955 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235238.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:40:19,743 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.15 vs. limit=5.0 2023-05-01 18:40:33,185 INFO [train.py:904] (0/8) Epoch 24, batch 1800, loss[loss=0.1693, simple_loss=0.2635, pruned_loss=0.0375, over 17157.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.255, pruned_loss=0.04043, over 3331149.07 frames. ], batch size: 46, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:40:36,451 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235255.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:40:56,878 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 18:41:29,746 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 18:41:33,485 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9309, 2.1324, 2.2813, 3.4425, 2.1577, 2.3862, 2.2283, 2.2770], device='cuda:0'), covar=tensor([0.1533, 0.3627, 0.2946, 0.0746, 0.3899, 0.2641, 0.3927, 0.3332], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0462, 0.0379, 0.0334, 0.0444, 0.0528, 0.0433, 0.0539], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:41:40,815 INFO [train.py:904] (0/8) Epoch 24, batch 1850, loss[loss=0.1834, simple_loss=0.2636, pruned_loss=0.05157, over 16435.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2557, pruned_loss=0.04079, over 3322265.32 frames. ], batch size: 146, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:41:55,868 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235314.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,477 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:41:59,624 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235316.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:11,310 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.696e+02 2.161e+02 2.587e+02 3.187e+02 4.962e+02, threshold=5.175e+02, percent-clipped=0.0 2023-05-01 18:42:11,683 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:42:49,550 INFO [train.py:904] (0/8) Epoch 24, batch 1900, loss[loss=0.1755, simple_loss=0.2572, pruned_loss=0.04692, over 16902.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2548, pruned_loss=0.04057, over 3326252.95 frames. ], batch size: 109, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:43:01,078 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235360.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:43:14,794 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0543, 4.8118, 5.0883, 5.2415, 5.4572, 4.7837, 5.4491, 5.4507], device='cuda:0'), covar=tensor([0.1910, 0.1452, 0.1789, 0.0804, 0.0573, 0.0959, 0.0620, 0.0667], device='cuda:0'), in_proj_covar=tensor([0.0677, 0.0840, 0.0963, 0.0849, 0.0644, 0.0667, 0.0700, 0.0814], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:43:17,589 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235373.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:43:31,601 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235382.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:00,963 INFO [train.py:904] (0/8) Epoch 24, batch 1950, loss[loss=0.1688, simple_loss=0.2524, pruned_loss=0.04255, over 16796.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2547, pruned_loss=0.03993, over 3327075.04 frames. ], batch size: 83, lr: 2.83e-03, grad_scale: 4.0 2023-05-01 18:44:26,887 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235421.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:44:31,115 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.100e+02 2.476e+02 2.993e+02 5.203e+02, threshold=4.952e+02, percent-clipped=1.0 2023-05-01 18:44:37,918 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:44:43,438 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-01 18:44:48,943 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235438.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:45:08,817 INFO [train.py:904] (0/8) Epoch 24, batch 2000, loss[loss=0.1385, simple_loss=0.2287, pruned_loss=0.02413, over 16345.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.254, pruned_loss=0.03949, over 3328253.86 frames. ], batch size: 36, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:45:53,864 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235486.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:46:09,595 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4143, 3.3641, 3.4504, 3.5083, 3.5690, 3.2839, 3.4753, 3.6461], device='cuda:0'), covar=tensor([0.1241, 0.0992, 0.1021, 0.0636, 0.0642, 0.2474, 0.1435, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0682, 0.0845, 0.0971, 0.0853, 0.0648, 0.0670, 0.0704, 0.0818], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:46:12,021 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-01 18:46:16,412 INFO [train.py:904] (0/8) Epoch 24, batch 2050, loss[loss=0.1824, simple_loss=0.2667, pruned_loss=0.04903, over 16823.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2536, pruned_loss=0.03952, over 3326165.24 frames. ], batch size: 83, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:46:46,287 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.111e+02 2.349e+02 3.134e+02 8.501e+02, threshold=4.699e+02, percent-clipped=2.0 2023-05-01 18:47:03,595 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235538.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:47:24,567 INFO [train.py:904] (0/8) Epoch 24, batch 2100, loss[loss=0.1772, simple_loss=0.2728, pruned_loss=0.04076, over 16622.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2543, pruned_loss=0.04024, over 3326125.98 frames. ], batch size: 62, lr: 2.83e-03, grad_scale: 8.0 2023-05-01 18:48:10,074 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 24, batch 2150, loss[loss=0.1658, simple_loss=0.2618, pruned_loss=0.03491, over 17118.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2557, pruned_loss=0.04072, over 3318921.37 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:48:43,107 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235611.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:46,961 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235614.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:48:49,859 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 18:49:24,511 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.54 vs. limit=5.0 2023-05-01 18:49:41,328 INFO [train.py:904] (0/8) Epoch 24, batch 2200, loss[loss=0.1705, simple_loss=0.2608, pruned_loss=0.04008, over 16671.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2562, pruned_loss=0.04112, over 3316021.36 frames. ], batch size: 62, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:49:53,770 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235662.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:49:57,654 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235664.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:18,206 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235679.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:50:50,809 INFO [train.py:904] (0/8) Epoch 24, batch 2250, loss[loss=0.1813, simple_loss=0.2731, pruned_loss=0.04476, over 17052.00 frames. ], tot_loss[loss=0.1703, simple_loss=0.2573, pruned_loss=0.04162, over 3317461.19 frames. ], batch size: 55, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:51:08,558 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=235716.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 18:51:20,159 INFO [optim.py:368] (0/8) 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 18:51:29,116 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0377, 2.0354, 2.3087, 3.7307, 2.0794, 2.2790, 2.1310, 2.2025], device='cuda:0'), covar=tensor([0.1845, 0.4220, 0.3180, 0.0794, 0.4807, 0.3118, 0.4309, 0.3862], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0461, 0.0378, 0.0334, 0.0441, 0.0527, 0.0432, 0.0538], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:51:41,211 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235740.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:51:54,911 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6817, 4.1005, 4.0277, 2.8552, 3.5454, 4.0401, 3.6837, 2.2270], device='cuda:0'), covar=tensor([0.0546, 0.0114, 0.0069, 0.0421, 0.0148, 0.0127, 0.0116, 0.0558], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0087, 0.0087, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 18:51:57,981 INFO [train.py:904] (0/8) Epoch 24, batch 2300, loss[loss=0.1624, simple_loss=0.2413, pruned_loss=0.04178, over 16874.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2572, pruned_loss=0.04136, over 3327435.81 frames. ], batch size: 96, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:52:10,238 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-05-01 18:52:50,302 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=235791.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:53:08,150 INFO [train.py:904] (0/8) Epoch 24, batch 2350, loss[loss=0.1495, simple_loss=0.2365, pruned_loss=0.0312, over 17016.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2573, pruned_loss=0.04137, over 3324245.25 frames. ], batch size: 41, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:53:17,442 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6446, 6.0120, 5.7378, 5.7866, 5.3919, 5.3976, 5.3962, 6.1318], device='cuda:0'), covar=tensor([0.1420, 0.0884, 0.1113, 0.0856, 0.0958, 0.0675, 0.1219, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0710, 0.0867, 0.0710, 0.0664, 0.0550, 0.0553, 0.0728, 0.0677], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:53:37,794 INFO [optim.py:368] (0/8) 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 18:54:14,050 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=235852.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:54:15,317 INFO [train.py:904] (0/8) Epoch 24, batch 2400, loss[loss=0.1838, simple_loss=0.2725, pruned_loss=0.04752, over 15338.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2581, pruned_loss=0.04151, over 3320674.01 frames. ], batch size: 190, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:54:40,585 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0117, 2.1035, 2.6421, 2.9977, 2.8355, 3.5074, 2.4055, 3.5180], device='cuda:0'), covar=tensor([0.0288, 0.0554, 0.0377, 0.0361, 0.0383, 0.0217, 0.0551, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0197, 0.0184, 0.0189, 0.0204, 0.0162, 0.0201, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:55:22,145 INFO [train.py:904] (0/8) Epoch 24, batch 2450, loss[loss=0.1682, simple_loss=0.2575, pruned_loss=0.03946, over 16788.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2581, pruned_loss=0.04113, over 3323614.01 frames. ], batch size: 83, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:55:33,308 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=235911.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:55:43,696 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0249, 4.0372, 3.9746, 3.4094, 3.9878, 1.7485, 3.7836, 3.4065], device='cuda:0'), covar=tensor([0.0143, 0.0132, 0.0200, 0.0253, 0.0104, 0.3073, 0.0148, 0.0296], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0166, 0.0208, 0.0183, 0.0183, 0.0214, 0.0197, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 18:55:51,880 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 2.168e+02 2.510e+02 2.858e+02 5.891e+02, threshold=5.019e+02, percent-clipped=1.0 2023-05-01 18:56:28,818 INFO [train.py:904] (0/8) Epoch 24, batch 2500, loss[loss=0.1604, simple_loss=0.2542, pruned_loss=0.03328, over 15808.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2581, pruned_loss=0.04101, over 3327268.65 frames. ], batch size: 35, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:56:37,232 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=235959.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:57:34,145 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-236000.pt 2023-05-01 18:57:41,347 INFO [train.py:904] (0/8) Epoch 24, batch 2550, loss[loss=0.1618, simple_loss=0.2479, pruned_loss=0.03791, over 16520.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2581, pruned_loss=0.0408, over 3312848.24 frames. ], batch size: 75, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:57:58,519 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236016.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 18:58:01,821 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236018.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:10,534 INFO [optim.py:368] (0/8) 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 18:58:24,329 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236035.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:58:49,135 INFO [train.py:904] (0/8) Epoch 24, batch 2600, loss[loss=0.1564, simple_loss=0.2433, pruned_loss=0.03474, over 16891.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2578, pruned_loss=0.04047, over 3317925.58 frames. ], batch size: 96, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 18:58:53,966 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.78 vs. limit=5.0 2023-05-01 18:59:03,927 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236064.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 18:59:25,511 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236079.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 18:59:58,181 INFO [train.py:904] (0/8) Epoch 24, batch 2650, loss[loss=0.1422, simple_loss=0.2317, pruned_loss=0.02633, over 17188.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2579, pruned_loss=0.04015, over 3326408.32 frames. ], batch size: 46, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:00:27,608 INFO [optim.py:368] (0/8) 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 19:00:55,006 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.37 vs. limit=5.0 2023-05-01 19:00:58,523 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:01:05,919 INFO [train.py:904] (0/8) Epoch 24, batch 2700, loss[loss=0.1913, simple_loss=0.2716, pruned_loss=0.05547, over 16746.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2582, pruned_loss=0.03998, over 3333395.03 frames. ], batch size: 124, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:01:14,246 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9944, 2.5865, 2.1152, 2.3678, 2.9218, 2.6942, 2.9335, 3.0420], device='cuda:0'), covar=tensor([0.0247, 0.0430, 0.0552, 0.0484, 0.0264, 0.0357, 0.0231, 0.0297], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0246, 0.0235, 0.0237, 0.0247, 0.0246, 0.0247, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:02:15,470 INFO [train.py:904] (0/8) Epoch 24, batch 2750, loss[loss=0.172, simple_loss=0.2569, pruned_loss=0.04354, over 16687.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2577, pruned_loss=0.03892, over 3340918.74 frames. ], batch size: 134, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:02:16,813 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9678, 4.2252, 4.0826, 4.1280, 3.7702, 3.8204, 3.8758, 4.2346], device='cuda:0'), covar=tensor([0.1224, 0.0995, 0.1029, 0.0847, 0.0890, 0.1759, 0.0971, 0.1097], device='cuda:0'), in_proj_covar=tensor([0.0719, 0.0876, 0.0719, 0.0672, 0.0555, 0.0559, 0.0735, 0.0684], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:02:44,671 INFO [optim.py:368] (0/8) 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 19:03:22,970 INFO [train.py:904] (0/8) Epoch 24, batch 2800, loss[loss=0.1541, simple_loss=0.2544, pruned_loss=0.02688, over 17113.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2574, pruned_loss=0.0394, over 3341943.33 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:04:32,917 INFO [train.py:904] (0/8) Epoch 24, batch 2850, loss[loss=0.1809, simple_loss=0.2747, pruned_loss=0.04351, over 17121.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2568, pruned_loss=0.03921, over 3337280.80 frames. ], batch size: 48, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:05:03,079 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.95 vs. limit=5.0 2023-05-01 19:05:04,035 INFO [optim.py:368] (0/8) 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 19:05:17,509 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236335.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:05:43,142 INFO [train.py:904] (0/8) Epoch 24, batch 2900, loss[loss=0.1377, simple_loss=0.2245, pruned_loss=0.02543, over 17236.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2553, pruned_loss=0.03979, over 3332239.17 frames. ], batch size: 44, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:06:12,652 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=236374.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:25,485 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236383.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:06:45,407 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0113, 4.0557, 3.9138, 3.6607, 3.7017, 4.0383, 3.6569, 3.8237], device='cuda:0'), covar=tensor([0.0735, 0.0774, 0.0347, 0.0312, 0.0709, 0.0526, 0.1204, 0.0629], device='cuda:0'), in_proj_covar=tensor([0.0314, 0.0466, 0.0366, 0.0366, 0.0371, 0.0424, 0.0250, 0.0442], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 19:06:53,578 INFO [train.py:904] (0/8) Epoch 24, batch 2950, loss[loss=0.1585, simple_loss=0.2583, pruned_loss=0.02935, over 17257.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2551, pruned_loss=0.04069, over 3317042.86 frames. ], batch size: 52, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:07:24,083 INFO [optim.py:368] (0/8) 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 19:07:30,651 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8120, 4.0001, 4.0189, 2.9423, 3.5773, 4.1142, 3.6579, 2.4950], device='cuda:0'), covar=tensor([0.0487, 0.0264, 0.0071, 0.0384, 0.0150, 0.0120, 0.0125, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0088, 0.0088, 0.0136, 0.0101, 0.0113, 0.0097, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 19:07:54,785 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236447.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:08:02,715 INFO [train.py:904] (0/8) Epoch 24, batch 3000, loss[loss=0.1665, simple_loss=0.2513, pruned_loss=0.04088, over 16721.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2556, pruned_loss=0.04098, over 3321584.00 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:08:02,716 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 19:08:12,056 INFO [train.py:938] (0/8) Epoch 24, validation: loss=0.1342, simple_loss=0.2393, pruned_loss=0.0145, over 944034.00 frames. 2023-05-01 19:08:12,057 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 19:09:12,139 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236495.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:09:22,991 INFO [train.py:904] (0/8) Epoch 24, batch 3050, loss[loss=0.1613, simple_loss=0.2427, pruned_loss=0.04002, over 16791.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2555, pruned_loss=0.04083, over 3329299.62 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:09:53,424 INFO [optim.py:368] (0/8) 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 19:10:32,487 INFO [train.py:904] (0/8) Epoch 24, batch 3100, loss[loss=0.1569, simple_loss=0.2386, pruned_loss=0.03761, over 16404.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.255, pruned_loss=0.0406, over 3333700.37 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:10:38,259 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0478, 2.1826, 2.3287, 3.4723, 2.2247, 2.4142, 2.3325, 2.2920], device='cuda:0'), covar=tensor([0.1395, 0.3461, 0.2935, 0.0709, 0.3922, 0.2468, 0.3348, 0.3544], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0461, 0.0379, 0.0334, 0.0441, 0.0529, 0.0432, 0.0539], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:11:02,841 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3279, 3.4211, 2.1684, 3.6087, 2.7584, 3.5685, 2.2889, 2.8068], device='cuda:0'), covar=tensor([0.0303, 0.0430, 0.1423, 0.0333, 0.0723, 0.0725, 0.1295, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0182, 0.0198, 0.0174, 0.0179, 0.0224, 0.0206, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:11:17,066 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6311, 3.6556, 4.2776, 2.3578, 3.4534, 2.6449, 4.0216, 3.9940], device='cuda:0'), covar=tensor([0.0241, 0.1048, 0.0411, 0.2044, 0.0760, 0.0968, 0.0584, 0.1105], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0168, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-01 19:11:43,896 INFO [train.py:904] (0/8) Epoch 24, batch 3150, loss[loss=0.1828, simple_loss=0.2578, pruned_loss=0.05388, over 16516.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2536, pruned_loss=0.04027, over 3332773.13 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:12:13,746 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.596e+02 2.103e+02 2.556e+02 2.857e+02 5.533e+02, threshold=5.112e+02, percent-clipped=2.0 2023-05-01 19:12:52,327 INFO [train.py:904] (0/8) Epoch 24, batch 3200, loss[loss=0.1382, simple_loss=0.2212, pruned_loss=0.02757, over 16679.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2527, pruned_loss=0.03973, over 3331067.29 frames. ], batch size: 89, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:13:21,991 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=236674.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:13:34,886 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7194, 4.5308, 4.7657, 4.9239, 5.0778, 4.5305, 5.0462, 5.0926], device='cuda:0'), covar=tensor([0.1986, 0.1355, 0.1582, 0.0751, 0.0644, 0.1065, 0.0967, 0.0874], device='cuda:0'), in_proj_covar=tensor([0.0685, 0.0850, 0.0976, 0.0851, 0.0652, 0.0678, 0.0707, 0.0822], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:13:50,025 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 19:14:01,578 INFO [train.py:904] (0/8) Epoch 24, batch 3250, loss[loss=0.1582, simple_loss=0.2539, pruned_loss=0.03123, over 17107.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2528, pruned_loss=0.04012, over 3332318.51 frames. ], batch size: 47, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:14:26,981 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=236722.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:14:31,668 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 24, batch 3300, loss[loss=0.1368, simple_loss=0.2246, pruned_loss=0.0245, over 17017.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2537, pruned_loss=0.04021, over 3332181.90 frames. ], batch size: 41, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:21,168 INFO [train.py:904] (0/8) Epoch 24, batch 3350, loss[loss=0.1719, simple_loss=0.2545, pruned_loss=0.04464, over 16749.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.254, pruned_loss=0.03992, over 3329155.73 frames. ], batch size: 134, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:16:51,759 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.239e+02 2.070e+02 2.526e+02 2.944e+02 6.359e+02, threshold=5.052e+02, percent-clipped=3.0 2023-05-01 19:16:53,527 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-01 19:17:04,618 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6469, 4.6358, 4.6064, 4.0017, 4.6280, 1.8154, 4.3855, 4.3121], device='cuda:0'), covar=tensor([0.0139, 0.0113, 0.0199, 0.0433, 0.0121, 0.2927, 0.0163, 0.0223], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0168, 0.0210, 0.0186, 0.0186, 0.0216, 0.0199, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:17:15,791 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-01 19:17:33,221 INFO [train.py:904] (0/8) Epoch 24, batch 3400, loss[loss=0.1623, simple_loss=0.2399, pruned_loss=0.04239, over 16686.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2545, pruned_loss=0.0402, over 3332907.25 frames. ], batch size: 134, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:17:39,729 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3558, 3.4965, 3.6156, 3.5869, 3.6062, 3.4560, 3.4842, 3.5146], device='cuda:0'), covar=tensor([0.0412, 0.0611, 0.0441, 0.0436, 0.0562, 0.0501, 0.0719, 0.0488], device='cuda:0'), in_proj_covar=tensor([0.0429, 0.0478, 0.0465, 0.0428, 0.0510, 0.0485, 0.0570, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 19:17:44,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3744, 4.6675, 4.4917, 4.5054, 4.2325, 4.1607, 4.2489, 4.7289], device='cuda:0'), covar=tensor([0.1243, 0.0867, 0.1082, 0.0839, 0.0841, 0.1626, 0.1070, 0.0892], device='cuda:0'), in_proj_covar=tensor([0.0715, 0.0871, 0.0717, 0.0670, 0.0555, 0.0557, 0.0731, 0.0680], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:18:31,118 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=236892.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:18:34,684 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-01 19:18:45,611 INFO [train.py:904] (0/8) Epoch 24, batch 3450, loss[loss=0.1835, simple_loss=0.2728, pruned_loss=0.04712, over 16739.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2532, pruned_loss=0.03951, over 3323970.47 frames. ], batch size: 57, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:15,852 INFO [optim.py:368] (0/8) 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 19:19:45,786 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 19:19:55,271 INFO [train.py:904] (0/8) Epoch 24, batch 3500, loss[loss=0.1583, simple_loss=0.2428, pruned_loss=0.03687, over 16810.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2526, pruned_loss=0.03969, over 3310748.75 frames. ], batch size: 39, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:19:56,838 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=236953.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 19:20:40,535 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1664, 2.1713, 2.3435, 3.8469, 2.1602, 2.5574, 2.2873, 2.4013], device='cuda:0'), covar=tensor([0.1549, 0.3919, 0.3025, 0.0653, 0.3966, 0.2745, 0.3921, 0.3147], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0461, 0.0380, 0.0335, 0.0441, 0.0531, 0.0432, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:20:41,681 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0488, 2.2110, 2.6025, 3.0139, 2.8191, 3.4658, 2.5330, 3.5035], device='cuda:0'), covar=tensor([0.0289, 0.0510, 0.0387, 0.0337, 0.0386, 0.0217, 0.0451, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0196, 0.0185, 0.0189, 0.0204, 0.0163, 0.0200, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:21:06,721 INFO [train.py:904] (0/8) Epoch 24, batch 3550, loss[loss=0.1694, simple_loss=0.2474, pruned_loss=0.04573, over 16234.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2523, pruned_loss=0.03958, over 3303829.55 frames. ], batch size: 165, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:21:35,733 INFO [optim.py:368] (0/8) 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 19:22:09,451 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8029, 2.5463, 2.3849, 3.8313, 3.1282, 3.8908, 1.5493, 2.8449], device='cuda:0'), covar=tensor([0.1442, 0.0760, 0.1255, 0.0209, 0.0165, 0.0381, 0.1713, 0.0844], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0177, 0.0197, 0.0197, 0.0206, 0.0218, 0.0205, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:22:15,052 INFO [train.py:904] (0/8) Epoch 24, batch 3600, loss[loss=0.1595, simple_loss=0.2419, pruned_loss=0.03855, over 16510.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2509, pruned_loss=0.03963, over 3303559.14 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:22:42,131 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 19:23:26,125 INFO [train.py:904] (0/8) Epoch 24, batch 3650, loss[loss=0.1834, simple_loss=0.2509, pruned_loss=0.05795, over 16837.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2496, pruned_loss=0.04017, over 3284433.12 frames. ], batch size: 116, lr: 2.82e-03, grad_scale: 8.0 2023-05-01 19:23:58,729 INFO [optim.py:368] (0/8) 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 19:24:39,874 INFO [train.py:904] (0/8) Epoch 24, batch 3700, loss[loss=0.1606, simple_loss=0.2368, pruned_loss=0.04218, over 16404.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2487, pruned_loss=0.04189, over 3276100.95 frames. ], batch size: 68, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:24:52,797 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8094, 2.6352, 2.5777, 4.1514, 3.5149, 4.0937, 1.5979, 2.9798], device='cuda:0'), covar=tensor([0.1460, 0.0720, 0.1175, 0.0169, 0.0158, 0.0376, 0.1658, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0197, 0.0206, 0.0218, 0.0205, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:25:53,338 INFO [train.py:904] (0/8) Epoch 24, batch 3750, loss[loss=0.2043, simple_loss=0.2892, pruned_loss=0.0597, over 12079.00 frames. ], tot_loss[loss=0.1678, simple_loss=0.2495, pruned_loss=0.04304, over 3252357.27 frames. ], batch size: 246, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:26:25,690 INFO [optim.py:368] (0/8) 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 19:26:57,495 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237248.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 19:27:03,611 INFO [train.py:904] (0/8) Epoch 24, batch 3800, loss[loss=0.1929, simple_loss=0.2874, pruned_loss=0.0492, over 17016.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2509, pruned_loss=0.04382, over 3252101.30 frames. ], batch size: 41, lr: 2.82e-03, grad_scale: 4.0 2023-05-01 19:27:32,004 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-01 19:27:40,939 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8753, 3.8419, 3.9515, 3.7508, 3.8642, 4.3235, 3.9660, 3.6348], device='cuda:0'), covar=tensor([0.2222, 0.2366, 0.2341, 0.2372, 0.2896, 0.1846, 0.1575, 0.2526], device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0621, 0.0682, 0.0515, 0.0681, 0.0712, 0.0537, 0.0677], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 19:28:20,070 INFO [train.py:904] (0/8) Epoch 24, batch 3850, loss[loss=0.1627, simple_loss=0.2399, pruned_loss=0.04277, over 16664.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2512, pruned_loss=0.04435, over 3248217.41 frames. ], batch size: 134, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:28:52,885 INFO [optim.py:368] (0/8) 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 19:28:53,465 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0164, 3.4633, 4.1857, 1.9974, 4.3779, 4.3828, 3.2349, 3.2800], device='cuda:0'), covar=tensor([0.0667, 0.0338, 0.0185, 0.1254, 0.0056, 0.0100, 0.0416, 0.0437], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0140, 0.0083, 0.0129, 0.0129, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:29:00,971 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9463, 2.0543, 2.4509, 2.8902, 2.8613, 2.9170, 2.2358, 3.1319], device='cuda:0'), covar=tensor([0.0193, 0.0493, 0.0389, 0.0278, 0.0317, 0.0296, 0.0507, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0196, 0.0185, 0.0189, 0.0204, 0.0163, 0.0201, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:29:30,848 INFO [train.py:904] (0/8) Epoch 24, batch 3900, loss[loss=0.1735, simple_loss=0.2543, pruned_loss=0.0463, over 16611.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2513, pruned_loss=0.0449, over 3250433.09 frames. ], batch size: 57, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:29:57,334 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9938, 5.0044, 4.8478, 4.1024, 4.9716, 2.0166, 4.7108, 4.4535], device='cuda:0'), covar=tensor([0.0104, 0.0081, 0.0207, 0.0443, 0.0090, 0.2770, 0.0130, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0168, 0.0210, 0.0186, 0.0186, 0.0215, 0.0198, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:30:43,387 INFO [train.py:904] (0/8) Epoch 24, batch 3950, loss[loss=0.192, simple_loss=0.2702, pruned_loss=0.05689, over 16311.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2509, pruned_loss=0.04549, over 3262569.78 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 4.0 2023-05-01 19:31:17,989 INFO [optim.py:368] (0/8) 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 19:31:57,081 INFO [train.py:904] (0/8) Epoch 24, batch 4000, loss[loss=0.1773, simple_loss=0.2576, pruned_loss=0.04848, over 17119.00 frames. ], tot_loss[loss=0.1707, simple_loss=0.2505, pruned_loss=0.04542, over 3262685.30 frames. ], batch size: 48, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:32:53,251 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237491.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:33:06,353 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1677, 3.1958, 1.9366, 3.3791, 2.3790, 3.5352, 2.0936, 2.6039], device='cuda:0'), covar=tensor([0.0320, 0.0454, 0.1810, 0.0190, 0.0944, 0.0402, 0.1808, 0.0853], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0174, 0.0179, 0.0223, 0.0205, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:33:10,924 INFO [train.py:904] (0/8) Epoch 24, batch 4050, loss[loss=0.1687, simple_loss=0.2631, pruned_loss=0.03717, over 15458.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2512, pruned_loss=0.0444, over 3263201.45 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:33:32,973 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7177, 2.8374, 2.3574, 2.7050, 3.1283, 2.8651, 3.2041, 3.3889], device='cuda:0'), covar=tensor([0.0091, 0.0426, 0.0526, 0.0398, 0.0257, 0.0356, 0.0254, 0.0197], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0244, 0.0234, 0.0235, 0.0245, 0.0245, 0.0246, 0.0243], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:33:43,754 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.357e+02 1.828e+02 2.122e+02 2.508e+02 6.067e+02, threshold=4.243e+02, percent-clipped=1.0 2023-05-01 19:34:17,624 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=237548.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:34:23,314 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 24, batch 4100, loss[loss=0.1702, simple_loss=0.2579, pruned_loss=0.04126, over 16638.00 frames. ], tot_loss[loss=0.1704, simple_loss=0.253, pruned_loss=0.04384, over 3267200.88 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:34:38,947 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237563.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:35:09,051 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0152, 4.1156, 4.3369, 4.2864, 4.3079, 4.0967, 4.0821, 4.0728], device='cuda:0'), covar=tensor([0.0333, 0.0467, 0.0332, 0.0392, 0.0475, 0.0391, 0.0792, 0.0482], device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0476, 0.0460, 0.0423, 0.0506, 0.0481, 0.0565, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 19:35:26,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5804, 1.9173, 2.2901, 2.6452, 2.6477, 2.9509, 1.9777, 2.9087], device='cuda:0'), covar=tensor([0.0267, 0.0510, 0.0328, 0.0349, 0.0310, 0.0227, 0.0553, 0.0143], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0195, 0.0184, 0.0189, 0.0204, 0.0163, 0.0201, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:35:30,063 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=237596.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:35:36,970 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8195, 5.0895, 4.8897, 4.9316, 4.6427, 4.5880, 4.6300, 5.2222], device='cuda:0'), covar=tensor([0.1387, 0.0920, 0.1033, 0.0847, 0.0807, 0.1020, 0.1034, 0.0874], device='cuda:0'), in_proj_covar=tensor([0.0707, 0.0858, 0.0709, 0.0664, 0.0547, 0.0551, 0.0724, 0.0671], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:35:40,173 INFO [train.py:904] (0/8) Epoch 24, batch 4150, loss[loss=0.2049, simple_loss=0.2976, pruned_loss=0.05611, over 16425.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2603, pruned_loss=0.04652, over 3233203.53 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:36:14,062 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237624.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:36:16,370 INFO [optim.py:368] (0/8) 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 19:36:22,067 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 19:36:56,406 INFO [train.py:904] (0/8) Epoch 24, batch 4200, loss[loss=0.2156, simple_loss=0.2947, pruned_loss=0.06829, over 11309.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2673, pruned_loss=0.04836, over 3197530.45 frames. ], batch size: 248, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:37:14,231 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 19:37:45,904 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7432, 4.7655, 5.1294, 5.0654, 5.1324, 4.7832, 4.6849, 4.6682], device='cuda:0'), covar=tensor([0.0294, 0.0460, 0.0376, 0.0456, 0.0354, 0.0347, 0.1204, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0474, 0.0458, 0.0421, 0.0504, 0.0479, 0.0562, 0.0385], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 19:38:10,757 INFO [train.py:904] (0/8) Epoch 24, batch 4250, loss[loss=0.1578, simple_loss=0.2505, pruned_loss=0.03254, over 16398.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2707, pruned_loss=0.04835, over 3176520.84 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:38:45,325 INFO [optim.py:368] (0/8) 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 19:39:26,159 INFO [train.py:904] (0/8) Epoch 24, batch 4300, loss[loss=0.1984, simple_loss=0.2862, pruned_loss=0.05531, over 16991.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.272, pruned_loss=0.04766, over 3164949.52 frames. ], batch size: 50, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:40:00,899 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8049, 2.6639, 2.4552, 4.3574, 3.3479, 3.9429, 1.6233, 2.9115], device='cuda:0'), covar=tensor([0.1252, 0.0799, 0.1262, 0.0181, 0.0291, 0.0385, 0.1602, 0.0873], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0194, 0.0205, 0.0215, 0.0203, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:40:23,436 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237790.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:40:41,743 INFO [train.py:904] (0/8) Epoch 24, batch 4350, loss[loss=0.195, simple_loss=0.2896, pruned_loss=0.05016, over 16657.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2753, pruned_loss=0.04838, over 3173665.84 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:41:02,497 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2285, 2.8040, 3.0911, 1.7244, 3.2052, 3.2641, 2.6117, 2.5071], device='cuda:0'), covar=tensor([0.0817, 0.0341, 0.0227, 0.1225, 0.0103, 0.0164, 0.0516, 0.0511], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0111, 0.0101, 0.0141, 0.0083, 0.0129, 0.0130, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:41:14,417 INFO [optim.py:368] (0/8) 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 19:41:45,920 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237847.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:49,068 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3635, 5.6335, 5.3978, 5.4143, 5.1478, 4.9998, 4.9892, 5.7539], device='cuda:0'), covar=tensor([0.1085, 0.0766, 0.0920, 0.0799, 0.0680, 0.0789, 0.1026, 0.0777], device='cuda:0'), in_proj_covar=tensor([0.0696, 0.0842, 0.0697, 0.0654, 0.0537, 0.0542, 0.0711, 0.0661], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:41:52,156 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237851.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:41:54,785 INFO [train.py:904] (0/8) Epoch 24, batch 4400, loss[loss=0.2001, simple_loss=0.3029, pruned_loss=0.04863, over 16841.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2775, pruned_loss=0.04955, over 3187095.32 frames. ], batch size: 83, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:05,646 INFO [train.py:904] (0/8) Epoch 24, batch 4450, loss[loss=0.1988, simple_loss=0.2945, pruned_loss=0.05152, over 16880.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2809, pruned_loss=0.05074, over 3207202.76 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:43:18,905 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=237912.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:43:28,560 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=237919.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 19:43:38,900 INFO [optim.py:368] (0/8) 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 19:44:16,779 INFO [train.py:904] (0/8) Epoch 24, batch 4500, loss[loss=0.184, simple_loss=0.2754, pruned_loss=0.04628, over 16495.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2809, pruned_loss=0.05131, over 3212391.51 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:44:31,691 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3993, 3.1003, 3.6433, 1.8051, 3.7811, 3.8017, 2.8819, 2.8702], device='cuda:0'), covar=tensor([0.0847, 0.0368, 0.0221, 0.1246, 0.0084, 0.0137, 0.0502, 0.0481], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0110, 0.0100, 0.0141, 0.0083, 0.0128, 0.0130, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:44:46,758 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=237973.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:44:52,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4277, 2.7920, 3.0307, 1.9580, 2.7119, 1.9312, 3.0347, 3.0217], device='cuda:0'), covar=tensor([0.0258, 0.0939, 0.0593, 0.2145, 0.0898, 0.1144, 0.0623, 0.0908], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0167, 0.0168, 0.0154, 0.0146, 0.0131, 0.0143, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:45:25,782 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-238000.pt 2023-05-01 19:45:32,218 INFO [train.py:904] (0/8) Epoch 24, batch 4550, loss[loss=0.2286, simple_loss=0.3139, pruned_loss=0.07168, over 16467.00 frames. ], tot_loss[loss=0.1934, simple_loss=0.2821, pruned_loss=0.05238, over 3221068.14 frames. ], batch size: 68, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:46:02,670 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1055, 3.0882, 2.5487, 2.8852, 3.4429, 3.0056, 3.6914, 3.6412], device='cuda:0'), covar=tensor([0.0060, 0.0354, 0.0466, 0.0350, 0.0236, 0.0321, 0.0166, 0.0198], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0239, 0.0229, 0.0231, 0.0240, 0.0240, 0.0241, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:46:03,169 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-01 19:46:04,581 INFO [optim.py:368] (0/8) 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 19:46:41,878 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9351, 5.1818, 4.9883, 5.0177, 4.7538, 4.6606, 4.6212, 5.2841], device='cuda:0'), covar=tensor([0.1315, 0.0861, 0.0963, 0.0818, 0.0796, 0.0980, 0.1115, 0.0786], device='cuda:0'), in_proj_covar=tensor([0.0692, 0.0837, 0.0694, 0.0650, 0.0535, 0.0540, 0.0707, 0.0659], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:46:44,390 INFO [train.py:904] (0/8) Epoch 24, batch 4600, loss[loss=0.198, simple_loss=0.2924, pruned_loss=0.05178, over 17243.00 frames. ], tot_loss[loss=0.1941, simple_loss=0.2827, pruned_loss=0.05272, over 3224358.44 frames. ], batch size: 45, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:47:00,270 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238064.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:15,154 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238074.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:47:56,654 INFO [train.py:904] (0/8) Epoch 24, batch 4650, loss[loss=0.1971, simple_loss=0.2789, pruned_loss=0.05762, over 16824.00 frames. ], tot_loss[loss=0.1937, simple_loss=0.2817, pruned_loss=0.05279, over 3217868.49 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:48:29,649 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 19:48:30,805 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238126.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:43,702 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238135.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:48:44,058 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 19:48:59,278 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238146.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:00,461 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238147.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:49:09,299 INFO [train.py:904] (0/8) Epoch 24, batch 4700, loss[loss=0.1511, simple_loss=0.2455, pruned_loss=0.02839, over 16853.00 frames. ], tot_loss[loss=0.1916, simple_loss=0.2794, pruned_loss=0.05191, over 3217164.50 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:49:58,177 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:09,159 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238195.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:50:10,677 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4653, 3.5350, 2.7505, 2.2361, 2.3438, 2.4082, 3.7690, 3.0738], device='cuda:0'), covar=tensor([0.3154, 0.0735, 0.2001, 0.2686, 0.2557, 0.2135, 0.0529, 0.1445], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0272, 0.0307, 0.0319, 0.0301, 0.0265, 0.0298, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 19:50:20,663 INFO [train.py:904] (0/8) Epoch 24, batch 4750, loss[loss=0.1734, simple_loss=0.2612, pruned_loss=0.0428, over 16916.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2753, pruned_loss=0.0501, over 3212735.13 frames. ], batch size: 109, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:50:33,316 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8104, 2.7724, 2.8221, 4.9376, 3.7647, 4.1771, 1.7359, 2.9823], device='cuda:0'), covar=tensor([0.1322, 0.0810, 0.1125, 0.0135, 0.0292, 0.0383, 0.1553, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0206, 0.0216, 0.0204, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:50:41,132 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.60 vs. limit=5.0 2023-05-01 19:50:43,591 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238219.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:50:53,744 INFO [optim.py:368] (0/8) 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 19:51:31,477 INFO [train.py:904] (0/8) Epoch 24, batch 4800, loss[loss=0.1815, simple_loss=0.2699, pruned_loss=0.04659, over 16907.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2718, pruned_loss=0.04789, over 3216929.97 frames. ], batch size: 116, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:51:49,242 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1450, 3.1306, 1.9482, 3.4418, 2.3633, 3.4469, 2.1199, 2.5174], device='cuda:0'), covar=tensor([0.0324, 0.0436, 0.1753, 0.0186, 0.0970, 0.0632, 0.1620, 0.0898], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0169, 0.0177, 0.0220, 0.0203, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 19:51:52,765 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238267.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:51:54,086 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238268.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:52:25,246 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0427, 2.1927, 2.1805, 3.7396, 2.1429, 2.5210, 2.2863, 2.3231], device='cuda:0'), covar=tensor([0.1564, 0.3845, 0.3089, 0.0593, 0.4137, 0.2581, 0.3773, 0.3315], device='cuda:0'), in_proj_covar=tensor([0.0409, 0.0460, 0.0375, 0.0331, 0.0440, 0.0528, 0.0429, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:52:46,493 INFO [train.py:904] (0/8) Epoch 24, batch 4850, loss[loss=0.1726, simple_loss=0.2707, pruned_loss=0.03726, over 15322.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2729, pruned_loss=0.04717, over 3196283.25 frames. ], batch size: 191, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:52:53,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2100, 3.4724, 3.5816, 3.5114, 3.5397, 3.4406, 3.2120, 3.4904], device='cuda:0'), covar=tensor([0.0570, 0.0719, 0.0580, 0.0705, 0.0660, 0.0604, 0.1241, 0.0573], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0460, 0.0449, 0.0413, 0.0495, 0.0468, 0.0552, 0.0376], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 19:53:22,733 INFO [optim.py:368] (0/8) 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 19:54:04,046 INFO [train.py:904] (0/8) Epoch 24, batch 4900, loss[loss=0.1677, simple_loss=0.2581, pruned_loss=0.0387, over 12272.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2725, pruned_loss=0.04623, over 3177633.25 frames. ], batch size: 247, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:17,240 INFO [train.py:904] (0/8) Epoch 24, batch 4950, loss[loss=0.2077, simple_loss=0.2956, pruned_loss=0.0599, over 15458.00 frames. ], tot_loss[loss=0.1815, simple_loss=0.272, pruned_loss=0.0455, over 3188864.46 frames. ], batch size: 190, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:55:41,598 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238420.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:41,728 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238420.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 19:55:49,725 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.002e+02 2.407e+02 2.827e+02 4.284e+02, threshold=4.815e+02, percent-clipped=0.0 2023-05-01 19:55:56,830 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238430.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:55:59,977 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5401, 4.4987, 4.3893, 3.3654, 4.4911, 1.5167, 4.1770, 4.0269], device='cuda:0'), covar=tensor([0.0162, 0.0128, 0.0242, 0.0662, 0.0166, 0.3492, 0.0193, 0.0359], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0163, 0.0204, 0.0181, 0.0180, 0.0209, 0.0192, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:56:20,218 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238446.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:56:29,588 INFO [train.py:904] (0/8) Epoch 24, batch 5000, loss[loss=0.1946, simple_loss=0.2929, pruned_loss=0.04815, over 16423.00 frames. ], tot_loss[loss=0.1824, simple_loss=0.2736, pruned_loss=0.04558, over 3199164.21 frames. ], batch size: 146, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:56:34,668 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2699, 4.1210, 4.3257, 4.4532, 4.5993, 4.2421, 4.5659, 4.6370], device='cuda:0'), covar=tensor([0.1608, 0.1235, 0.1487, 0.0703, 0.0515, 0.1123, 0.0700, 0.0627], device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0797, 0.0918, 0.0806, 0.0615, 0.0636, 0.0665, 0.0775], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 19:57:11,141 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238481.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 19:57:11,983 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238482.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:30,550 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238494.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 19:57:42,738 INFO [train.py:904] (0/8) Epoch 24, batch 5050, loss[loss=0.1691, simple_loss=0.2655, pruned_loss=0.03629, over 16409.00 frames. ], tot_loss[loss=0.183, simple_loss=0.2746, pruned_loss=0.04573, over 3216350.44 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:58:15,455 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 24, batch 5100, loss[loss=0.1713, simple_loss=0.2631, pruned_loss=0.03976, over 16758.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2724, pruned_loss=0.04498, over 3234071.56 frames. ], batch size: 76, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 19:59:16,170 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 24, batch 5150, loss[loss=0.1716, simple_loss=0.2736, pruned_loss=0.03478, over 16811.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2725, pruned_loss=0.04381, over 3229471.59 frames. ], batch size: 102, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:00:26,119 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:00:38,785 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.286e+02 1.999e+02 2.272e+02 2.615e+02 4.924e+02, threshold=4.544e+02, percent-clipped=1.0 2023-05-01 20:00:52,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1921, 3.5370, 3.6211, 2.1632, 3.0764, 2.4761, 3.6493, 3.7060], device='cuda:0'), covar=tensor([0.0256, 0.0771, 0.0603, 0.1942, 0.0844, 0.0940, 0.0552, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0145, 0.0130, 0.0143, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:01:17,686 INFO [train.py:904] (0/8) Epoch 24, batch 5200, loss[loss=0.1651, simple_loss=0.2537, pruned_loss=0.03822, over 16539.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2701, pruned_loss=0.0431, over 3223875.76 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:01:57,605 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238681.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:02:28,278 INFO [train.py:904] (0/8) Epoch 24, batch 5250, loss[loss=0.1667, simple_loss=0.2652, pruned_loss=0.03409, over 16878.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2679, pruned_loss=0.04278, over 3215886.19 frames. ], batch size: 96, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:02:51,468 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 20:03:07,647 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238730.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:25,295 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238742.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:03:39,617 INFO [train.py:904] (0/8) Epoch 24, batch 5300, loss[loss=0.1563, simple_loss=0.2409, pruned_loss=0.03585, over 16693.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2646, pruned_loss=0.04194, over 3213262.15 frames. ], batch size: 124, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:04:00,409 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238768.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:12,311 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=238776.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:04:15,076 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238778.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:20,297 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=238782.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:04:50,580 INFO [train.py:904] (0/8) Epoch 24, batch 5350, loss[loss=0.1854, simple_loss=0.2872, pruned_loss=0.0418, over 16326.00 frames. ], tot_loss[loss=0.1737, simple_loss=0.2637, pruned_loss=0.04183, over 3202181.34 frames. ], batch size: 165, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:04:56,357 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2478, 4.1406, 4.3292, 4.4532, 4.5973, 4.2048, 4.5835, 4.6301], device='cuda:0'), covar=tensor([0.1741, 0.1202, 0.1514, 0.0745, 0.0524, 0.1178, 0.0674, 0.0650], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0797, 0.0918, 0.0808, 0.0616, 0.0637, 0.0664, 0.0776], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:05:17,645 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 20:05:21,928 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=238830.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:06:01,813 INFO [train.py:904] (0/8) Epoch 24, batch 5400, loss[loss=0.1873, simple_loss=0.278, pruned_loss=0.04825, over 12274.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2656, pruned_loss=0.04196, over 3214709.36 frames. ], batch size: 247, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:17,036 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6218, 2.5929, 1.8691, 2.7677, 2.1051, 2.8273, 2.1542, 2.3711], device='cuda:0'), covar=tensor([0.0328, 0.0383, 0.1389, 0.0214, 0.0772, 0.0475, 0.1287, 0.0660], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0166, 0.0176, 0.0216, 0.0201, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:07:18,023 INFO [train.py:904] (0/8) Epoch 24, batch 5450, loss[loss=0.2001, simple_loss=0.2858, pruned_loss=0.05718, over 12193.00 frames. ], tot_loss[loss=0.1787, simple_loss=0.2695, pruned_loss=0.04393, over 3196947.98 frames. ], batch size: 248, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:07:54,284 INFO [optim.py:368] (0/8) 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 20:08:11,058 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=238936.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:08:17,962 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9239, 3.2531, 3.4225, 2.0080, 3.0490, 2.2178, 3.3877, 3.5931], device='cuda:0'), covar=tensor([0.0245, 0.0792, 0.0570, 0.2132, 0.0774, 0.1016, 0.0632, 0.0952], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0166, 0.0168, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:08:33,259 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1341, 5.1378, 4.9534, 4.2995, 5.0847, 1.8190, 4.8519, 4.7076], device='cuda:0'), covar=tensor([0.0093, 0.0085, 0.0183, 0.0400, 0.0092, 0.2903, 0.0114, 0.0230], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0163, 0.0205, 0.0182, 0.0180, 0.0210, 0.0192, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:08:35,279 INFO [train.py:904] (0/8) Epoch 24, batch 5500, loss[loss=0.1965, simple_loss=0.2888, pruned_loss=0.05211, over 16395.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2759, pruned_loss=0.04829, over 3145767.77 frames. ], batch size: 75, lr: 2.81e-03, grad_scale: 8.0 2023-05-01 20:09:43,403 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=238997.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:09:53,117 INFO [train.py:904] (0/8) Epoch 24, batch 5550, loss[loss=0.1805, simple_loss=0.2716, pruned_loss=0.04467, over 17247.00 frames. ], tot_loss[loss=0.1942, simple_loss=0.2822, pruned_loss=0.05311, over 3115572.27 frames. ], batch size: 52, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:10:29,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7416, 4.5243, 4.4497, 4.8902, 5.0458, 4.6105, 5.0654, 5.0356], device='cuda:0'), covar=tensor([0.1544, 0.1300, 0.2314, 0.0894, 0.0808, 0.1174, 0.0866, 0.0942], device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0793, 0.0916, 0.0805, 0.0613, 0.0636, 0.0663, 0.0775], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:10:30,509 INFO [optim.py:368] (0/8) 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 20:10:48,382 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239037.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:11:12,516 INFO [train.py:904] (0/8) Epoch 24, batch 5600, loss[loss=0.2568, simple_loss=0.3187, pruned_loss=0.09742, over 10874.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2866, pruned_loss=0.05651, over 3091689.72 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:11:52,332 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239076.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:12:14,950 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0552, 2.3865, 2.4040, 2.8277, 1.9404, 3.2068, 1.9061, 2.7525], device='cuda:0'), covar=tensor([0.1096, 0.0617, 0.0978, 0.0193, 0.0117, 0.0371, 0.1363, 0.0645], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0178, 0.0196, 0.0195, 0.0206, 0.0216, 0.0205, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:12:36,914 INFO [train.py:904] (0/8) Epoch 24, batch 5650, loss[loss=0.2405, simple_loss=0.3227, pruned_loss=0.07917, over 16284.00 frames. ], tot_loss[loss=0.2066, simple_loss=0.2915, pruned_loss=0.06078, over 3064155.71 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:12:50,374 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2601, 1.7140, 2.0425, 2.2499, 2.4197, 2.5353, 1.8425, 2.4351], device='cuda:0'), covar=tensor([0.0257, 0.0521, 0.0318, 0.0356, 0.0297, 0.0227, 0.0523, 0.0149], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0193, 0.0181, 0.0185, 0.0200, 0.0159, 0.0198, 0.0156], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:13:11,112 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239124.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:13:14,768 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.394e+02 3.339e+02 4.323e+02 5.614e+02 1.229e+03, threshold=8.646e+02, percent-clipped=9.0 2023-05-01 20:13:55,698 INFO [train.py:904] (0/8) Epoch 24, batch 5700, loss[loss=0.2711, simple_loss=0.3252, pruned_loss=0.1084, over 11365.00 frames. ], tot_loss[loss=0.2079, simple_loss=0.2922, pruned_loss=0.06173, over 3071603.69 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:13:59,428 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6297, 3.8956, 4.1516, 2.4035, 3.5506, 2.8287, 4.1141, 4.1852], device='cuda:0'), covar=tensor([0.0242, 0.0745, 0.0485, 0.1912, 0.0713, 0.0885, 0.0525, 0.0766], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0165, 0.0167, 0.0153, 0.0145, 0.0130, 0.0143, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:15:14,252 INFO [train.py:904] (0/8) Epoch 24, batch 5750, loss[loss=0.2071, simple_loss=0.2979, pruned_loss=0.05815, over 16873.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2954, pruned_loss=0.06347, over 3053001.21 frames. ], batch size: 116, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:15:53,492 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 24, batch 5800, loss[loss=0.1944, simple_loss=0.2836, pruned_loss=0.05254, over 15226.00 frames. ], tot_loss[loss=0.2101, simple_loss=0.2955, pruned_loss=0.06238, over 3050042.27 frames. ], batch size: 191, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:17:20,270 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5627, 2.2428, 1.8635, 2.0345, 2.5627, 2.2434, 2.3137, 2.6925], device='cuda:0'), covar=tensor([0.0244, 0.0454, 0.0581, 0.0521, 0.0271, 0.0400, 0.0246, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0238, 0.0230, 0.0230, 0.0240, 0.0238, 0.0239, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:17:33,851 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3584, 3.4282, 2.0659, 3.7281, 2.6008, 3.8017, 2.1887, 2.7550], device='cuda:0'), covar=tensor([0.0347, 0.0412, 0.1750, 0.0304, 0.0853, 0.0461, 0.1667, 0.0848], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0180, 0.0196, 0.0169, 0.0179, 0.0219, 0.0205, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:17:40,222 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239292.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:17:56,790 INFO [train.py:904] (0/8) Epoch 24, batch 5850, loss[loss=0.1871, simple_loss=0.2661, pruned_loss=0.05404, over 17118.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2927, pruned_loss=0.06016, over 3079160.48 frames. ], batch size: 47, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:18:33,904 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.882e+02 2.891e+02 3.438e+02 4.490e+02 7.448e+02, threshold=6.875e+02, percent-clipped=1.0 2023-05-01 20:18:52,934 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239337.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:19:01,629 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1656, 3.6043, 3.6744, 2.3133, 3.3938, 3.6813, 3.3584, 2.2373], device='cuda:0'), covar=tensor([0.0605, 0.0066, 0.0060, 0.0471, 0.0110, 0.0117, 0.0116, 0.0459], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0087, 0.0086, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 20:19:19,097 INFO [train.py:904] (0/8) Epoch 24, batch 5900, loss[loss=0.2082, simple_loss=0.289, pruned_loss=0.06368, over 16906.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.292, pruned_loss=0.05984, over 3084618.79 frames. ], batch size: 116, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:20:02,545 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 20:20:14,710 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239385.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:20:42,793 INFO [train.py:904] (0/8) Epoch 24, batch 5950, loss[loss=0.2026, simple_loss=0.2879, pruned_loss=0.05866, over 16868.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2929, pruned_loss=0.05869, over 3092255.31 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:21:12,648 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-01 20:21:21,253 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.785e+02 2.477e+02 3.014e+02 3.724e+02 7.833e+02, threshold=6.029e+02, percent-clipped=2.0 2023-05-01 20:21:31,923 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9963, 2.3236, 2.3093, 2.8141, 1.8954, 3.1813, 1.8141, 2.7109], device='cuda:0'), covar=tensor([0.1143, 0.0645, 0.1050, 0.0196, 0.0126, 0.0361, 0.1409, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0176, 0.0195, 0.0193, 0.0204, 0.0215, 0.0203, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:22:03,339 INFO [train.py:904] (0/8) Epoch 24, batch 6000, loss[loss=0.2212, simple_loss=0.2918, pruned_loss=0.0753, over 11789.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2925, pruned_loss=0.05868, over 3072946.68 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:22:03,340 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 20:22:14,271 INFO [train.py:938] (0/8) Epoch 24, validation: loss=0.1493, simple_loss=0.2618, pruned_loss=0.01837, over 944034.00 frames. 2023-05-01 20:22:14,272 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 20:22:34,680 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4157, 3.2646, 2.7166, 2.1817, 2.2655, 2.3211, 3.3845, 3.0416], device='cuda:0'), covar=tensor([0.2928, 0.0651, 0.1747, 0.2670, 0.2588, 0.2148, 0.0523, 0.1320], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0272, 0.0308, 0.0320, 0.0301, 0.0266, 0.0299, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 20:23:19,146 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6971, 4.9764, 4.7500, 4.7628, 4.4853, 4.4647, 4.3467, 5.0414], device='cuda:0'), covar=tensor([0.1180, 0.0840, 0.0950, 0.0884, 0.0840, 0.1118, 0.1199, 0.0843], device='cuda:0'), in_proj_covar=tensor([0.0687, 0.0828, 0.0686, 0.0640, 0.0529, 0.0530, 0.0696, 0.0649], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:23:32,277 INFO [train.py:904] (0/8) Epoch 24, batch 6050, loss[loss=0.1873, simple_loss=0.2822, pruned_loss=0.04613, over 16239.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2911, pruned_loss=0.0578, over 3087436.74 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:24:09,880 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.713e+02 3.363e+02 4.065e+02 7.660e+02, threshold=6.725e+02, percent-clipped=3.0 2023-05-01 20:24:42,772 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.17 vs. limit=5.0 2023-05-01 20:24:51,444 INFO [train.py:904] (0/8) Epoch 24, batch 6100, loss[loss=0.1736, simple_loss=0.2676, pruned_loss=0.03982, over 16755.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2905, pruned_loss=0.05659, over 3108428.59 frames. ], batch size: 76, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:25:03,158 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8995, 2.1460, 2.2876, 3.4143, 2.0518, 2.3793, 2.2425, 2.2693], device='cuda:0'), covar=tensor([0.1503, 0.3557, 0.2930, 0.0649, 0.4258, 0.2555, 0.3530, 0.3190], device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0455, 0.0372, 0.0328, 0.0435, 0.0521, 0.0425, 0.0530], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:25:56,156 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=239592.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 20:26:14,100 INFO [train.py:904] (0/8) Epoch 24, batch 6150, loss[loss=0.1934, simple_loss=0.2815, pruned_loss=0.05259, over 16824.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2881, pruned_loss=0.05574, over 3131836.19 frames. ], batch size: 116, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:26:45,468 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=239622.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:26:53,242 INFO [optim.py:368] (0/8) 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 20:27:14,523 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=239640.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 20:27:34,966 INFO [train.py:904] (0/8) Epoch 24, batch 6200, loss[loss=0.1939, simple_loss=0.2849, pruned_loss=0.05144, over 16911.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.286, pruned_loss=0.05511, over 3126835.88 frames. ], batch size: 96, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:27:42,834 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7316, 1.8977, 2.3065, 2.6782, 2.6508, 3.0613, 1.9347, 3.0056], device='cuda:0'), covar=tensor([0.0235, 0.0544, 0.0369, 0.0348, 0.0339, 0.0209, 0.0591, 0.0145], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0193, 0.0182, 0.0184, 0.0200, 0.0160, 0.0199, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:28:21,934 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=239683.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:28:52,037 INFO [train.py:904] (0/8) Epoch 24, batch 6250, loss[loss=0.2042, simple_loss=0.302, pruned_loss=0.05322, over 16443.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2864, pruned_loss=0.05557, over 3113048.36 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:29:29,993 INFO [optim.py:368] (0/8) 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 20:29:30,520 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8791, 5.0815, 5.3761, 5.3028, 5.3658, 5.0931, 4.9076, 4.7982], device='cuda:0'), covar=tensor([0.0517, 0.0709, 0.0579, 0.0651, 0.0595, 0.0548, 0.1399, 0.0584], device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0471, 0.0459, 0.0421, 0.0505, 0.0478, 0.0562, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 20:30:05,927 INFO [train.py:904] (0/8) Epoch 24, batch 6300, loss[loss=0.1938, simple_loss=0.2822, pruned_loss=0.05271, over 16418.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2858, pruned_loss=0.05496, over 3112210.14 frames. ], batch size: 68, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:24,183 INFO [train.py:904] (0/8) Epoch 24, batch 6350, loss[loss=0.2434, simple_loss=0.3105, pruned_loss=0.08809, over 11240.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2868, pruned_loss=0.05656, over 3096369.00 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:31:24,717 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1384, 2.1159, 1.8003, 1.7683, 2.3226, 1.9800, 1.9614, 2.3990], device='cuda:0'), covar=tensor([0.0223, 0.0363, 0.0505, 0.0456, 0.0259, 0.0351, 0.0206, 0.0266], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0237, 0.0229, 0.0230, 0.0239, 0.0237, 0.0237, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:32:03,930 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.770e+02 2.790e+02 3.396e+02 4.270e+02 1.026e+03, threshold=6.791e+02, percent-clipped=1.0 2023-05-01 20:32:42,138 INFO [train.py:904] (0/8) Epoch 24, batch 6400, loss[loss=0.2069, simple_loss=0.2924, pruned_loss=0.0607, over 17012.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2868, pruned_loss=0.05691, over 3118549.37 frames. ], batch size: 53, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:33:03,632 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 20:33:19,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-01 20:33:55,865 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 20:33:58,192 INFO [train.py:904] (0/8) Epoch 24, batch 6450, loss[loss=0.1948, simple_loss=0.2739, pruned_loss=0.05789, over 11664.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2871, pruned_loss=0.05696, over 3092591.20 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:34:04,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8329, 2.2620, 1.8073, 2.0256, 2.6504, 2.2359, 2.4758, 2.8269], device='cuda:0'), covar=tensor([0.0280, 0.0446, 0.0629, 0.0526, 0.0326, 0.0453, 0.0342, 0.0307], device='cuda:0'), in_proj_covar=tensor([0.0217, 0.0237, 0.0228, 0.0229, 0.0239, 0.0236, 0.0236, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:34:15,357 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-01 20:34:37,421 INFO [optim.py:368] (0/8) 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 20:35:16,109 INFO [train.py:904] (0/8) Epoch 24, batch 6500, loss[loss=0.2122, simple_loss=0.2815, pruned_loss=0.07146, over 11548.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2853, pruned_loss=0.05658, over 3100061.43 frames. ], batch size: 250, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:35:55,238 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=239978.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:35:55,365 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7776, 3.5578, 4.2110, 1.9986, 4.3814, 4.3841, 3.1591, 3.2875], device='cuda:0'), covar=tensor([0.0773, 0.0300, 0.0187, 0.1248, 0.0062, 0.0169, 0.0415, 0.0440], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0109, 0.0100, 0.0138, 0.0082, 0.0128, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:36:24,785 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9439, 2.1254, 2.0838, 3.5885, 2.0077, 2.4037, 2.2171, 2.2477], device='cuda:0'), covar=tensor([0.1518, 0.3821, 0.3228, 0.0623, 0.4287, 0.2655, 0.3867, 0.3256], device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0455, 0.0372, 0.0328, 0.0436, 0.0522, 0.0426, 0.0532], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:36:31,257 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-240000.pt 2023-05-01 20:36:39,182 INFO [train.py:904] (0/8) Epoch 24, batch 6550, loss[loss=0.1921, simple_loss=0.2897, pruned_loss=0.04718, over 16426.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2882, pruned_loss=0.05735, over 3109991.98 frames. ], batch size: 146, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:36:42,854 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0768, 3.0655, 1.9146, 3.2515, 2.3208, 3.2790, 2.0896, 2.6058], device='cuda:0'), covar=tensor([0.0308, 0.0396, 0.1592, 0.0305, 0.0804, 0.0761, 0.1510, 0.0742], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0178, 0.0195, 0.0167, 0.0177, 0.0217, 0.0203, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:37:06,112 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6140, 3.7301, 2.8495, 2.2385, 2.5281, 2.3630, 3.9457, 3.3563], device='cuda:0'), covar=tensor([0.2873, 0.0642, 0.1726, 0.2891, 0.2573, 0.2213, 0.0483, 0.1337], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0272, 0.0307, 0.0320, 0.0301, 0.0265, 0.0299, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 20:37:16,981 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.521e+02 3.197e+02 3.720e+02 9.360e+02, threshold=6.395e+02, percent-clipped=2.0 2023-05-01 20:37:54,315 INFO [train.py:904] (0/8) Epoch 24, batch 6600, loss[loss=0.2084, simple_loss=0.2948, pruned_loss=0.06097, over 16270.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2904, pruned_loss=0.05752, over 3126825.41 frames. ], batch size: 165, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:11,706 INFO [train.py:904] (0/8) Epoch 24, batch 6650, loss[loss=0.1651, simple_loss=0.2554, pruned_loss=0.03742, over 17097.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2911, pruned_loss=0.05856, over 3124987.58 frames. ], batch size: 49, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:39:50,367 INFO [optim.py:368] (0/8) 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 20:40:28,846 INFO [train.py:904] (0/8) Epoch 24, batch 6700, loss[loss=0.2221, simple_loss=0.2957, pruned_loss=0.07428, over 11528.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.29, pruned_loss=0.05895, over 3116395.11 frames. ], batch size: 246, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:41:45,725 INFO [train.py:904] (0/8) Epoch 24, batch 6750, loss[loss=0.2361, simple_loss=0.3081, pruned_loss=0.08202, over 11733.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.289, pruned_loss=0.05902, over 3103366.53 frames. ], batch size: 247, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:42:23,538 INFO [optim.py:368] (0/8) 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 20:43:01,413 INFO [train.py:904] (0/8) Epoch 24, batch 6800, loss[loss=0.2368, simple_loss=0.331, pruned_loss=0.07133, over 16886.00 frames. ], tot_loss[loss=0.2042, simple_loss=0.2895, pruned_loss=0.05947, over 3084994.02 frames. ], batch size: 109, lr: 2.80e-03, grad_scale: 8.0 2023-05-01 20:43:33,049 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.86 vs. limit=5.0 2023-05-01 20:43:42,209 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240278.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:44:06,658 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9721, 4.9118, 4.7926, 3.8811, 4.8503, 1.5771, 4.5256, 4.3538], device='cuda:0'), covar=tensor([0.0163, 0.0168, 0.0220, 0.0571, 0.0154, 0.3288, 0.0286, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0181, 0.0212, 0.0194, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:44:07,893 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8702, 2.6587, 2.6233, 4.5155, 3.2308, 4.1174, 1.7692, 3.0723], device='cuda:0'), covar=tensor([0.1318, 0.0801, 0.1176, 0.0171, 0.0274, 0.0396, 0.1537, 0.0754], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0193, 0.0205, 0.0216, 0.0203, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:44:21,126 INFO [train.py:904] (0/8) Epoch 24, batch 6850, loss[loss=0.2258, simple_loss=0.2963, pruned_loss=0.07762, over 11410.00 frames. ], tot_loss[loss=0.204, simple_loss=0.2898, pruned_loss=0.05904, over 3085586.22 frames. ], batch size: 248, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:44:56,068 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=240326.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:45:00,027 INFO [optim.py:368] (0/8) 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 20:45:17,488 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240340.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 20:45:35,455 INFO [train.py:904] (0/8) Epoch 24, batch 6900, loss[loss=0.2022, simple_loss=0.3091, pruned_loss=0.0476, over 16661.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.292, pruned_loss=0.05832, over 3103419.78 frames. ], batch size: 76, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:46:25,906 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-01 20:46:33,699 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-01 20:46:50,846 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240401.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 20:46:52,761 INFO [train.py:904] (0/8) Epoch 24, batch 6950, loss[loss=0.1854, simple_loss=0.2754, pruned_loss=0.04768, over 16716.00 frames. ], tot_loss[loss=0.2065, simple_loss=0.2936, pruned_loss=0.05968, over 3089556.07 frames. ], batch size: 57, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:47:33,343 INFO [optim.py:368] (0/8) 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 20:48:07,690 INFO [train.py:904] (0/8) Epoch 24, batch 7000, loss[loss=0.1877, simple_loss=0.2879, pruned_loss=0.04373, over 16512.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2935, pruned_loss=0.05938, over 3084324.88 frames. ], batch size: 75, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:48:31,507 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3211, 3.4372, 3.5912, 3.5593, 3.5826, 3.4223, 3.4327, 3.4702], device='cuda:0'), covar=tensor([0.0443, 0.0709, 0.0507, 0.0486, 0.0579, 0.0593, 0.0905, 0.0555], device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0464, 0.0451, 0.0416, 0.0497, 0.0472, 0.0555, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 20:48:48,816 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5669, 3.3226, 3.8378, 1.8961, 3.9768, 4.0101, 2.9766, 2.9128], device='cuda:0'), covar=tensor([0.0782, 0.0300, 0.0202, 0.1246, 0.0079, 0.0163, 0.0460, 0.0476], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0138, 0.0082, 0.0128, 0.0129, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 20:49:13,700 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8237, 2.1169, 2.4445, 3.0750, 2.1924, 2.3016, 2.2920, 2.2310], device='cuda:0'), covar=tensor([0.1386, 0.3267, 0.2613, 0.0739, 0.4239, 0.2375, 0.3128, 0.3373], device='cuda:0'), in_proj_covar=tensor([0.0408, 0.0457, 0.0375, 0.0330, 0.0440, 0.0523, 0.0429, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:49:21,372 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3316, 4.2941, 4.2174, 3.4673, 4.2787, 1.7586, 4.0360, 3.8379], device='cuda:0'), covar=tensor([0.0128, 0.0111, 0.0208, 0.0387, 0.0104, 0.2956, 0.0153, 0.0281], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0165, 0.0207, 0.0183, 0.0181, 0.0213, 0.0194, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:49:23,909 INFO [train.py:904] (0/8) Epoch 24, batch 7050, loss[loss=0.2056, simple_loss=0.3044, pruned_loss=0.05341, over 16666.00 frames. ], tot_loss[loss=0.2054, simple_loss=0.2935, pruned_loss=0.05864, over 3111248.50 frames. ], batch size: 89, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:49:24,354 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0846, 4.1492, 4.4442, 4.3882, 4.4118, 4.1563, 4.1474, 4.1136], device='cuda:0'), covar=tensor([0.0366, 0.0599, 0.0372, 0.0436, 0.0526, 0.0423, 0.0909, 0.0547], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0465, 0.0451, 0.0416, 0.0498, 0.0473, 0.0555, 0.0377], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 20:50:06,659 INFO [optim.py:368] (0/8) 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 20:50:42,275 INFO [train.py:904] (0/8) Epoch 24, batch 7100, loss[loss=0.206, simple_loss=0.2945, pruned_loss=0.05871, over 16518.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2924, pruned_loss=0.05864, over 3101847.10 frames. ], batch size: 75, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:51:59,040 INFO [train.py:904] (0/8) Epoch 24, batch 7150, loss[loss=0.1712, simple_loss=0.2625, pruned_loss=0.03988, over 16827.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2898, pruned_loss=0.0575, over 3109666.52 frames. ], batch size: 96, lr: 2.80e-03, grad_scale: 2.0 2023-05-01 20:52:39,168 INFO [optim.py:368] (0/8) 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 20:52:49,486 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-01 20:53:12,653 INFO [train.py:904] (0/8) Epoch 24, batch 7200, loss[loss=0.182, simple_loss=0.2731, pruned_loss=0.04543, over 16659.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2875, pruned_loss=0.05563, over 3116289.95 frames. ], batch size: 134, lr: 2.80e-03, grad_scale: 4.0 2023-05-01 20:54:20,531 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=240696.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 20:54:32,017 INFO [train.py:904] (0/8) Epoch 24, batch 7250, loss[loss=0.2114, simple_loss=0.2879, pruned_loss=0.0674, over 11324.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2857, pruned_loss=0.05502, over 3088415.18 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:54:44,334 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-01 20:55:12,201 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.572e+02 2.642e+02 3.264e+02 4.366e+02 7.559e+02, threshold=6.528e+02, percent-clipped=1.0 2023-05-01 20:55:45,071 INFO [train.py:904] (0/8) Epoch 24, batch 7300, loss[loss=0.196, simple_loss=0.2924, pruned_loss=0.04975, over 16731.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.286, pruned_loss=0.05577, over 3071143.35 frames. ], batch size: 83, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:56:03,638 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6111, 3.8144, 2.8548, 2.2861, 2.5331, 2.4700, 4.2401, 3.3564], device='cuda:0'), covar=tensor([0.3226, 0.0711, 0.1982, 0.2839, 0.3062, 0.2261, 0.0456, 0.1460], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0272, 0.0308, 0.0321, 0.0301, 0.0266, 0.0300, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 20:56:58,478 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2579, 4.2315, 4.1184, 3.3279, 4.2175, 1.6635, 3.9933, 3.6621], device='cuda:0'), covar=tensor([0.0110, 0.0092, 0.0207, 0.0352, 0.0093, 0.3058, 0.0129, 0.0305], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0163, 0.0204, 0.0180, 0.0178, 0.0209, 0.0191, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:57:02,851 INFO [train.py:904] (0/8) Epoch 24, batch 7350, loss[loss=0.2622, simple_loss=0.3222, pruned_loss=0.1012, over 11185.00 frames. ], tot_loss[loss=0.1996, simple_loss=0.2867, pruned_loss=0.05626, over 3074469.03 frames. ], batch size: 250, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:57:44,660 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.735e+02 3.201e+02 4.098e+02 9.519e+02, threshold=6.402e+02, percent-clipped=6.0 2023-05-01 20:58:18,676 INFO [train.py:904] (0/8) Epoch 24, batch 7400, loss[loss=0.1782, simple_loss=0.2751, pruned_loss=0.04066, over 16812.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2875, pruned_loss=0.05677, over 3069735.76 frames. ], batch size: 102, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 20:58:29,794 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2029, 4.2964, 4.1246, 3.8318, 3.8237, 4.2192, 3.9061, 3.9883], device='cuda:0'), covar=tensor([0.0648, 0.0674, 0.0298, 0.0293, 0.0781, 0.0484, 0.0816, 0.0630], device='cuda:0'), in_proj_covar=tensor([0.0288, 0.0433, 0.0338, 0.0338, 0.0342, 0.0391, 0.0233, 0.0407], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:58:54,972 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6666, 1.7952, 1.5559, 1.5235, 1.9501, 1.6123, 1.6096, 1.9215], device='cuda:0'), covar=tensor([0.0280, 0.0328, 0.0458, 0.0418, 0.0250, 0.0274, 0.0234, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0211, 0.0232, 0.0225, 0.0226, 0.0234, 0.0232, 0.0232, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 20:59:34,873 INFO [train.py:904] (0/8) Epoch 24, batch 7450, loss[loss=0.2134, simple_loss=0.3046, pruned_loss=0.06112, over 16912.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2891, pruned_loss=0.05813, over 3066947.44 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:00:19,564 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.977e+02 2.843e+02 3.471e+02 4.367e+02 9.441e+02, threshold=6.943e+02, percent-clipped=5.0 2023-05-01 21:00:22,560 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240932.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:00:55,593 INFO [train.py:904] (0/8) Epoch 24, batch 7500, loss[loss=0.1899, simple_loss=0.2779, pruned_loss=0.05097, over 16757.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2889, pruned_loss=0.05741, over 3065028.95 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:01:34,747 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4593, 3.5555, 2.7297, 2.2132, 2.3919, 2.3551, 3.7574, 3.2186], device='cuda:0'), covar=tensor([0.3048, 0.0589, 0.1821, 0.2697, 0.2538, 0.2124, 0.0453, 0.1250], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0269, 0.0306, 0.0317, 0.0298, 0.0264, 0.0297, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 21:01:45,352 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=240986.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:01:57,404 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=240993.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:02:00,905 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=240996.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:02:11,654 INFO [train.py:904] (0/8) Epoch 24, batch 7550, loss[loss=0.2191, simple_loss=0.2922, pruned_loss=0.07302, over 11459.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2874, pruned_loss=0.05729, over 3070571.29 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:02:53,688 INFO [optim.py:368] (0/8) 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 21:03:15,803 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241044.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:20,292 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241047.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:03:29,246 INFO [train.py:904] (0/8) Epoch 24, batch 7600, loss[loss=0.2022, simple_loss=0.2858, pruned_loss=0.0593, over 16467.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2868, pruned_loss=0.05719, over 3096345.93 frames. ], batch size: 68, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:03:48,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6696, 2.5605, 2.3444, 3.4353, 2.4331, 3.6782, 1.5164, 2.8057], device='cuda:0'), covar=tensor([0.1426, 0.0764, 0.1326, 0.0210, 0.0213, 0.0413, 0.1785, 0.0810], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0177, 0.0197, 0.0194, 0.0206, 0.0216, 0.0205, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 21:04:21,729 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3312, 2.9655, 2.7114, 2.3284, 2.3281, 2.3367, 2.9117, 2.8643], device='cuda:0'), covar=tensor([0.2362, 0.0709, 0.1528, 0.2488, 0.2362, 0.2080, 0.0503, 0.1206], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0270, 0.0306, 0.0318, 0.0299, 0.0265, 0.0298, 0.0340], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 21:04:47,285 INFO [train.py:904] (0/8) Epoch 24, batch 7650, loss[loss=0.2012, simple_loss=0.2887, pruned_loss=0.05688, over 16606.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2877, pruned_loss=0.05823, over 3063994.88 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:05:30,620 INFO [optim.py:368] (0/8) 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 21:06:04,732 INFO [train.py:904] (0/8) Epoch 24, batch 7700, loss[loss=0.2227, simple_loss=0.299, pruned_loss=0.07321, over 17253.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2881, pruned_loss=0.05859, over 3076125.30 frames. ], batch size: 52, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:06:09,590 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4248, 4.4161, 4.2899, 3.5410, 4.3774, 1.6201, 4.1175, 3.9260], device='cuda:0'), covar=tensor([0.0140, 0.0126, 0.0219, 0.0395, 0.0101, 0.3044, 0.0157, 0.0290], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0163, 0.0204, 0.0180, 0.0179, 0.0209, 0.0191, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:07:23,966 INFO [train.py:904] (0/8) Epoch 24, batch 7750, loss[loss=0.2258, simple_loss=0.3013, pruned_loss=0.07519, over 11186.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2883, pruned_loss=0.05855, over 3069257.14 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:06,387 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.032e+02 2.844e+02 3.408e+02 3.883e+02 6.664e+02, threshold=6.815e+02, percent-clipped=0.0 2023-05-01 21:08:38,515 INFO [train.py:904] (0/8) Epoch 24, batch 7800, loss[loss=0.1889, simple_loss=0.2799, pruned_loss=0.04896, over 16728.00 frames. ], tot_loss[loss=0.2038, simple_loss=0.2892, pruned_loss=0.05926, over 3053366.15 frames. ], batch size: 124, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:08:52,531 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6890, 1.8538, 1.6534, 1.5165, 1.9812, 1.6054, 1.6547, 1.9229], device='cuda:0'), covar=tensor([0.0194, 0.0258, 0.0372, 0.0336, 0.0212, 0.0239, 0.0190, 0.0186], device='cuda:0'), in_proj_covar=tensor([0.0213, 0.0234, 0.0226, 0.0227, 0.0236, 0.0233, 0.0234, 0.0231], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:09:31,605 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.56 vs. limit=5.0 2023-05-01 21:09:33,425 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241288.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 21:09:56,323 INFO [train.py:904] (0/8) Epoch 24, batch 7850, loss[loss=0.1909, simple_loss=0.2849, pruned_loss=0.04842, over 16343.00 frames. ], tot_loss[loss=0.2046, simple_loss=0.2901, pruned_loss=0.05958, over 3034887.09 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:10:38,479 INFO [optim.py:368] (0/8) 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 21:10:54,883 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241342.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:10:59,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5325, 3.4826, 3.4425, 2.6009, 3.3664, 2.0579, 3.1766, 2.7659], device='cuda:0'), covar=tensor([0.0206, 0.0169, 0.0255, 0.0318, 0.0134, 0.2742, 0.0180, 0.0339], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0163, 0.0205, 0.0181, 0.0179, 0.0210, 0.0191, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:11:11,152 INFO [train.py:904] (0/8) Epoch 24, batch 7900, loss[loss=0.2357, simple_loss=0.3078, pruned_loss=0.08175, over 11593.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.289, pruned_loss=0.05894, over 3044563.30 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:12:24,153 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6670, 1.8100, 1.6192, 1.5453, 1.9210, 1.5948, 1.6231, 1.8827], device='cuda:0'), covar=tensor([0.0186, 0.0273, 0.0417, 0.0328, 0.0227, 0.0257, 0.0188, 0.0197], device='cuda:0'), in_proj_covar=tensor([0.0210, 0.0231, 0.0223, 0.0224, 0.0233, 0.0230, 0.0231, 0.0228], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:12:29,007 INFO [train.py:904] (0/8) Epoch 24, batch 7950, loss[loss=0.1922, simple_loss=0.2779, pruned_loss=0.05318, over 16697.00 frames. ], tot_loss[loss=0.2052, simple_loss=0.29, pruned_loss=0.06015, over 3012008.66 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:12:34,471 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 21:13:15,444 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.79 vs. limit=5.0 2023-05-01 21:13:46,685 INFO [train.py:904] (0/8) Epoch 24, batch 8000, loss[loss=0.1873, simple_loss=0.2842, pruned_loss=0.04516, over 16584.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2914, pruned_loss=0.061, over 3009893.34 frames. ], batch size: 62, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:14:09,543 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241467.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:14:41,183 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.72 vs. limit=5.0 2023-05-01 21:14:45,015 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0662, 3.4244, 3.7139, 1.6163, 3.9090, 3.9728, 2.9265, 2.7729], device='cuda:0'), covar=tensor([0.1322, 0.0262, 0.0208, 0.1548, 0.0102, 0.0186, 0.0520, 0.0659], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0109, 0.0100, 0.0139, 0.0082, 0.0128, 0.0129, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 21:15:04,291 INFO [train.py:904] (0/8) Epoch 24, batch 8050, loss[loss=0.2044, simple_loss=0.2923, pruned_loss=0.05824, over 15459.00 frames. ], tot_loss[loss=0.205, simple_loss=0.2904, pruned_loss=0.05982, over 3040464.98 frames. ], batch size: 191, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:15:47,627 INFO [optim.py:368] (0/8) 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 21:16:22,144 INFO [train.py:904] (0/8) Epoch 24, batch 8100, loss[loss=0.1953, simple_loss=0.2813, pruned_loss=0.05466, over 16570.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2894, pruned_loss=0.05881, over 3050702.32 frames. ], batch size: 68, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:17:14,345 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241588.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:17:20,085 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 21:17:38,331 INFO [train.py:904] (0/8) Epoch 24, batch 8150, loss[loss=0.1662, simple_loss=0.2502, pruned_loss=0.04116, over 16986.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.286, pruned_loss=0.05717, over 3078767.47 frames. ], batch size: 41, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:18:06,812 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241621.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:22,060 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241636.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:18:36,622 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0407, 2.1586, 2.1600, 3.6171, 2.1392, 2.4594, 2.2381, 2.2880], device='cuda:0'), covar=tensor([0.1499, 0.3537, 0.3263, 0.0660, 0.4315, 0.2664, 0.3762, 0.3454], device='cuda:0'), in_proj_covar=tensor([0.0408, 0.0459, 0.0374, 0.0331, 0.0441, 0.0524, 0.0429, 0.0535], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:18:39,442 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=241642.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:18:55,855 INFO [train.py:904] (0/8) Epoch 24, batch 8200, loss[loss=0.1853, simple_loss=0.277, pruned_loss=0.04681, over 16325.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.283, pruned_loss=0.05579, over 3108403.86 frames. ], batch size: 146, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:19:38,360 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1100, 3.9311, 4.2008, 4.3035, 4.4580, 4.0456, 4.3922, 4.4852], device='cuda:0'), covar=tensor([0.1909, 0.1348, 0.1621, 0.0804, 0.0641, 0.1281, 0.0866, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0789, 0.0904, 0.0793, 0.0609, 0.0627, 0.0661, 0.0767], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:19:42,917 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241682.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:19:53,138 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 21:19:56,099 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=241690.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:20:16,393 INFO [train.py:904] (0/8) Epoch 24, batch 8250, loss[loss=0.1719, simple_loss=0.2594, pruned_loss=0.04218, over 11644.00 frames. ], tot_loss[loss=0.1946, simple_loss=0.2824, pruned_loss=0.05343, over 3093621.60 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:03,235 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.735e+02 2.441e+02 2.851e+02 3.404e+02 7.117e+02, threshold=5.702e+02, percent-clipped=1.0 2023-05-01 21:21:05,722 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.63 vs. limit=5.0 2023-05-01 21:21:19,576 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9131, 2.5916, 2.5376, 1.9679, 2.5514, 2.7164, 2.5723, 1.8292], device='cuda:0'), covar=tensor([0.0420, 0.0128, 0.0128, 0.0380, 0.0150, 0.0163, 0.0147, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0085, 0.0086, 0.0133, 0.0098, 0.0110, 0.0095, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-01 21:21:39,006 INFO [train.py:904] (0/8) Epoch 24, batch 8300, loss[loss=0.191, simple_loss=0.2679, pruned_loss=0.05708, over 12053.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.28, pruned_loss=0.05088, over 3072234.59 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:21:41,891 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5464, 3.4905, 3.5119, 2.6641, 3.3787, 2.0531, 3.1758, 2.8193], device='cuda:0'), covar=tensor([0.0161, 0.0159, 0.0186, 0.0256, 0.0122, 0.2441, 0.0152, 0.0249], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0162, 0.0203, 0.0179, 0.0178, 0.0208, 0.0191, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:21:54,711 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241762.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 21:23:02,698 INFO [train.py:904] (0/8) Epoch 24, batch 8350, loss[loss=0.2073, simple_loss=0.2829, pruned_loss=0.06582, over 11886.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2794, pruned_loss=0.04872, over 3078962.33 frames. ], batch size: 246, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:23:03,343 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241803.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:23:37,194 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0554, 3.6953, 4.0472, 2.1938, 4.2670, 4.2800, 3.3261, 3.3604], device='cuda:0'), covar=tensor([0.0571, 0.0240, 0.0211, 0.1123, 0.0070, 0.0151, 0.0341, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0107, 0.0097, 0.0135, 0.0080, 0.0125, 0.0126, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 21:23:48,198 INFO [optim.py:368] (0/8) 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 21:23:48,806 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=241831.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:24:23,538 INFO [train.py:904] (0/8) Epoch 24, batch 8400, loss[loss=0.1781, simple_loss=0.2663, pruned_loss=0.04494, over 12054.00 frames. ], tot_loss[loss=0.185, simple_loss=0.277, pruned_loss=0.04652, over 3093889.45 frames. ], batch size: 247, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:24:42,812 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241864.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:25:14,354 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4330, 2.4022, 2.2458, 4.2314, 2.3120, 2.7094, 2.4243, 2.5221], device='cuda:0'), covar=tensor([0.1205, 0.3604, 0.3300, 0.0463, 0.4310, 0.2599, 0.3730, 0.3346], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0450, 0.0368, 0.0324, 0.0433, 0.0513, 0.0421, 0.0525], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:25:28,728 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=241892.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 21:25:45,946 INFO [train.py:904] (0/8) Epoch 24, batch 8450, loss[loss=0.186, simple_loss=0.2745, pruned_loss=0.04876, over 12355.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2754, pruned_loss=0.0451, over 3087054.16 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:26:31,271 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.437e+02 2.077e+02 2.419e+02 2.958e+02 5.415e+02, threshold=4.839e+02, percent-clipped=2.0 2023-05-01 21:27:07,284 INFO [train.py:904] (0/8) Epoch 24, batch 8500, loss[loss=0.1642, simple_loss=0.2554, pruned_loss=0.03648, over 15240.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2719, pruned_loss=0.04295, over 3104982.49 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:27:47,074 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=241977.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:27:59,884 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5578, 4.8321, 4.6618, 4.6648, 4.4242, 4.4121, 4.3665, 4.9033], device='cuda:0'), covar=tensor([0.1317, 0.1058, 0.1072, 0.0865, 0.0874, 0.1252, 0.1078, 0.0897], device='cuda:0'), in_proj_covar=tensor([0.0684, 0.0821, 0.0682, 0.0638, 0.0522, 0.0530, 0.0692, 0.0644], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:28:25,403 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-242000.pt 2023-05-01 21:28:33,831 INFO [train.py:904] (0/8) Epoch 24, batch 8550, loss[loss=0.1569, simple_loss=0.2569, pruned_loss=0.02846, over 16902.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.27, pruned_loss=0.04214, over 3098922.78 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:29:05,828 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 21:29:26,921 INFO [optim.py:368] (0/8) 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 21:30:13,214 INFO [train.py:904] (0/8) Epoch 24, batch 8600, loss[loss=0.1538, simple_loss=0.2398, pruned_loss=0.03392, over 12103.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2695, pruned_loss=0.04118, over 3076200.35 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:30:31,556 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242062.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:31:51,605 INFO [train.py:904] (0/8) Epoch 24, batch 8650, loss[loss=0.1449, simple_loss=0.2459, pruned_loss=0.02193, over 15287.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2681, pruned_loss=0.04, over 3056024.91 frames. ], batch size: 190, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:32:10,579 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242110.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:32:56,843 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.586e+02 2.159e+02 2.524e+02 3.190e+02 5.272e+02, threshold=5.049e+02, percent-clipped=1.0 2023-05-01 21:33:30,355 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-01 21:33:37,179 INFO [train.py:904] (0/8) Epoch 24, batch 8700, loss[loss=0.1621, simple_loss=0.2512, pruned_loss=0.0365, over 12128.00 frames. ], tot_loss[loss=0.1713, simple_loss=0.2652, pruned_loss=0.03873, over 3061343.87 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:33:38,189 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242153.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:33:38,356 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2543, 2.1702, 2.1903, 3.8598, 2.1884, 2.5252, 2.2875, 2.3220], device='cuda:0'), covar=tensor([0.1264, 0.3885, 0.3325, 0.0526, 0.4151, 0.2665, 0.3732, 0.3484], device='cuda:0'), in_proj_covar=tensor([0.0399, 0.0448, 0.0367, 0.0322, 0.0431, 0.0510, 0.0419, 0.0522], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:33:49,880 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242159.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:34:42,498 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242187.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 21:35:13,457 INFO [train.py:904] (0/8) Epoch 24, batch 8750, loss[loss=0.1754, simple_loss=0.2737, pruned_loss=0.03862, over 16274.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2646, pruned_loss=0.03854, over 3042262.73 frames. ], batch size: 165, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:35:42,278 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242214.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:10,039 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242226.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:36:21,486 INFO [optim.py:368] (0/8) 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 21:37:05,089 INFO [train.py:904] (0/8) Epoch 24, batch 8800, loss[loss=0.1611, simple_loss=0.2522, pruned_loss=0.03502, over 12478.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2632, pruned_loss=0.03716, over 3058477.54 frames. ], batch size: 248, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:37:56,097 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242277.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:17,632 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242287.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:38:23,524 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-01 21:38:50,072 INFO [train.py:904] (0/8) Epoch 24, batch 8850, loss[loss=0.1789, simple_loss=0.2838, pruned_loss=0.03706, over 16658.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2663, pruned_loss=0.03699, over 3055506.19 frames. ], batch size: 134, lr: 2.79e-03, grad_scale: 8.0 2023-05-01 21:39:03,944 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242309.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:38,585 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242325.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:39:51,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6302, 4.4530, 4.7311, 4.8477, 5.0068, 4.5288, 5.0093, 5.0200], device='cuda:0'), covar=tensor([0.1978, 0.1300, 0.1615, 0.0783, 0.0502, 0.0879, 0.0518, 0.0603], device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0772, 0.0884, 0.0780, 0.0596, 0.0616, 0.0647, 0.0750], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:39:52,559 INFO [optim.py:368] (0/8) 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 21:40:33,960 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7770, 4.9993, 5.1709, 4.9255, 4.9992, 5.5091, 5.0291, 4.7605], device='cuda:0'), covar=tensor([0.0984, 0.1646, 0.1720, 0.2069, 0.2155, 0.0862, 0.1382, 0.2210], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0597, 0.0656, 0.0490, 0.0644, 0.0681, 0.0511, 0.0652], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 21:40:38,017 INFO [train.py:904] (0/8) Epoch 24, batch 8900, loss[loss=0.1593, simple_loss=0.2624, pruned_loss=0.02812, over 16898.00 frames. ], tot_loss[loss=0.1698, simple_loss=0.2665, pruned_loss=0.03656, over 3049027.07 frames. ], batch size: 96, lr: 2.79e-03, grad_scale: 4.0 2023-05-01 21:41:12,532 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242370.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:42:06,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8315, 3.7104, 3.9083, 4.0051, 4.0912, 3.6669, 4.0384, 4.1237], device='cuda:0'), covar=tensor([0.1624, 0.1259, 0.1317, 0.0729, 0.0566, 0.1897, 0.0704, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0623, 0.0770, 0.0883, 0.0778, 0.0595, 0.0616, 0.0645, 0.0748], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:42:44,447 INFO [train.py:904] (0/8) Epoch 24, batch 8950, loss[loss=0.1487, simple_loss=0.2457, pruned_loss=0.02591, over 16408.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2666, pruned_loss=0.03692, over 3066905.91 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:43:49,727 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.585e+02 2.299e+02 2.560e+02 2.923e+02 8.049e+02, threshold=5.120e+02, percent-clipped=2.0 2023-05-01 21:44:35,822 INFO [train.py:904] (0/8) Epoch 24, batch 9000, loss[loss=0.1744, simple_loss=0.2582, pruned_loss=0.0453, over 11931.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2631, pruned_loss=0.03581, over 3058571.82 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:44:35,823 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 21:44:45,533 INFO [train.py:938] (0/8) Epoch 24, validation: loss=0.1445, simple_loss=0.2484, pruned_loss=0.02026, over 944034.00 frames. 2023-05-01 21:44:45,534 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 21:44:58,954 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242459.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:14,340 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-01 21:45:21,171 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9821, 4.2368, 4.1108, 4.1012, 3.7653, 3.8550, 3.9101, 4.2436], device='cuda:0'), covar=tensor([0.1104, 0.0892, 0.0872, 0.0818, 0.0807, 0.1485, 0.0901, 0.0951], device='cuda:0'), in_proj_covar=tensor([0.0673, 0.0807, 0.0670, 0.0626, 0.0514, 0.0521, 0.0680, 0.0635], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:45:48,515 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242483.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:45:58,409 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242487.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 21:46:27,733 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5816, 3.6656, 2.8028, 2.2750, 2.2651, 2.4050, 3.8756, 3.1857], device='cuda:0'), covar=tensor([0.3106, 0.0612, 0.1904, 0.2954, 0.2703, 0.2217, 0.0399, 0.1378], device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0265, 0.0301, 0.0312, 0.0291, 0.0262, 0.0292, 0.0334], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 21:46:30,352 INFO [train.py:904] (0/8) Epoch 24, batch 9050, loss[loss=0.17, simple_loss=0.271, pruned_loss=0.03448, over 12471.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.264, pruned_loss=0.03627, over 3057325.21 frames. ], batch size: 246, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:46:40,250 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242507.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:46:44,686 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242509.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:47:02,640 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1220, 2.1689, 2.5695, 2.9926, 2.7913, 3.4892, 2.3849, 3.4678], device='cuda:0'), covar=tensor([0.0204, 0.0500, 0.0363, 0.0303, 0.0358, 0.0168, 0.0492, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0191, 0.0179, 0.0181, 0.0196, 0.0155, 0.0195, 0.0153], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:47:30,076 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.227e+02 2.575e+02 3.137e+02 5.273e+02, threshold=5.150e+02, percent-clipped=1.0 2023-05-01 21:47:35,732 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242535.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 21:47:38,028 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 21:48:00,552 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242544.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:48:17,901 INFO [train.py:904] (0/8) Epoch 24, batch 9100, loss[loss=0.1434, simple_loss=0.2408, pruned_loss=0.02297, over 17116.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2633, pruned_loss=0.03663, over 3062196.67 frames. ], batch size: 49, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:48:23,316 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242555.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:49:27,700 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242582.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:15,754 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242602.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:50:16,000 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-01 21:50:16,956 INFO [train.py:904] (0/8) Epoch 24, batch 9150, loss[loss=0.1634, simple_loss=0.2512, pruned_loss=0.03781, over 16897.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2643, pruned_loss=0.03644, over 3081687.43 frames. ], batch size: 125, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 21:50:46,735 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242616.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:51:18,320 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1825, 4.0612, 4.2891, 4.3952, 4.4969, 4.0611, 4.4802, 4.5366], device='cuda:0'), covar=tensor([0.1786, 0.1227, 0.1314, 0.0641, 0.0576, 0.1265, 0.0741, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0624, 0.0771, 0.0885, 0.0779, 0.0597, 0.0615, 0.0647, 0.0749], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:51:21,274 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.667e+02 2.227e+02 2.578e+02 3.148e+02 6.071e+02, threshold=5.155e+02, percent-clipped=3.0 2023-05-01 21:52:01,747 INFO [train.py:904] (0/8) Epoch 24, batch 9200, loss[loss=0.1673, simple_loss=0.2612, pruned_loss=0.03674, over 16305.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2597, pruned_loss=0.03546, over 3078563.73 frames. ], batch size: 166, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:52:21,795 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242663.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:52:24,776 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242665.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:52:34,100 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 21:53:35,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 21:53:38,107 INFO [train.py:904] (0/8) Epoch 24, batch 9250, loss[loss=0.1575, simple_loss=0.2508, pruned_loss=0.03211, over 16143.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2596, pruned_loss=0.03558, over 3062855.42 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:54:40,873 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 2.148e+02 2.469e+02 2.996e+02 5.681e+02, threshold=4.938e+02, percent-clipped=3.0 2023-05-01 21:55:27,533 INFO [train.py:904] (0/8) Epoch 24, batch 9300, loss[loss=0.169, simple_loss=0.2688, pruned_loss=0.03463, over 16402.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2582, pruned_loss=0.0351, over 3044272.74 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:56:13,863 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242772.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:56:44,418 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8303, 1.3917, 1.6789, 1.7763, 1.8425, 1.9335, 1.7083, 1.8794], device='cuda:0'), covar=tensor([0.0290, 0.0460, 0.0282, 0.0337, 0.0340, 0.0256, 0.0449, 0.0184], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0189, 0.0177, 0.0179, 0.0195, 0.0155, 0.0194, 0.0152], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:57:14,576 INFO [train.py:904] (0/8) Epoch 24, batch 9350, loss[loss=0.1709, simple_loss=0.2522, pruned_loss=0.04474, over 12009.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.258, pruned_loss=0.03503, over 3037766.79 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:57:27,080 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242809.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:57:54,736 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9398, 2.1940, 2.4025, 3.2247, 2.2275, 2.3725, 2.3620, 2.3066], device='cuda:0'), covar=tensor([0.1297, 0.3611, 0.2769, 0.0728, 0.4420, 0.2638, 0.3552, 0.3493], device='cuda:0'), in_proj_covar=tensor([0.0398, 0.0447, 0.0369, 0.0322, 0.0432, 0.0509, 0.0419, 0.0522], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 21:58:13,371 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.321e+02 1.952e+02 2.311e+02 3.086e+02 6.417e+02, threshold=4.623e+02, percent-clipped=1.0 2023-05-01 21:58:14,168 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242833.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:25,327 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242839.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:58:54,549 INFO [train.py:904] (0/8) Epoch 24, batch 9400, loss[loss=0.1492, simple_loss=0.2368, pruned_loss=0.03083, over 12534.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2572, pruned_loss=0.03449, over 3030230.67 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 21:59:03,673 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242857.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 21:59:55,062 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242882.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:00:35,991 INFO [train.py:904] (0/8) Epoch 24, batch 9450, loss[loss=0.176, simple_loss=0.2629, pruned_loss=0.04454, over 12313.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2589, pruned_loss=0.03473, over 3020105.55 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:00:51,469 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=242921.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:32,439 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=242930.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:01:37,831 INFO [optim.py:368] (0/8) 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 22:02:16,745 INFO [train.py:904] (0/8) Epoch 24, batch 9500, loss[loss=0.1528, simple_loss=0.2511, pruned_loss=0.02724, over 17029.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2578, pruned_loss=0.03414, over 3025500.04 frames. ], batch size: 50, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:02:30,117 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=242958.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:02:43,800 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=242965.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:02:45,714 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-01 22:02:59,244 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9684, 5.3195, 5.5052, 5.2244, 5.2805, 5.8347, 5.3754, 5.0088], device='cuda:0'), covar=tensor([0.0860, 0.1672, 0.2005, 0.1894, 0.2230, 0.0785, 0.1478, 0.2356], device='cuda:0'), in_proj_covar=tensor([0.0393, 0.0586, 0.0645, 0.0480, 0.0634, 0.0673, 0.0501, 0.0637], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:03:15,729 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=242982.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:04:01,168 INFO [train.py:904] (0/8) Epoch 24, batch 9550, loss[loss=0.1508, simple_loss=0.2471, pruned_loss=0.02726, over 12454.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2575, pruned_loss=0.03408, over 3023967.72 frames. ], batch size: 246, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:04:24,736 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243013.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:05:06,081 INFO [optim.py:368] (0/8) 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 22:05:43,255 INFO [train.py:904] (0/8) Epoch 24, batch 9600, loss[loss=0.1931, simple_loss=0.3, pruned_loss=0.04313, over 16153.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2594, pruned_loss=0.03537, over 3024324.69 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:05:54,088 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4370, 4.4260, 4.7898, 4.7438, 4.7774, 4.5367, 4.4954, 4.4109], device='cuda:0'), covar=tensor([0.0376, 0.0597, 0.0423, 0.0446, 0.0482, 0.0356, 0.0916, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0402, 0.0448, 0.0438, 0.0403, 0.0480, 0.0456, 0.0532, 0.0367], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 22:06:01,501 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.30 vs. limit=5.0 2023-05-01 22:07:32,101 INFO [train.py:904] (0/8) Epoch 24, batch 9650, loss[loss=0.1677, simple_loss=0.2661, pruned_loss=0.03464, over 16386.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2614, pruned_loss=0.03579, over 3031783.56 frames. ], batch size: 146, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:08:29,801 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243128.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:08:37,735 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.193e+02 2.679e+02 3.134e+02 6.217e+02, threshold=5.359e+02, percent-clipped=4.0 2023-05-01 22:08:42,596 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4339, 3.0718, 2.7527, 2.2699, 2.2020, 2.3308, 3.0137, 2.8645], device='cuda:0'), covar=tensor([0.2692, 0.0678, 0.1612, 0.2769, 0.2576, 0.2188, 0.0451, 0.1466], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0263, 0.0300, 0.0310, 0.0288, 0.0260, 0.0290, 0.0331], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:08:50,200 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243139.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:09:19,549 INFO [train.py:904] (0/8) Epoch 24, batch 9700, loss[loss=0.159, simple_loss=0.2494, pruned_loss=0.03424, over 12895.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2603, pruned_loss=0.03555, over 3032275.78 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:10:30,930 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243187.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:01,233 INFO [train.py:904] (0/8) Epoch 24, batch 9750, loss[loss=0.1886, simple_loss=0.2669, pruned_loss=0.05514, over 12199.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2594, pruned_loss=0.03596, over 3039681.85 frames. ], batch size: 247, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:11:16,953 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243211.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:11:54,831 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8627, 5.1106, 5.2353, 5.0053, 5.1199, 5.6243, 5.1091, 4.8322], device='cuda:0'), covar=tensor([0.0918, 0.1844, 0.2464, 0.1887, 0.2088, 0.0914, 0.1629, 0.2392], device='cuda:0'), in_proj_covar=tensor([0.0388, 0.0579, 0.0638, 0.0474, 0.0625, 0.0662, 0.0495, 0.0628], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:12:03,365 INFO [optim.py:368] (0/8) 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 22:12:37,399 INFO [train.py:904] (0/8) Epoch 24, batch 9800, loss[loss=0.1493, simple_loss=0.2396, pruned_loss=0.02948, over 12255.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2594, pruned_loss=0.03507, over 3032747.46 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:12:49,021 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243258.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:12:50,365 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243259.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:13:22,150 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243277.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 22:13:35,453 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-01 22:13:58,482 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=243293.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:14:01,633 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-01 22:14:21,694 INFO [train.py:904] (0/8) Epoch 24, batch 9850, loss[loss=0.1769, simple_loss=0.2732, pruned_loss=0.04031, over 16249.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2606, pruned_loss=0.03485, over 3037881.25 frames. ], batch size: 165, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:14:28,525 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243306.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:15:22,939 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.302e+02 1.939e+02 2.274e+02 2.876e+02 6.181e+02, threshold=4.547e+02, percent-clipped=2.0 2023-05-01 22:16:11,639 INFO [train.py:904] (0/8) Epoch 24, batch 9900, loss[loss=0.1851, simple_loss=0.2845, pruned_loss=0.04284, over 16793.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2613, pruned_loss=0.0347, over 3047776.36 frames. ], batch size: 134, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:16:15,147 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=243354.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:18:09,548 INFO [train.py:904] (0/8) Epoch 24, batch 9950, loss[loss=0.1643, simple_loss=0.2656, pruned_loss=0.03156, over 16912.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2641, pruned_loss=0.03482, over 3078295.04 frames. ], batch size: 102, lr: 2.78e-03, grad_scale: 4.0 2023-05-01 22:19:11,958 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243428.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:19:24,207 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-01 22:19:24,684 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.448e+02 2.110e+02 2.470e+02 3.179e+02 5.008e+02, threshold=4.941e+02, percent-clipped=4.0 2023-05-01 22:19:47,276 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4271, 4.5788, 4.6920, 4.4512, 4.5399, 5.0446, 4.6027, 4.2806], device='cuda:0'), covar=tensor([0.1323, 0.1863, 0.1970, 0.2081, 0.2337, 0.0946, 0.1587, 0.2467], device='cuda:0'), in_proj_covar=tensor([0.0386, 0.0576, 0.0634, 0.0472, 0.0623, 0.0658, 0.0493, 0.0625], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:20:10,934 INFO [train.py:904] (0/8) Epoch 24, batch 10000, loss[loss=0.1728, simple_loss=0.2588, pruned_loss=0.0434, over 12906.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2631, pruned_loss=0.03461, over 3072325.68 frames. ], batch size: 248, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:20:25,360 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1981, 3.5511, 3.5808, 2.5238, 3.2964, 3.6261, 3.4103, 2.0550], device='cuda:0'), covar=tensor([0.0628, 0.0058, 0.0062, 0.0402, 0.0118, 0.0086, 0.0077, 0.0556], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0084, 0.0084, 0.0132, 0.0097, 0.0107, 0.0093, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:20:55,210 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243476.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:21:51,124 INFO [train.py:904] (0/8) Epoch 24, batch 10050, loss[loss=0.1568, simple_loss=0.259, pruned_loss=0.02732, over 16824.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2626, pruned_loss=0.03438, over 3070037.14 frames. ], batch size: 83, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:22:52,104 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.427e+02 2.109e+02 2.585e+02 3.087e+02 5.366e+02, threshold=5.170e+02, percent-clipped=3.0 2023-05-01 22:23:25,158 INFO [train.py:904] (0/8) Epoch 24, batch 10100, loss[loss=0.1501, simple_loss=0.2373, pruned_loss=0.0315, over 12677.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2627, pruned_loss=0.03464, over 3069882.42 frames. ], batch size: 250, lr: 2.78e-03, grad_scale: 8.0 2023-05-01 22:24:15,453 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243577.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 22:24:46,537 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-24.pt 2023-05-01 22:25:09,207 INFO [train.py:904] (0/8) Epoch 25, batch 0, loss[loss=0.1768, simple_loss=0.2655, pruned_loss=0.0441, over 17119.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2655, pruned_loss=0.0441, over 17119.00 frames. ], batch size: 47, lr: 2.72e-03, grad_scale: 8.0 2023-05-01 22:25:09,208 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 22:25:16,826 INFO [train.py:938] (0/8) Epoch 25, validation: loss=0.1443, simple_loss=0.2477, pruned_loss=0.02048, over 944034.00 frames. 2023-05-01 22:25:16,826 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 22:25:48,058 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243625.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:26:03,180 INFO [optim.py:368] (0/8) 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 22:26:21,440 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=243649.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:26:26,915 INFO [train.py:904] (0/8) Epoch 25, batch 50, loss[loss=0.211, simple_loss=0.2851, pruned_loss=0.06847, over 16758.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.2628, pruned_loss=0.0455, over 753237.27 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:27:35,827 INFO [train.py:904] (0/8) Epoch 25, batch 100, loss[loss=0.2096, simple_loss=0.2892, pruned_loss=0.06502, over 12205.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2627, pruned_loss=0.04597, over 1315175.12 frames. ], batch size: 246, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:28:22,082 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.579e+02 2.266e+02 2.864e+02 3.484e+02 1.313e+03, threshold=5.728e+02, percent-clipped=5.0 2023-05-01 22:28:45,151 INFO [train.py:904] (0/8) Epoch 25, batch 150, loss[loss=0.1929, simple_loss=0.271, pruned_loss=0.0574, over 16296.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2602, pruned_loss=0.04485, over 1760767.31 frames. ], batch size: 165, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:29:24,929 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0228, 2.6653, 2.6095, 4.3537, 3.4572, 4.0998, 1.6222, 3.0078], device='cuda:0'), covar=tensor([0.1294, 0.0753, 0.1200, 0.0166, 0.0192, 0.0417, 0.1655, 0.0837], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0175, 0.0195, 0.0190, 0.0199, 0.0213, 0.0205, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:29:28,313 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0496, 5.5361, 5.6361, 5.3913, 5.4073, 6.0115, 5.5090, 5.2936], device='cuda:0'), covar=tensor([0.0999, 0.1931, 0.2408, 0.2044, 0.2763, 0.0963, 0.1500, 0.2136], device='cuda:0'), in_proj_covar=tensor([0.0396, 0.0592, 0.0652, 0.0484, 0.0637, 0.0674, 0.0505, 0.0641], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:29:55,594 INFO [train.py:904] (0/8) Epoch 25, batch 200, loss[loss=0.2037, simple_loss=0.2879, pruned_loss=0.05978, over 12140.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2605, pruned_loss=0.04434, over 2108563.34 frames. ], batch size: 246, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:29:58,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8373, 4.5963, 4.8857, 5.0353, 5.2232, 4.6537, 5.1949, 5.2205], device='cuda:0'), covar=tensor([0.1985, 0.1447, 0.1757, 0.0838, 0.0647, 0.0932, 0.0699, 0.0708], device='cuda:0'), in_proj_covar=tensor([0.0630, 0.0777, 0.0892, 0.0785, 0.0602, 0.0618, 0.0652, 0.0754], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:30:40,585 INFO [optim.py:368] (0/8) 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 22:30:45,500 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3413, 2.4055, 2.5664, 4.1372, 2.4294, 2.8109, 2.4679, 2.6125], device='cuda:0'), covar=tensor([0.1406, 0.3273, 0.2747, 0.0637, 0.3633, 0.2210, 0.3394, 0.2935], device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0453, 0.0375, 0.0327, 0.0438, 0.0517, 0.0426, 0.0530], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:31:04,169 INFO [train.py:904] (0/8) Epoch 25, batch 250, loss[loss=0.164, simple_loss=0.2507, pruned_loss=0.03866, over 17225.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2588, pruned_loss=0.04421, over 2383783.59 frames. ], batch size: 44, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:32:14,701 INFO [train.py:904] (0/8) Epoch 25, batch 300, loss[loss=0.1754, simple_loss=0.2659, pruned_loss=0.0424, over 16763.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2563, pruned_loss=0.04272, over 2589391.94 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:32:37,463 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7471, 5.0097, 5.1135, 4.9586, 4.9823, 5.5952, 5.0115, 4.7377], device='cuda:0'), covar=tensor([0.1447, 0.2077, 0.2893, 0.2230, 0.2656, 0.1079, 0.2157, 0.2830], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0603, 0.0665, 0.0493, 0.0650, 0.0685, 0.0514, 0.0652], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:33:00,866 INFO [optim.py:368] (0/8) 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 22:33:03,770 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9673, 4.0546, 2.7895, 4.7147, 3.3002, 4.6531, 2.7449, 3.4826], device='cuda:0'), covar=tensor([0.0321, 0.0374, 0.1462, 0.0272, 0.0800, 0.0457, 0.1458, 0.0711], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0177, 0.0193, 0.0166, 0.0177, 0.0215, 0.0203, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:33:19,460 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=243949.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:33:24,548 INFO [train.py:904] (0/8) Epoch 25, batch 350, loss[loss=0.1535, simple_loss=0.2374, pruned_loss=0.03486, over 15926.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2533, pruned_loss=0.04151, over 2738308.11 frames. ], batch size: 35, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:34:25,844 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=243997.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:34:30,122 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-244000.pt 2023-05-01 22:34:38,365 INFO [train.py:904] (0/8) Epoch 25, batch 400, loss[loss=0.1584, simple_loss=0.2436, pruned_loss=0.0366, over 16583.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2518, pruned_loss=0.04039, over 2864231.73 frames. ], batch size: 68, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:34:50,894 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9283, 2.6862, 2.6064, 4.2987, 3.5094, 4.0766, 1.6374, 2.9112], device='cuda:0'), covar=tensor([0.1462, 0.0776, 0.1232, 0.0194, 0.0242, 0.0471, 0.1765, 0.0927], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0176, 0.0195, 0.0191, 0.0200, 0.0213, 0.0204, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:34:54,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6379, 3.8231, 2.9414, 2.2243, 2.4240, 2.3998, 3.9380, 3.2601], device='cuda:0'), covar=tensor([0.2960, 0.0589, 0.1743, 0.3298, 0.2872, 0.2248, 0.0524, 0.1562], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0268, 0.0305, 0.0315, 0.0294, 0.0265, 0.0295, 0.0339], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:35:02,335 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.57 vs. limit=2.0 2023-05-01 22:35:22,992 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.463e+02 2.193e+02 2.643e+02 3.297e+02 6.329e+02, threshold=5.285e+02, percent-clipped=2.0 2023-05-01 22:35:42,203 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.94 vs. limit=5.0 2023-05-01 22:35:47,044 INFO [train.py:904] (0/8) Epoch 25, batch 450, loss[loss=0.1816, simple_loss=0.2574, pruned_loss=0.05286, over 16887.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2513, pruned_loss=0.04002, over 2970479.61 frames. ], batch size: 109, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:36:50,454 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4691, 4.4187, 4.8316, 4.8051, 4.8457, 4.5751, 4.5472, 4.4295], device='cuda:0'), covar=tensor([0.0393, 0.0763, 0.0385, 0.0406, 0.0476, 0.0452, 0.0834, 0.0608], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0459, 0.0447, 0.0412, 0.0491, 0.0470, 0.0546, 0.0375], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 22:36:55,218 INFO [train.py:904] (0/8) Epoch 25, batch 500, loss[loss=0.1607, simple_loss=0.2526, pruned_loss=0.03435, over 17140.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.251, pruned_loss=0.03916, over 3037730.04 frames. ], batch size: 48, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:37:18,533 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1803, 4.4707, 4.4496, 3.3247, 3.8787, 4.4941, 4.0245, 2.6446], device='cuda:0'), covar=tensor([0.0465, 0.0069, 0.0047, 0.0356, 0.0127, 0.0082, 0.0083, 0.0498], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0088, 0.0087, 0.0136, 0.0100, 0.0111, 0.0096, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-01 22:37:42,050 INFO [optim.py:368] (0/8) 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 22:38:05,723 INFO [train.py:904] (0/8) Epoch 25, batch 550, loss[loss=0.1649, simple_loss=0.2608, pruned_loss=0.03455, over 17140.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2501, pruned_loss=0.03871, over 3098337.56 frames. ], batch size: 47, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:39:15,774 INFO [train.py:904] (0/8) Epoch 25, batch 600, loss[loss=0.1813, simple_loss=0.2487, pruned_loss=0.05697, over 16883.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2496, pruned_loss=0.03935, over 3138460.53 frames. ], batch size: 116, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:39:18,619 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2694, 5.2474, 5.1417, 4.5955, 4.7664, 5.1670, 5.0400, 4.7611], device='cuda:0'), covar=tensor([0.0561, 0.0458, 0.0300, 0.0370, 0.1093, 0.0435, 0.0325, 0.0843], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0446, 0.0346, 0.0349, 0.0350, 0.0401, 0.0239, 0.0418], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:40:02,991 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.384e+02 2.116e+02 2.478e+02 2.900e+02 1.597e+03, threshold=4.957e+02, percent-clipped=3.0 2023-05-01 22:40:25,356 INFO [train.py:904] (0/8) Epoch 25, batch 650, loss[loss=0.1528, simple_loss=0.2355, pruned_loss=0.03507, over 15600.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2491, pruned_loss=0.03934, over 3171842.54 frames. ], batch size: 190, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:40:32,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1628, 2.2079, 2.4015, 3.8582, 2.2296, 2.5553, 2.2775, 2.3979], device='cuda:0'), covar=tensor([0.1489, 0.3888, 0.3050, 0.0683, 0.3906, 0.2678, 0.4049, 0.3061], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0459, 0.0378, 0.0332, 0.0443, 0.0524, 0.0432, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:40:53,532 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3632, 3.3414, 2.1974, 3.5563, 2.6490, 3.5591, 2.2944, 2.7916], device='cuda:0'), covar=tensor([0.0334, 0.0505, 0.1541, 0.0378, 0.0808, 0.0846, 0.1469, 0.0732], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0169, 0.0179, 0.0218, 0.0205, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:40:53,571 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7874, 2.4663, 2.4587, 3.8502, 3.0278, 3.8735, 1.6051, 2.8426], device='cuda:0'), covar=tensor([0.1423, 0.0815, 0.1245, 0.0224, 0.0158, 0.0401, 0.1722, 0.0860], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0175, 0.0194, 0.0191, 0.0199, 0.0213, 0.0204, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:41:08,473 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8350, 4.0078, 2.7640, 4.6058, 3.2344, 4.5373, 2.7109, 3.3565], device='cuda:0'), covar=tensor([0.0368, 0.0404, 0.1508, 0.0344, 0.0878, 0.0572, 0.1601, 0.0733], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0179, 0.0195, 0.0169, 0.0179, 0.0218, 0.0205, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:41:33,627 INFO [train.py:904] (0/8) Epoch 25, batch 700, loss[loss=0.167, simple_loss=0.2602, pruned_loss=0.03689, over 17081.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2483, pruned_loss=0.03854, over 3206810.46 frames. ], batch size: 53, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:41:50,755 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7434, 4.9421, 5.0512, 4.8447, 4.9105, 5.5313, 5.0027, 4.6903], device='cuda:0'), covar=tensor([0.1477, 0.2176, 0.2805, 0.2312, 0.2965, 0.1077, 0.1887, 0.2722], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0612, 0.0677, 0.0502, 0.0665, 0.0698, 0.0525, 0.0665], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:41:55,081 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3639, 4.4373, 4.5654, 4.3495, 4.4218, 5.0114, 4.5219, 4.2038], device='cuda:0'), covar=tensor([0.1759, 0.2256, 0.3104, 0.2324, 0.3004, 0.1198, 0.1825, 0.2734], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0613, 0.0677, 0.0503, 0.0666, 0.0699, 0.0525, 0.0665], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:42:16,533 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5733, 5.9989, 5.7902, 5.8206, 5.3940, 5.4623, 5.3435, 6.1359], device='cuda:0'), covar=tensor([0.1495, 0.1064, 0.0982, 0.0873, 0.0863, 0.0599, 0.1230, 0.0844], device='cuda:0'), in_proj_covar=tensor([0.0700, 0.0845, 0.0692, 0.0651, 0.0535, 0.0541, 0.0710, 0.0661], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:42:20,955 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 2.146e+02 2.447e+02 2.875e+02 6.652e+02, threshold=4.894e+02, percent-clipped=4.0 2023-05-01 22:42:42,420 INFO [train.py:904] (0/8) Epoch 25, batch 750, loss[loss=0.1489, simple_loss=0.2323, pruned_loss=0.03278, over 16402.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2489, pruned_loss=0.03881, over 3230142.43 frames. ], batch size: 68, lr: 2.72e-03, grad_scale: 2.0 2023-05-01 22:42:42,779 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244353.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:43:11,817 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9607, 4.4496, 3.0712, 2.3516, 2.8603, 2.6186, 4.6699, 3.6314], device='cuda:0'), covar=tensor([0.2893, 0.0587, 0.2012, 0.3086, 0.2959, 0.2243, 0.0455, 0.1522], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0269, 0.0308, 0.0317, 0.0297, 0.0267, 0.0298, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-01 22:43:34,882 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3662, 4.2876, 4.2614, 3.9447, 4.0100, 4.3194, 4.0155, 4.0728], device='cuda:0'), covar=tensor([0.0667, 0.0914, 0.0376, 0.0348, 0.0867, 0.0469, 0.0810, 0.0753], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0451, 0.0351, 0.0353, 0.0355, 0.0407, 0.0242, 0.0424], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:43:52,160 INFO [train.py:904] (0/8) Epoch 25, batch 800, loss[loss=0.1475, simple_loss=0.2325, pruned_loss=0.03122, over 16841.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2494, pruned_loss=0.03882, over 3248816.79 frames. ], batch size: 42, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:44:08,208 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244414.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:44:13,694 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9245, 3.6759, 4.1853, 2.1351, 4.3410, 4.4141, 3.2222, 3.4753], device='cuda:0'), covar=tensor([0.0769, 0.0316, 0.0247, 0.1219, 0.0092, 0.0243, 0.0492, 0.0431], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0140, 0.0083, 0.0130, 0.0129, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:44:39,345 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.613e+02 2.216e+02 2.518e+02 3.114e+02 7.971e+02, threshold=5.036e+02, percent-clipped=2.0 2023-05-01 22:45:03,354 INFO [train.py:904] (0/8) Epoch 25, batch 850, loss[loss=0.1591, simple_loss=0.2538, pruned_loss=0.0322, over 16787.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2485, pruned_loss=0.03854, over 3267322.76 frames. ], batch size: 62, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:46:12,675 INFO [train.py:904] (0/8) Epoch 25, batch 900, loss[loss=0.1593, simple_loss=0.2544, pruned_loss=0.03206, over 16437.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2475, pruned_loss=0.03806, over 3285263.56 frames. ], batch size: 75, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:46:34,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0063, 5.0114, 4.7967, 4.2561, 4.8897, 1.9299, 4.6385, 4.6241], device='cuda:0'), covar=tensor([0.0102, 0.0091, 0.0233, 0.0429, 0.0118, 0.2909, 0.0147, 0.0243], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0166, 0.0205, 0.0180, 0.0182, 0.0213, 0.0194, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:47:00,726 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.397e+02 1.999e+02 2.403e+02 2.874e+02 4.869e+02, threshold=4.806e+02, percent-clipped=0.0 2023-05-01 22:47:03,427 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-01 22:47:23,404 INFO [train.py:904] (0/8) Epoch 25, batch 950, loss[loss=0.1479, simple_loss=0.2351, pruned_loss=0.03035, over 17021.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2475, pruned_loss=0.03807, over 3292071.58 frames. ], batch size: 41, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:47:44,411 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-01 22:48:27,768 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-01 22:48:33,905 INFO [train.py:904] (0/8) Epoch 25, batch 1000, loss[loss=0.1687, simple_loss=0.2366, pruned_loss=0.05039, over 16759.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2466, pruned_loss=0.03793, over 3293278.65 frames. ], batch size: 124, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:49:10,343 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-01 22:49:21,002 INFO [optim.py:368] (0/8) 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 22:49:42,540 INFO [train.py:904] (0/8) Epoch 25, batch 1050, loss[loss=0.1526, simple_loss=0.2322, pruned_loss=0.03646, over 16829.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.2456, pruned_loss=0.03776, over 3294211.76 frames. ], batch size: 96, lr: 2.72e-03, grad_scale: 4.0 2023-05-01 22:50:21,610 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7401, 3.8340, 2.4956, 4.3109, 3.0565, 4.3412, 2.5988, 3.2163], device='cuda:0'), covar=tensor([0.0308, 0.0438, 0.1660, 0.0363, 0.0816, 0.0568, 0.1560, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0179, 0.0196, 0.0170, 0.0179, 0.0219, 0.0204, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:50:39,803 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8003, 4.8557, 5.1962, 5.1835, 5.2148, 4.9129, 4.8623, 4.7088], device='cuda:0'), covar=tensor([0.0399, 0.0641, 0.0486, 0.0454, 0.0477, 0.0464, 0.0938, 0.0508], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0476, 0.0464, 0.0428, 0.0509, 0.0488, 0.0566, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 22:50:53,106 INFO [train.py:904] (0/8) Epoch 25, batch 1100, loss[loss=0.1504, simple_loss=0.2469, pruned_loss=0.02693, over 17101.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2448, pruned_loss=0.03722, over 3297775.22 frames. ], batch size: 47, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:51:01,211 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=244709.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:51:13,176 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244717.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 22:51:40,295 INFO [optim.py:368] (0/8) 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 22:52:02,018 INFO [train.py:904] (0/8) Epoch 25, batch 1150, loss[loss=0.1429, simple_loss=0.2253, pruned_loss=0.03028, over 16814.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2448, pruned_loss=0.03684, over 3297791.85 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 22:52:36,508 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244778.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 22:53:11,759 INFO [train.py:904] (0/8) Epoch 25, batch 1200, loss[loss=0.1436, simple_loss=0.228, pruned_loss=0.02961, over 16795.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2444, pruned_loss=0.03629, over 3304718.39 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:53:57,424 INFO [optim.py:368] (0/8) 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 22:54:19,243 INFO [train.py:904] (0/8) Epoch 25, batch 1250, loss[loss=0.1875, simple_loss=0.2774, pruned_loss=0.04882, over 17040.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2447, pruned_loss=0.03686, over 3313672.24 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:54:37,918 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2472, 2.8545, 3.1775, 1.9165, 3.2473, 3.2738, 2.8217, 2.5540], device='cuda:0'), covar=tensor([0.0843, 0.0328, 0.0253, 0.1108, 0.0151, 0.0285, 0.0463, 0.0491], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0110, 0.0100, 0.0139, 0.0083, 0.0130, 0.0129, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:55:06,362 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244887.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:55:20,092 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-01 22:55:26,321 INFO [train.py:904] (0/8) Epoch 25, batch 1300, loss[loss=0.1972, simple_loss=0.2668, pruned_loss=0.06383, over 16866.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2449, pruned_loss=0.03689, over 3325145.55 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:56:12,139 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.526e+02 2.186e+02 2.539e+02 3.062e+02 9.189e+02, threshold=5.078e+02, percent-clipped=4.0 2023-05-01 22:56:27,616 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=244948.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:56:35,076 INFO [train.py:904] (0/8) Epoch 25, batch 1350, loss[loss=0.1709, simple_loss=0.2602, pruned_loss=0.04083, over 17051.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2456, pruned_loss=0.03695, over 3328555.88 frames. ], batch size: 50, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:56:36,697 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=244954.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:57:43,165 INFO [train.py:904] (0/8) Epoch 25, batch 1400, loss[loss=0.1406, simple_loss=0.2248, pruned_loss=0.02818, over 15444.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2464, pruned_loss=0.03753, over 3317221.90 frames. ], batch size: 190, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:57:51,760 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245009.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:57:59,988 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245015.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:58:28,654 INFO [optim.py:368] (0/8) 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 22:58:31,539 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9209, 4.9775, 5.3937, 5.3691, 5.3847, 5.0683, 5.0340, 4.8253], device='cuda:0'), covar=tensor([0.0379, 0.0562, 0.0415, 0.0431, 0.0479, 0.0438, 0.0842, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0478, 0.0464, 0.0429, 0.0510, 0.0490, 0.0567, 0.0388], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 22:58:37,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1754, 5.1366, 5.0037, 4.4224, 4.6232, 5.0907, 5.0252, 4.6587], device='cuda:0'), covar=tensor([0.0690, 0.0579, 0.0350, 0.0426, 0.1262, 0.0535, 0.0329, 0.0856], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0458, 0.0357, 0.0359, 0.0362, 0.0413, 0.0245, 0.0431], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-01 22:58:51,290 INFO [train.py:904] (0/8) Epoch 25, batch 1450, loss[loss=0.1647, simple_loss=0.2375, pruned_loss=0.04593, over 16839.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2457, pruned_loss=0.0379, over 3313047.89 frames. ], batch size: 96, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 22:58:56,916 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245057.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 22:59:00,704 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9282, 4.0427, 2.9242, 4.6753, 3.3597, 4.5954, 2.8747, 3.4859], device='cuda:0'), covar=tensor([0.0329, 0.0417, 0.1421, 0.0258, 0.0794, 0.0571, 0.1382, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0180, 0.0196, 0.0171, 0.0180, 0.0221, 0.0206, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 22:59:18,206 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245073.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:00:00,093 INFO [train.py:904] (0/8) Epoch 25, batch 1500, loss[loss=0.1664, simple_loss=0.2455, pruned_loss=0.04361, over 16683.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2463, pruned_loss=0.03834, over 3309958.00 frames. ], batch size: 76, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:00:28,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1677, 4.1698, 4.1292, 3.5406, 4.1657, 1.8877, 3.9521, 3.6656], device='cuda:0'), covar=tensor([0.0159, 0.0150, 0.0207, 0.0277, 0.0116, 0.2743, 0.0159, 0.0246], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0169, 0.0209, 0.0183, 0.0186, 0.0215, 0.0198, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:00:46,573 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.188e+02 2.473e+02 2.927e+02 4.606e+02, threshold=4.945e+02, percent-clipped=0.0 2023-05-01 23:00:52,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8964, 4.0499, 2.3731, 4.7066, 3.1192, 4.5409, 2.3314, 3.2868], device='cuda:0'), covar=tensor([0.0383, 0.0400, 0.1982, 0.0247, 0.0927, 0.0460, 0.2136, 0.0809], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0172, 0.0181, 0.0223, 0.0207, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:01:08,941 INFO [train.py:904] (0/8) Epoch 25, batch 1550, loss[loss=0.1652, simple_loss=0.257, pruned_loss=0.03672, over 17113.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2467, pruned_loss=0.03907, over 3318525.81 frames. ], batch size: 48, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:01:11,842 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9821, 2.2161, 2.3187, 2.6520, 2.0174, 3.2108, 1.7901, 2.7342], device='cuda:0'), covar=tensor([0.1126, 0.0754, 0.1106, 0.0175, 0.0121, 0.0380, 0.1463, 0.0698], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0196, 0.0204, 0.0218, 0.0206, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:02:19,898 INFO [train.py:904] (0/8) Epoch 25, batch 1600, loss[loss=0.173, simple_loss=0.2774, pruned_loss=0.03428, over 17142.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2491, pruned_loss=0.04005, over 3317049.57 frames. ], batch size: 49, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:02:25,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8565, 2.7598, 2.5894, 4.2028, 3.4480, 4.0984, 1.6613, 2.9988], device='cuda:0'), covar=tensor([0.1396, 0.0692, 0.1158, 0.0199, 0.0158, 0.0398, 0.1595, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0196, 0.0203, 0.0218, 0.0206, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:02:33,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7961, 3.5641, 4.0459, 2.1639, 4.0540, 4.0675, 3.3026, 3.0671], device='cuda:0'), covar=tensor([0.0772, 0.0278, 0.0195, 0.1178, 0.0122, 0.0214, 0.0416, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0140, 0.0084, 0.0131, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:02:48,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8102, 2.8034, 2.7103, 4.9964, 3.9727, 4.3904, 1.6548, 3.2127], device='cuda:0'), covar=tensor([0.1426, 0.0855, 0.1273, 0.0213, 0.0197, 0.0382, 0.1691, 0.0783], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0196, 0.0204, 0.0218, 0.0206, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:03:06,949 INFO [optim.py:368] (0/8) 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 23:03:16,809 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245243.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:03:29,747 INFO [train.py:904] (0/8) Epoch 25, batch 1650, loss[loss=0.1762, simple_loss=0.2506, pruned_loss=0.0509, over 16889.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2502, pruned_loss=0.0403, over 3310601.45 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:03:50,393 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6225, 3.6543, 2.2822, 3.9212, 2.7964, 3.9030, 2.3211, 2.9524], device='cuda:0'), covar=tensor([0.0257, 0.0360, 0.1548, 0.0346, 0.0798, 0.0701, 0.1475, 0.0725], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0173, 0.0181, 0.0222, 0.0206, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:04:41,937 INFO [train.py:904] (0/8) Epoch 25, batch 1700, loss[loss=0.1838, simple_loss=0.2683, pruned_loss=0.04965, over 16829.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2509, pruned_loss=0.03986, over 3316215.79 frames. ], batch size: 83, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:04:51,856 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245310.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:04:52,093 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9400, 2.6948, 2.8055, 2.0863, 2.6277, 2.1047, 2.6850, 2.8810], device='cuda:0'), covar=tensor([0.0291, 0.0828, 0.0585, 0.1785, 0.0850, 0.0959, 0.0592, 0.0847], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0166, 0.0169, 0.0155, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:05:30,612 INFO [optim.py:368] (0/8) 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 23:05:52,661 INFO [train.py:904] (0/8) Epoch 25, batch 1750, loss[loss=0.1962, simple_loss=0.2856, pruned_loss=0.05336, over 16652.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2519, pruned_loss=0.04003, over 3313381.97 frames. ], batch size: 62, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:06:20,908 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245373.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:07:03,962 INFO [train.py:904] (0/8) Epoch 25, batch 1800, loss[loss=0.1667, simple_loss=0.2771, pruned_loss=0.0282, over 17286.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2536, pruned_loss=0.03993, over 3316472.42 frames. ], batch size: 52, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:07:29,361 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245421.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:07:51,081 INFO [optim.py:368] (0/8) 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 23:08:12,921 INFO [train.py:904] (0/8) Epoch 25, batch 1850, loss[loss=0.1382, simple_loss=0.2254, pruned_loss=0.02549, over 16790.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2539, pruned_loss=0.03954, over 3329321.44 frames. ], batch size: 39, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:08:21,633 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5457, 4.5948, 4.9440, 4.9121, 4.9651, 4.6449, 4.6528, 4.5174], device='cuda:0'), covar=tensor([0.0447, 0.0745, 0.0454, 0.0467, 0.0616, 0.0486, 0.0939, 0.0623], device='cuda:0'), in_proj_covar=tensor([0.0433, 0.0482, 0.0469, 0.0432, 0.0516, 0.0495, 0.0573, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 23:08:23,492 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245460.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:09:23,473 INFO [train.py:904] (0/8) Epoch 25, batch 1900, loss[loss=0.169, simple_loss=0.2614, pruned_loss=0.0383, over 17122.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2527, pruned_loss=0.03892, over 3333982.72 frames. ], batch size: 48, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:09:49,179 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245521.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:11,814 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.101e+02 2.496e+02 2.952e+02 1.304e+03, threshold=4.992e+02, percent-clipped=2.0 2023-05-01 23:10:20,409 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245543.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:25,096 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245546.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:10:33,952 INFO [train.py:904] (0/8) Epoch 25, batch 1950, loss[loss=0.1852, simple_loss=0.271, pruned_loss=0.0497, over 12156.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2529, pruned_loss=0.03874, over 3329175.10 frames. ], batch size: 245, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:26,901 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245591.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:11:41,409 INFO [train.py:904] (0/8) Epoch 25, batch 2000, loss[loss=0.177, simple_loss=0.2563, pruned_loss=0.04883, over 16791.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2525, pruned_loss=0.03854, over 3331519.53 frames. ], batch size: 124, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:11:46,173 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4089, 3.4173, 4.0131, 2.1381, 3.2893, 2.5145, 3.9080, 3.6409], device='cuda:0'), covar=tensor([0.0245, 0.1051, 0.0477, 0.2072, 0.0794, 0.1020, 0.0553, 0.1082], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0168, 0.0171, 0.0156, 0.0147, 0.0132, 0.0146, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:11:46,737 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-01 23:11:49,089 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245607.0, num_to_drop=1, layers_to_drop={0} 2023-05-01 23:11:53,139 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=245610.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:12:20,558 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0137, 2.2095, 2.6776, 2.9627, 2.7370, 3.4651, 2.6514, 3.5457], device='cuda:0'), covar=tensor([0.0284, 0.0560, 0.0359, 0.0391, 0.0419, 0.0237, 0.0482, 0.0182], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0198, 0.0184, 0.0189, 0.0205, 0.0162, 0.0201, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:12:31,549 INFO [optim.py:368] (0/8) 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 23:12:50,409 INFO [train.py:904] (0/8) Epoch 25, batch 2050, loss[loss=0.1744, simple_loss=0.2745, pruned_loss=0.03718, over 17121.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2525, pruned_loss=0.03876, over 3324592.65 frames. ], batch size: 48, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:12:57,320 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=245658.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:13:19,025 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4460, 4.0656, 4.5131, 2.5697, 4.7019, 4.7753, 3.6191, 3.8500], device='cuda:0'), covar=tensor([0.0620, 0.0281, 0.0210, 0.1057, 0.0081, 0.0178, 0.0389, 0.0385], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0084, 0.0131, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:13:34,851 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245686.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:13:58,112 INFO [train.py:904] (0/8) Epoch 25, batch 2100, loss[loss=0.161, simple_loss=0.2456, pruned_loss=0.03814, over 16838.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2528, pruned_loss=0.03901, over 3320443.25 frames. ], batch size: 102, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:14:31,314 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-01 23:14:49,017 INFO [optim.py:368] (0/8) 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 23:14:59,063 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245747.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:15:08,433 INFO [train.py:904] (0/8) Epoch 25, batch 2150, loss[loss=0.1773, simple_loss=0.2728, pruned_loss=0.04087, over 17061.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2546, pruned_loss=0.0397, over 3313576.12 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:00,596 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245792.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:16:15,996 INFO [train.py:904] (0/8) Epoch 25, batch 2200, loss[loss=0.1474, simple_loss=0.2411, pruned_loss=0.02682, over 17219.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2552, pruned_loss=0.04034, over 3308419.97 frames. ], batch size: 46, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:16:34,729 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245816.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:17:06,940 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.642e+02 2.208e+02 2.555e+02 2.978e+02 8.197e+02, threshold=5.110e+02, percent-clipped=1.0 2023-05-01 23:17:09,663 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5188, 4.4772, 4.4242, 4.1397, 4.1895, 4.4950, 4.2156, 4.2518], device='cuda:0'), covar=tensor([0.0725, 0.0818, 0.0331, 0.0330, 0.0790, 0.0551, 0.0589, 0.0652], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0469, 0.0364, 0.0368, 0.0369, 0.0423, 0.0249, 0.0441], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 23:17:23,667 INFO [train.py:904] (0/8) Epoch 25, batch 2250, loss[loss=0.1616, simple_loss=0.2414, pruned_loss=0.04086, over 15854.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2557, pruned_loss=0.03987, over 3324208.09 frames. ], batch size: 35, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:17:24,161 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245853.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:17:35,275 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5474, 3.6187, 3.3726, 2.9672, 3.2202, 3.4991, 3.3082, 3.3539], device='cuda:0'), covar=tensor([0.0603, 0.0689, 0.0305, 0.0327, 0.0503, 0.0500, 0.1373, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0315, 0.0470, 0.0365, 0.0369, 0.0369, 0.0424, 0.0250, 0.0443], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-01 23:17:36,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9726, 4.9743, 4.7674, 4.0720, 4.8623, 1.8998, 4.5893, 4.5301], device='cuda:0'), covar=tensor([0.0118, 0.0097, 0.0270, 0.0473, 0.0123, 0.2929, 0.0170, 0.0297], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0170, 0.0210, 0.0184, 0.0187, 0.0216, 0.0200, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:18:14,292 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0089, 2.1427, 2.5795, 2.8628, 2.8677, 2.9527, 2.1844, 3.2041], device='cuda:0'), covar=tensor([0.0223, 0.0486, 0.0353, 0.0300, 0.0317, 0.0306, 0.0573, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0198, 0.0185, 0.0190, 0.0205, 0.0162, 0.0201, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:18:16,582 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=245891.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:18:32,163 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=245902.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:18:32,956 INFO [train.py:904] (0/8) Epoch 25, batch 2300, loss[loss=0.173, simple_loss=0.2524, pruned_loss=0.04678, over 16852.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2554, pruned_loss=0.04032, over 3318536.77 frames. ], batch size: 102, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:18:49,968 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5331, 3.9837, 4.2490, 2.0489, 4.4325, 4.6485, 3.4153, 3.2411], device='cuda:0'), covar=tensor([0.1185, 0.0235, 0.0251, 0.1437, 0.0119, 0.0191, 0.0435, 0.0586], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0101, 0.0140, 0.0084, 0.0131, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:19:24,140 INFO [optim.py:368] (0/8) 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 23:19:42,203 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=245952.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:19:42,922 INFO [train.py:904] (0/8) Epoch 25, batch 2350, loss[loss=0.1672, simple_loss=0.2473, pruned_loss=0.04349, over 16816.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2553, pruned_loss=0.04102, over 3316396.07 frames. ], batch size: 83, lr: 2.71e-03, grad_scale: 2.0 2023-05-01 23:19:45,460 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4941, 4.5253, 4.8431, 4.8267, 4.8914, 4.5628, 4.5673, 4.4809], device='cuda:0'), covar=tensor([0.0459, 0.0952, 0.0517, 0.0513, 0.0596, 0.0571, 0.1094, 0.0699], device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0483, 0.0472, 0.0433, 0.0517, 0.0497, 0.0575, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 23:19:45,608 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0190, 2.9554, 2.6905, 4.4185, 3.5767, 4.1825, 1.7535, 3.0928], device='cuda:0'), covar=tensor([0.1270, 0.0667, 0.1104, 0.0164, 0.0198, 0.0392, 0.1571, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0178, 0.0197, 0.0197, 0.0204, 0.0217, 0.0206, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:20:33,926 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8361, 3.5837, 4.0302, 2.1898, 4.1062, 4.1349, 3.2384, 3.0924], device='cuda:0'), covar=tensor([0.0728, 0.0269, 0.0205, 0.1150, 0.0100, 0.0228, 0.0423, 0.0471], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0101, 0.0141, 0.0084, 0.0131, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:20:47,516 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-246000.pt 2023-05-01 23:20:54,782 INFO [train.py:904] (0/8) Epoch 25, batch 2400, loss[loss=0.1771, simple_loss=0.265, pruned_loss=0.0446, over 16718.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2559, pruned_loss=0.04145, over 3327082.78 frames. ], batch size: 134, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:21:18,009 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.90 vs. limit=2.0 2023-05-01 23:21:46,454 INFO [optim.py:368] (0/8) 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 23:21:49,231 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246042.0, num_to_drop=1, layers_to_drop={3} 2023-05-01 23:22:04,195 INFO [train.py:904] (0/8) Epoch 25, batch 2450, loss[loss=0.1608, simple_loss=0.2481, pruned_loss=0.03679, over 16991.00 frames. ], tot_loss[loss=0.169, simple_loss=0.256, pruned_loss=0.041, over 3331480.81 frames. ], batch size: 41, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:10,971 INFO [train.py:904] (0/8) Epoch 25, batch 2500, loss[loss=0.1741, simple_loss=0.2694, pruned_loss=0.03945, over 17053.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2564, pruned_loss=0.04018, over 3334854.97 frames. ], batch size: 53, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:23:21,599 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246111.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:27,344 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-01 23:23:28,139 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246116.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:23:59,035 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4684, 1.6434, 2.2064, 2.3782, 2.4580, 2.4889, 1.7177, 2.5728], device='cuda:0'), covar=tensor([0.0207, 0.0622, 0.0319, 0.0306, 0.0328, 0.0349, 0.0647, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0198, 0.0185, 0.0190, 0.0205, 0.0163, 0.0202, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:24:01,588 INFO [optim.py:368] (0/8) 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 23:24:12,439 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246148.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:18,464 INFO [train.py:904] (0/8) Epoch 25, batch 2550, loss[loss=0.1688, simple_loss=0.2494, pruned_loss=0.04408, over 16914.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2563, pruned_loss=0.03991, over 3342251.49 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:24:24,355 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1004, 3.8601, 4.3916, 2.1961, 4.5559, 4.6621, 3.3472, 3.5875], device='cuda:0'), covar=tensor([0.0720, 0.0282, 0.0236, 0.1212, 0.0079, 0.0170, 0.0453, 0.0423], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0112, 0.0101, 0.0141, 0.0084, 0.0132, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:24:28,667 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6043, 1.7351, 2.2297, 2.5181, 2.5293, 2.5602, 1.9173, 2.7418], device='cuda:0'), covar=tensor([0.0217, 0.0542, 0.0351, 0.0279, 0.0326, 0.0346, 0.0575, 0.0210], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0197, 0.0185, 0.0189, 0.0204, 0.0163, 0.0201, 0.0162], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:24:35,190 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246164.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:24:44,815 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246172.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:25:06,235 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8761, 3.0875, 2.8544, 5.0589, 4.1281, 4.4964, 1.7920, 3.3152], device='cuda:0'), covar=tensor([0.1391, 0.0758, 0.1173, 0.0267, 0.0254, 0.0420, 0.1637, 0.0781], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0196, 0.0204, 0.0217, 0.0205, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:25:17,114 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246196.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:25:26,254 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246202.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:25:26,993 INFO [train.py:904] (0/8) Epoch 25, batch 2600, loss[loss=0.174, simple_loss=0.2713, pruned_loss=0.03837, over 17031.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2562, pruned_loss=0.03978, over 3342185.55 frames. ], batch size: 50, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:25:52,272 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 23:26:28,326 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246247.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:33,159 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246250.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:26:37,205 INFO [train.py:904] (0/8) Epoch 25, batch 2650, loss[loss=0.1707, simple_loss=0.2645, pruned_loss=0.03842, over 16521.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2563, pruned_loss=0.03936, over 3347388.62 frames. ], batch size: 75, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:26:42,364 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246257.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:27:17,003 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246282.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:27:44,617 INFO [train.py:904] (0/8) Epoch 25, batch 2700, loss[loss=0.1847, simple_loss=0.283, pruned_loss=0.04319, over 17005.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2569, pruned_loss=0.03918, over 3354103.04 frames. ], batch size: 55, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:28:34,094 INFO [optim.py:368] (0/8) 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 23:28:36,991 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246342.0, num_to_drop=1, layers_to_drop={2} 2023-05-01 23:28:52,096 INFO [train.py:904] (0/8) Epoch 25, batch 2750, loss[loss=0.176, simple_loss=0.2592, pruned_loss=0.04638, over 16915.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2577, pruned_loss=0.03905, over 3352648.98 frames. ], batch size: 109, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:29:28,484 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-01 23:29:43,590 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246390.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:29:45,502 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8882, 3.6171, 4.1358, 2.0171, 4.2768, 4.3246, 3.0803, 3.2880], device='cuda:0'), covar=tensor([0.0749, 0.0297, 0.0229, 0.1276, 0.0098, 0.0242, 0.0515, 0.0455], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0102, 0.0142, 0.0085, 0.0133, 0.0132, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:30:01,774 INFO [train.py:904] (0/8) Epoch 25, batch 2800, loss[loss=0.1645, simple_loss=0.2641, pruned_loss=0.03244, over 17112.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2576, pruned_loss=0.03881, over 3354016.74 frames. ], batch size: 48, lr: 2.71e-03, grad_scale: 8.0 2023-05-01 23:30:27,743 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8373, 4.8945, 5.2672, 5.2769, 5.2895, 4.9804, 4.9222, 4.7506], device='cuda:0'), covar=tensor([0.0372, 0.0543, 0.0427, 0.0392, 0.0521, 0.0423, 0.0997, 0.0480], device='cuda:0'), in_proj_covar=tensor([0.0432, 0.0484, 0.0472, 0.0433, 0.0517, 0.0497, 0.0575, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 23:30:34,358 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7468, 2.8358, 2.7468, 4.8960, 3.6407, 4.3160, 1.5751, 3.1158], device='cuda:0'), covar=tensor([0.1494, 0.0875, 0.1227, 0.0172, 0.0192, 0.0364, 0.1821, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0179, 0.0198, 0.0198, 0.0206, 0.0219, 0.0206, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:30:54,341 INFO [optim.py:368] (0/8) 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 23:31:04,187 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246448.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:31:10,740 INFO [train.py:904] (0/8) Epoch 25, batch 2850, loss[loss=0.1723, simple_loss=0.2612, pruned_loss=0.04172, over 16675.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2572, pruned_loss=0.03912, over 3346411.64 frames. ], batch size: 83, lr: 2.71e-03, grad_scale: 4.0 2023-05-01 23:31:31,884 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246467.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:10,940 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246496.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:32:20,912 INFO [train.py:904] (0/8) Epoch 25, batch 2900, loss[loss=0.1458, simple_loss=0.2474, pruned_loss=0.0221, over 17122.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2561, pruned_loss=0.03966, over 3333256.58 frames. ], batch size: 48, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:32:31,070 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0674, 5.0409, 5.5317, 5.5212, 5.5400, 5.1676, 5.1318, 4.9902], device='cuda:0'), covar=tensor([0.0379, 0.0629, 0.0362, 0.0401, 0.0489, 0.0447, 0.1033, 0.0437], device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0487, 0.0474, 0.0435, 0.0518, 0.0499, 0.0579, 0.0394], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-01 23:33:11,806 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6384, 5.9830, 5.7792, 5.8091, 5.4094, 5.4509, 5.4467, 6.0888], device='cuda:0'), covar=tensor([0.1391, 0.1002, 0.1014, 0.0908, 0.0902, 0.0700, 0.1224, 0.0920], device='cuda:0'), in_proj_covar=tensor([0.0722, 0.0874, 0.0714, 0.0676, 0.0554, 0.0555, 0.0735, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:33:14,959 INFO [optim.py:368] (0/8) 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 23:33:25,244 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246547.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:33:32,620 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246552.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:33:33,513 INFO [train.py:904] (0/8) Epoch 25, batch 2950, loss[loss=0.1812, simple_loss=0.2742, pruned_loss=0.04408, over 16743.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2557, pruned_loss=0.04013, over 3325966.08 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:06,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=246577.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:32,278 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246595.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:34:42,716 INFO [train.py:904] (0/8) Epoch 25, batch 3000, loss[loss=0.1587, simple_loss=0.2582, pruned_loss=0.02967, over 17193.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2556, pruned_loss=0.04038, over 3325270.27 frames. ], batch size: 46, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:34:42,717 INFO [train.py:929] (0/8) Computing validation loss 2023-05-01 23:34:52,580 INFO [train.py:938] (0/8) Epoch 25, validation: loss=0.1341, simple_loss=0.239, pruned_loss=0.01457, over 944034.00 frames. 2023-05-01 23:34:52,581 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-01 23:34:58,595 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1683, 3.8961, 4.3659, 2.3464, 4.5773, 4.6317, 3.3869, 3.6998], device='cuda:0'), covar=tensor([0.0723, 0.0292, 0.0243, 0.1167, 0.0084, 0.0199, 0.0452, 0.0370], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0102, 0.0141, 0.0085, 0.0132, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:35:46,639 INFO [optim.py:368] (0/8) 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 23:36:05,627 INFO [train.py:904] (0/8) Epoch 25, batch 3050, loss[loss=0.1439, simple_loss=0.2373, pruned_loss=0.02519, over 17190.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2555, pruned_loss=0.04023, over 3330940.37 frames. ], batch size: 46, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:36:38,417 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246677.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:37:13,560 INFO [train.py:904] (0/8) Epoch 25, batch 3100, loss[loss=0.1632, simple_loss=0.2585, pruned_loss=0.03396, over 16711.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2547, pruned_loss=0.04026, over 3336349.38 frames. ], batch size: 62, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:37:41,132 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.50 vs. limit=2.0 2023-05-01 23:38:04,935 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5901, 4.5554, 4.4962, 3.8526, 4.5662, 1.7248, 4.2752, 4.1979], device='cuda:0'), covar=tensor([0.0152, 0.0114, 0.0219, 0.0448, 0.0121, 0.2979, 0.0190, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0172, 0.0211, 0.0186, 0.0189, 0.0217, 0.0201, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:38:05,007 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246738.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:38:08,782 INFO [optim.py:368] (0/8) 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 23:38:25,024 INFO [train.py:904] (0/8) Epoch 25, batch 3150, loss[loss=0.1572, simple_loss=0.2407, pruned_loss=0.03686, over 16872.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2534, pruned_loss=0.03995, over 3341300.73 frames. ], batch size: 90, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:38:44,312 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246767.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:39:34,228 INFO [train.py:904] (0/8) Epoch 25, batch 3200, loss[loss=0.1489, simple_loss=0.2368, pruned_loss=0.03051, over 15933.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2532, pruned_loss=0.03951, over 3339138.53 frames. ], batch size: 35, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:39:51,044 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246815.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:40:27,531 INFO [optim.py:368] (0/8) 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 23:40:42,141 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246852.0, num_to_drop=1, layers_to_drop={1} 2023-05-01 23:40:42,946 INFO [train.py:904] (0/8) Epoch 25, batch 3250, loss[loss=0.1364, simple_loss=0.2262, pruned_loss=0.02327, over 17012.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2528, pruned_loss=0.03877, over 3341604.24 frames. ], batch size: 41, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:41:16,221 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=246877.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:41:49,127 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 25, batch 3300, loss[loss=0.1881, simple_loss=0.2682, pruned_loss=0.05397, over 12249.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2542, pruned_loss=0.0394, over 3339094.75 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:42:24,638 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=246925.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:36,848 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=246934.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:42:46,637 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 25, batch 3350, loss[loss=0.185, simple_loss=0.263, pruned_loss=0.05347, over 16226.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2545, pruned_loss=0.03952, over 3334891.09 frames. ], batch size: 165, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:43:42,474 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9491, 3.1489, 2.9917, 5.1168, 4.2371, 4.5047, 1.8586, 3.2498], device='cuda:0'), covar=tensor([0.1306, 0.0704, 0.1078, 0.0199, 0.0218, 0.0407, 0.1602, 0.0747], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0197, 0.0198, 0.0205, 0.0218, 0.0206, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-01 23:44:01,665 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=246995.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:44:12,424 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9573, 2.6943, 2.7654, 2.1563, 2.6027, 2.1064, 2.7646, 2.8941], device='cuda:0'), covar=tensor([0.0312, 0.0842, 0.0600, 0.1775, 0.0894, 0.1004, 0.0631, 0.0842], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0169, 0.0171, 0.0156, 0.0148, 0.0132, 0.0147, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-01 23:44:13,558 INFO [train.py:904] (0/8) Epoch 25, batch 3400, loss[loss=0.2292, simple_loss=0.3015, pruned_loss=0.07845, over 12177.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2546, pruned_loss=0.03943, over 3334311.49 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:44:39,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4163, 3.6548, 3.9939, 2.2256, 3.3591, 2.5588, 3.9090, 3.8542], device='cuda:0'), covar=tensor([0.0271, 0.0937, 0.0481, 0.2080, 0.0757, 0.1009, 0.0632, 0.1080], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0169, 0.0171, 0.0156, 0.0147, 0.0132, 0.0146, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-01 23:44:48,714 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-01 23:44:56,892 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247033.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:45:08,922 INFO [optim.py:368] (0/8) 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 23:45:12,407 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9623, 4.6613, 5.0099, 5.1680, 5.3776, 4.6688, 5.3607, 5.3444], device='cuda:0'), covar=tensor([0.1814, 0.1376, 0.1835, 0.0830, 0.0519, 0.1074, 0.0516, 0.0650], device='cuda:0'), in_proj_covar=tensor([0.0692, 0.0856, 0.0987, 0.0864, 0.0661, 0.0683, 0.0707, 0.0829], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:45:16,783 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-01 23:45:26,159 INFO [train.py:904] (0/8) Epoch 25, batch 3450, loss[loss=0.1927, simple_loss=0.27, pruned_loss=0.05771, over 16560.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2535, pruned_loss=0.03927, over 3328053.74 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:46:35,602 INFO [train.py:904] (0/8) Epoch 25, batch 3500, loss[loss=0.1622, simple_loss=0.2442, pruned_loss=0.04006, over 16727.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2525, pruned_loss=0.03895, over 3323436.64 frames. ], batch size: 134, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:47:30,024 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.124e+02 2.460e+02 2.748e+02 9.578e+02, threshold=4.920e+02, percent-clipped=1.0 2023-05-01 23:47:45,787 INFO [train.py:904] (0/8) Epoch 25, batch 3550, loss[loss=0.1499, simple_loss=0.2371, pruned_loss=0.03137, over 16855.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2507, pruned_loss=0.03837, over 3325211.57 frames. ], batch size: 42, lr: 2.70e-03, grad_scale: 4.0 2023-05-01 23:48:52,366 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-01 23:48:55,177 INFO [train.py:904] (0/8) Epoch 25, batch 3600, loss[loss=0.1511, simple_loss=0.2383, pruned_loss=0.03193, over 17231.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2508, pruned_loss=0.03869, over 3319070.51 frames. ], batch size: 45, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:49:07,553 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-01 23:49:12,277 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-05-01 23:49:49,108 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.664e+02 2.145e+02 2.569e+02 3.333e+02 7.279e+02, threshold=5.139e+02, percent-clipped=3.0 2023-05-01 23:49:54,607 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0674, 3.3564, 3.2984, 2.2639, 2.9616, 2.5893, 3.6146, 3.6762], device='cuda:0'), covar=tensor([0.0286, 0.0865, 0.0683, 0.1902, 0.0949, 0.0971, 0.0584, 0.0828], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0170, 0.0171, 0.0156, 0.0148, 0.0132, 0.0147, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-01 23:50:05,485 INFO [train.py:904] (0/8) Epoch 25, batch 3650, loss[loss=0.1488, simple_loss=0.2303, pruned_loss=0.03367, over 15929.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2491, pruned_loss=0.03896, over 3309768.76 frames. ], batch size: 35, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:50:58,845 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247290.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:51:17,367 INFO [train.py:904] (0/8) Epoch 25, batch 3700, loss[loss=0.181, simple_loss=0.2555, pruned_loss=0.05325, over 16770.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2484, pruned_loss=0.04031, over 3290649.63 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:52:00,701 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247333.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:52:12,952 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.806e+02 2.292e+02 2.615e+02 3.085e+02 5.082e+02, threshold=5.229e+02, percent-clipped=0.0 2023-05-01 23:52:30,283 INFO [train.py:904] (0/8) Epoch 25, batch 3750, loss[loss=0.1632, simple_loss=0.248, pruned_loss=0.03924, over 16245.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2497, pruned_loss=0.04167, over 3290243.55 frames. ], batch size: 165, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:53:06,861 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247379.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:10,242 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247381.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:53:42,911 INFO [train.py:904] (0/8) Epoch 25, batch 3800, loss[loss=0.2017, simple_loss=0.284, pruned_loss=0.05971, over 12111.00 frames. ], tot_loss[loss=0.1681, simple_loss=0.2504, pruned_loss=0.04289, over 3290783.21 frames. ], batch size: 248, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:54:06,930 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3243, 2.5275, 2.5188, 4.1645, 2.3363, 2.7830, 2.5461, 2.7061], device='cuda:0'), covar=tensor([0.1323, 0.3301, 0.2711, 0.0491, 0.3888, 0.2289, 0.3162, 0.2907], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0465, 0.0383, 0.0337, 0.0444, 0.0533, 0.0438, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:54:15,320 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247426.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:35,832 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247440.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:54:38,774 INFO [optim.py:368] (0/8) 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 23:54:54,730 INFO [train.py:904] (0/8) Epoch 25, batch 3850, loss[loss=0.188, simple_loss=0.2669, pruned_loss=0.05456, over 16375.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2508, pruned_loss=0.04365, over 3287869.88 frames. ], batch size: 146, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:55:39,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3633, 5.6859, 5.4244, 5.5338, 5.1641, 5.0649, 5.1116, 5.8045], device='cuda:0'), covar=tensor([0.1298, 0.0827, 0.1030, 0.0824, 0.0829, 0.0770, 0.1129, 0.0825], device='cuda:0'), in_proj_covar=tensor([0.0724, 0.0873, 0.0719, 0.0678, 0.0555, 0.0557, 0.0738, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:55:43,354 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247487.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:02,867 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247500.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:56:06,083 INFO [train.py:904] (0/8) Epoch 25, batch 3900, loss[loss=0.1589, simple_loss=0.2348, pruned_loss=0.04151, over 16770.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.251, pruned_loss=0.04445, over 3279539.80 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:56:18,573 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4655, 3.4455, 3.7543, 2.1031, 3.1898, 2.4959, 3.9016, 3.9190], device='cuda:0'), covar=tensor([0.0197, 0.0868, 0.0567, 0.2162, 0.0881, 0.0956, 0.0478, 0.0776], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0168, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-01 23:56:39,110 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247525.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:03,109 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.630e+02 2.139e+02 2.490e+02 2.933e+02 5.052e+02, threshold=4.980e+02, percent-clipped=0.0 2023-05-01 23:57:17,844 INFO [train.py:904] (0/8) Epoch 25, batch 3950, loss[loss=0.1666, simple_loss=0.2443, pruned_loss=0.04449, over 16424.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2498, pruned_loss=0.04449, over 3288086.51 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:57:20,040 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247554.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:30,897 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247561.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:57:43,132 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8963, 5.2223, 4.9502, 4.9651, 4.7460, 4.6824, 4.6084, 5.3314], device='cuda:0'), covar=tensor([0.1205, 0.0846, 0.1049, 0.0899, 0.0885, 0.1098, 0.1182, 0.0866], device='cuda:0'), in_proj_covar=tensor([0.0726, 0.0875, 0.0721, 0.0679, 0.0557, 0.0559, 0.0741, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:58:05,310 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247586.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:12,111 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=247590.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:58:30,199 INFO [train.py:904] (0/8) Epoch 25, batch 4000, loss[loss=0.1977, simple_loss=0.2656, pruned_loss=0.06492, over 16767.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2501, pruned_loss=0.04501, over 3291038.69 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 8.0 2023-05-01 23:58:49,593 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247615.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:03,872 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2341, 2.3532, 2.3661, 3.9287, 2.2829, 2.7184, 2.3867, 2.5095], device='cuda:0'), covar=tensor([0.1386, 0.3416, 0.2886, 0.0604, 0.3850, 0.2329, 0.3547, 0.2978], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0467, 0.0384, 0.0338, 0.0445, 0.0535, 0.0439, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-01 23:59:06,797 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-01 23:59:21,868 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=247638.0, num_to_drop=0, layers_to_drop=set() 2023-05-01 23:59:27,472 INFO [optim.py:368] (0/8) 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 23:59:44,170 INFO [train.py:904] (0/8) Epoch 25, batch 4050, loss[loss=0.15, simple_loss=0.2416, pruned_loss=0.0292, over 16831.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2507, pruned_loss=0.04431, over 3284810.59 frames. ], batch size: 83, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:00:59,194 INFO [train.py:904] (0/8) Epoch 25, batch 4100, loss[loss=0.1876, simple_loss=0.2745, pruned_loss=0.05035, over 16509.00 frames. ], tot_loss[loss=0.1705, simple_loss=0.2527, pruned_loss=0.04414, over 3277383.11 frames. ], batch size: 75, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:01:40,437 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2694, 2.5176, 2.5964, 4.0885, 2.4329, 2.7947, 2.5227, 2.6618], device='cuda:0'), covar=tensor([0.1360, 0.3111, 0.2614, 0.0527, 0.3551, 0.2294, 0.3178, 0.3031], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0466, 0.0383, 0.0337, 0.0444, 0.0535, 0.0438, 0.0546], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:01:48,431 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247735.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:01:57,931 INFO [optim.py:368] (0/8) 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 00:02:14,952 INFO [train.py:904] (0/8) Epoch 25, batch 4150, loss[loss=0.2124, simple_loss=0.2905, pruned_loss=0.0671, over 11140.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2594, pruned_loss=0.04636, over 3232148.08 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:02:45,961 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0578, 4.0459, 3.9327, 3.0193, 3.9230, 1.7929, 3.7376, 3.2898], device='cuda:0'), covar=tensor([0.0120, 0.0122, 0.0207, 0.0302, 0.0096, 0.3205, 0.0128, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0169, 0.0210, 0.0186, 0.0188, 0.0216, 0.0200, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:03:01,204 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:04,951 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247784.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:03:32,736 INFO [train.py:904] (0/8) Epoch 25, batch 4200, loss[loss=0.2372, simple_loss=0.3107, pruned_loss=0.08179, over 11309.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.2658, pruned_loss=0.04746, over 3201721.38 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:22,935 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8693, 2.1790, 2.3593, 3.1015, 2.1741, 2.3426, 2.3024, 2.3174], device='cuda:0'), covar=tensor([0.1402, 0.3390, 0.2579, 0.0739, 0.4230, 0.2401, 0.3363, 0.3184], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0463, 0.0380, 0.0334, 0.0441, 0.0531, 0.0435, 0.0542], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:04:30,301 INFO [optim.py:368] (0/8) 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 00:04:35,268 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=247845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:04:46,256 INFO [train.py:904] (0/8) Epoch 25, batch 4250, loss[loss=0.1652, simple_loss=0.258, pruned_loss=0.03621, over 16765.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2693, pruned_loss=0.04714, over 3198437.95 frames. ], batch size: 124, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:04:51,428 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247856.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:05:28,749 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:06:02,544 INFO [train.py:904] (0/8) Epoch 25, batch 4300, loss[loss=0.2115, simple_loss=0.2977, pruned_loss=0.06268, over 15362.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2709, pruned_loss=0.04653, over 3193196.81 frames. ], batch size: 191, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:06:13,456 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=247910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:06:23,454 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6102, 2.5488, 1.9538, 2.7175, 2.0458, 2.7836, 2.1552, 2.3193], device='cuda:0'), covar=tensor([0.0269, 0.0329, 0.1217, 0.0204, 0.0639, 0.0384, 0.1185, 0.0627], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0171, 0.0179, 0.0222, 0.0204, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:06:23,809 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 00:06:57,269 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4269, 4.6931, 4.4528, 4.5035, 4.2514, 4.1826, 4.1221, 4.7377], device='cuda:0'), covar=tensor([0.1104, 0.0767, 0.1055, 0.0849, 0.0776, 0.1500, 0.1094, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0713, 0.0860, 0.0708, 0.0666, 0.0546, 0.0549, 0.0725, 0.0674], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:07:00,836 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3522, 3.2085, 2.5466, 2.1578, 2.2348, 2.2451, 3.3634, 2.9700], device='cuda:0'), covar=tensor([0.3090, 0.0822, 0.1988, 0.2531, 0.2562, 0.2169, 0.0540, 0.1334], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0273, 0.0310, 0.0319, 0.0303, 0.0269, 0.0301, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:07:01,393 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.557e+02 2.163e+02 2.477e+02 3.039e+02 5.024e+02, threshold=4.955e+02, percent-clipped=0.0 2023-05-02 00:07:17,423 INFO [train.py:904] (0/8) Epoch 25, batch 4350, loss[loss=0.1795, simple_loss=0.2722, pruned_loss=0.04335, over 17218.00 frames. ], tot_loss[loss=0.1847, simple_loss=0.2741, pruned_loss=0.04766, over 3182240.84 frames. ], batch size: 45, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:08:05,203 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247984.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:23,238 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=247996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:08:29,324 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-248000.pt 2023-05-02 00:08:36,581 INFO [train.py:904] (0/8) Epoch 25, batch 4400, loss[loss=0.2139, simple_loss=0.2977, pruned_loss=0.06499, over 16771.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2761, pruned_loss=0.04876, over 3175810.46 frames. ], batch size: 39, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:04,022 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0978, 4.8912, 5.1242, 5.2723, 5.4381, 4.8299, 5.4561, 5.4665], device='cuda:0'), covar=tensor([0.1477, 0.1119, 0.1328, 0.0611, 0.0402, 0.0750, 0.0469, 0.0505], device='cuda:0'), in_proj_covar=tensor([0.0667, 0.0821, 0.0945, 0.0829, 0.0637, 0.0653, 0.0681, 0.0796], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:09:24,858 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:29,258 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4200, 4.0600, 3.9407, 2.6918, 3.6418, 4.0673, 3.5933, 2.3421], device='cuda:0'), covar=tensor([0.0546, 0.0035, 0.0058, 0.0399, 0.0091, 0.0099, 0.0090, 0.0470], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0134, 0.0100, 0.0111, 0.0097, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 00:09:34,456 INFO [optim.py:368] (0/8) 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 00:09:39,350 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248045.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:09:50,304 INFO [train.py:904] (0/8) Epoch 25, batch 4450, loss[loss=0.2129, simple_loss=0.3049, pruned_loss=0.06047, over 16514.00 frames. ], tot_loss[loss=0.1902, simple_loss=0.2797, pruned_loss=0.05031, over 3178664.07 frames. ], batch size: 75, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:09:56,548 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,173 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:33,248 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:10:34,259 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248083.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:04,518 INFO [train.py:904] (0/8) Epoch 25, batch 4500, loss[loss=0.1889, simple_loss=0.2769, pruned_loss=0.05047, over 17087.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2803, pruned_loss=0.05088, over 3195528.39 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:11:44,504 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:11:59,991 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:01,969 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.331e+02 1.873e+02 2.071e+02 2.399e+02 5.095e+02, threshold=4.142e+02, percent-clipped=1.0 2023-05-02 00:12:02,453 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6079, 2.5341, 2.3804, 3.4303, 2.5342, 3.6434, 1.5906, 2.7822], device='cuda:0'), covar=tensor([0.1381, 0.0783, 0.1244, 0.0163, 0.0183, 0.0335, 0.1666, 0.0823], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0196, 0.0205, 0.0216, 0.0206, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:12:03,683 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:12:17,521 INFO [train.py:904] (0/8) Epoch 25, batch 4550, loss[loss=0.1922, simple_loss=0.2805, pruned_loss=0.05193, over 17103.00 frames. ], tot_loss[loss=0.1922, simple_loss=0.2808, pruned_loss=0.05185, over 3189980.78 frames. ], batch size: 49, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:12:22,906 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:00,080 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248181.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:32,565 INFO [train.py:904] (0/8) Epoch 25, batch 4600, loss[loss=0.2288, simple_loss=0.2991, pruned_loss=0.07928, over 11840.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2818, pruned_loss=0.05212, over 3205803.87 frames. ], batch size: 246, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:13:34,236 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248204.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:13:42,889 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:14:01,176 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 00:14:11,283 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 1.740e+02 1.903e+02 2.276e+02 6.899e+02, threshold=3.806e+02, percent-clipped=0.0 2023-05-02 00:14:46,977 INFO [train.py:904] (0/8) Epoch 25, batch 4650, loss[loss=0.1866, simple_loss=0.2719, pruned_loss=0.0506, over 16539.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2815, pruned_loss=0.05256, over 3217303.56 frames. ], batch size: 68, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:14:54,079 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:15:07,352 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8736, 2.6535, 2.8574, 2.0583, 2.7001, 2.0852, 2.7233, 2.7858], device='cuda:0'), covar=tensor([0.0262, 0.0818, 0.0482, 0.1891, 0.0784, 0.0973, 0.0604, 0.0756], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0168, 0.0170, 0.0155, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 00:16:01,797 INFO [train.py:904] (0/8) Epoch 25, batch 4700, loss[loss=0.1666, simple_loss=0.255, pruned_loss=0.03908, over 16352.00 frames. ], tot_loss[loss=0.1907, simple_loss=0.2785, pruned_loss=0.05148, over 3202957.22 frames. ], batch size: 35, lr: 2.70e-03, grad_scale: 8.0 2023-05-02 00:16:05,172 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3896, 4.5454, 4.6424, 4.4672, 4.5228, 5.0341, 4.5673, 4.2558], device='cuda:0'), covar=tensor([0.1461, 0.1707, 0.2110, 0.2012, 0.2428, 0.0966, 0.1474, 0.2521], device='cuda:0'), in_proj_covar=tensor([0.0424, 0.0624, 0.0680, 0.0507, 0.0674, 0.0711, 0.0529, 0.0678], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:16:42,007 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 00:16:43,649 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:56,037 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:16:57,978 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 1.897e+02 2.230e+02 2.532e+02 4.236e+02, threshold=4.460e+02, percent-clipped=2.0 2023-05-02 00:16:59,697 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1994, 4.0771, 4.2569, 4.4117, 4.5193, 4.1667, 4.4513, 4.5611], device='cuda:0'), covar=tensor([0.1655, 0.1241, 0.1544, 0.0719, 0.0577, 0.1211, 0.0884, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0662, 0.0816, 0.0938, 0.0826, 0.0632, 0.0650, 0.0676, 0.0792], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:17:13,872 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:17:14,742 INFO [train.py:904] (0/8) Epoch 25, batch 4750, loss[loss=0.1643, simple_loss=0.2539, pruned_loss=0.03731, over 16280.00 frames. ], tot_loss[loss=0.1864, simple_loss=0.2745, pruned_loss=0.04918, over 3211699.39 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:17:59,847 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:13,235 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:18:29,682 INFO [train.py:904] (0/8) Epoch 25, batch 4800, loss[loss=0.1521, simple_loss=0.2402, pruned_loss=0.03196, over 17243.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2702, pruned_loss=0.04675, over 3214792.20 frames. ], batch size: 52, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:18:41,960 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0301, 2.1767, 2.5822, 2.9618, 2.8624, 3.4937, 2.2507, 3.4051], device='cuda:0'), covar=tensor([0.0212, 0.0493, 0.0348, 0.0359, 0.0358, 0.0192, 0.0552, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0188, 0.0203, 0.0161, 0.0199, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:19:21,484 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:24,818 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248440.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:27,485 INFO [optim.py:368] (0/8) 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 00:19:31,498 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:19:43,856 INFO [train.py:904] (0/8) Epoch 25, batch 4850, loss[loss=0.1795, simple_loss=0.2739, pruned_loss=0.04254, over 16506.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2711, pruned_loss=0.0461, over 3201931.54 frames. ], batch size: 68, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:20:24,652 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-02 00:20:36,776 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248488.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:20:50,375 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 00:20:58,432 INFO [train.py:904] (0/8) Epoch 25, batch 4900, loss[loss=0.1696, simple_loss=0.2673, pruned_loss=0.03596, over 16847.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2695, pruned_loss=0.04437, over 3214772.92 frames. ], batch size: 96, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:21:08,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8237, 1.3929, 1.6938, 1.7829, 1.8924, 1.9546, 1.6733, 1.8294], device='cuda:0'), covar=tensor([0.0273, 0.0447, 0.0234, 0.0331, 0.0294, 0.0195, 0.0443, 0.0151], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0187, 0.0202, 0.0160, 0.0199, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:22:03,445 INFO [optim.py:368] (0/8) 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 00:22:20,192 INFO [train.py:904] (0/8) Epoch 25, batch 4950, loss[loss=0.1741, simple_loss=0.2653, pruned_loss=0.04142, over 16157.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2689, pruned_loss=0.04373, over 3211642.24 frames. ], batch size: 35, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:06,770 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6587, 3.4747, 4.1144, 1.9582, 4.2449, 4.2843, 3.1495, 3.1483], device='cuda:0'), covar=tensor([0.0829, 0.0308, 0.0172, 0.1301, 0.0064, 0.0120, 0.0419, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0101, 0.0139, 0.0084, 0.0130, 0.0130, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:23:31,264 INFO [train.py:904] (0/8) Epoch 25, batch 5000, loss[loss=0.1729, simple_loss=0.2706, pruned_loss=0.03759, over 15378.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2712, pruned_loss=0.04401, over 3210143.20 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:23:51,580 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0642, 3.6902, 3.6372, 2.2741, 3.3430, 3.6908, 3.3278, 2.0920], device='cuda:0'), covar=tensor([0.0584, 0.0048, 0.0054, 0.0458, 0.0100, 0.0083, 0.0107, 0.0469], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0134, 0.0100, 0.0111, 0.0097, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 00:24:25,388 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248640.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:26,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5306, 2.5878, 2.1839, 2.4150, 2.9283, 2.5982, 2.9004, 3.1639], device='cuda:0'), covar=tensor([0.0099, 0.0452, 0.0577, 0.0466, 0.0285, 0.0393, 0.0201, 0.0272], device='cuda:0'), in_proj_covar=tensor([0.0222, 0.0242, 0.0230, 0.0233, 0.0244, 0.0242, 0.0243, 0.0241], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:24:27,924 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.446e+02 2.126e+02 2.416e+02 2.928e+02 6.487e+02, threshold=4.832e+02, percent-clipped=1.0 2023-05-02 00:24:34,857 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 00:24:42,919 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:24:43,841 INFO [train.py:904] (0/8) Epoch 25, batch 5050, loss[loss=0.1689, simple_loss=0.2659, pruned_loss=0.03597, over 16714.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2718, pruned_loss=0.04407, over 3214560.18 frames. ], batch size: 76, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:25:35,139 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:35,155 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248688.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:39,538 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:53,430 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:25:57,289 INFO [train.py:904] (0/8) Epoch 25, batch 5100, loss[loss=0.1829, simple_loss=0.2653, pruned_loss=0.0503, over 12248.00 frames. ], tot_loss[loss=0.179, simple_loss=0.2706, pruned_loss=0.04371, over 3213244.92 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:26:15,828 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=248715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:45,333 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0794, 2.2068, 2.5501, 3.0420, 2.9040, 3.5453, 2.2528, 3.4582], device='cuda:0'), covar=tensor([0.0230, 0.0510, 0.0409, 0.0359, 0.0312, 0.0150, 0.0582, 0.0126], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0194, 0.0182, 0.0187, 0.0201, 0.0159, 0.0198, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:26:50,007 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:52,088 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=248739.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:26:55,448 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8693, 4.3243, 3.1524, 2.5205, 2.9793, 2.7835, 4.6411, 3.7698], device='cuda:0'), covar=tensor([0.2686, 0.0534, 0.1768, 0.2588, 0.2387, 0.1788, 0.0350, 0.1136], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0272, 0.0310, 0.0319, 0.0302, 0.0269, 0.0301, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:26:56,589 INFO [optim.py:368] (0/8) 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 00:27:12,740 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:27:13,425 INFO [train.py:904] (0/8) Epoch 25, batch 5150, loss[loss=0.1833, simple_loss=0.2848, pruned_loss=0.04089, over 15455.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2702, pruned_loss=0.04305, over 3212448.67 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:27:46,964 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=248776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:27:51,260 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8479, 4.7338, 4.6147, 3.3000, 3.9098, 4.6467, 3.9137, 2.6017], device='cuda:0'), covar=tensor([0.0515, 0.0037, 0.0038, 0.0342, 0.0106, 0.0093, 0.0114, 0.0459], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0100, 0.0111, 0.0097, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 00:28:03,432 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=248786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:28:28,608 INFO [train.py:904] (0/8) Epoch 25, batch 5200, loss[loss=0.1742, simple_loss=0.2615, pruned_loss=0.04346, over 16865.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2693, pruned_loss=0.04276, over 3211407.99 frames. ], batch size: 42, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:29:14,787 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 00:29:25,197 INFO [optim.py:368] (0/8) 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 00:29:41,121 INFO [train.py:904] (0/8) Epoch 25, batch 5250, loss[loss=0.2008, simple_loss=0.2833, pruned_loss=0.05915, over 12189.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2672, pruned_loss=0.04244, over 3209394.39 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:30:23,989 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6567, 2.7172, 2.8419, 4.5276, 2.5418, 3.0808, 2.7577, 2.9792], device='cuda:0'), covar=tensor([0.1278, 0.3162, 0.2666, 0.0432, 0.3841, 0.2387, 0.3230, 0.2911], device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0464, 0.0380, 0.0335, 0.0443, 0.0532, 0.0436, 0.0542], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:30:43,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7080, 2.3895, 2.4006, 3.3673, 2.2436, 3.6984, 1.5940, 2.8968], device='cuda:0'), covar=tensor([0.1351, 0.0854, 0.1267, 0.0165, 0.0168, 0.0358, 0.1649, 0.0733], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0194, 0.0203, 0.0214, 0.0204, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:30:53,684 INFO [train.py:904] (0/8) Epoch 25, batch 5300, loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03893, over 16585.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2637, pruned_loss=0.04142, over 3207113.01 frames. ], batch size: 62, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:31:51,343 INFO [optim.py:368] (0/8) 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 00:32:03,706 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6979, 1.9125, 2.3144, 2.6977, 2.6006, 3.0831, 1.9883, 3.0142], device='cuda:0'), covar=tensor([0.0286, 0.0575, 0.0381, 0.0377, 0.0381, 0.0220, 0.0619, 0.0177], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0195, 0.0182, 0.0187, 0.0202, 0.0160, 0.0199, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:32:08,043 INFO [train.py:904] (0/8) Epoch 25, batch 5350, loss[loss=0.1725, simple_loss=0.2694, pruned_loss=0.0378, over 16778.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2629, pruned_loss=0.04099, over 3211218.95 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:32:21,420 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1959, 4.0876, 4.2727, 4.3994, 4.5527, 4.1530, 4.4963, 4.5539], device='cuda:0'), covar=tensor([0.1697, 0.1203, 0.1502, 0.0752, 0.0512, 0.1234, 0.0791, 0.0685], device='cuda:0'), in_proj_covar=tensor([0.0659, 0.0811, 0.0933, 0.0820, 0.0625, 0.0644, 0.0673, 0.0786], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:32:29,336 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6834, 2.5780, 2.5099, 3.8706, 2.6751, 3.8882, 1.5059, 2.9385], device='cuda:0'), covar=tensor([0.1342, 0.0776, 0.1165, 0.0149, 0.0180, 0.0351, 0.1654, 0.0749], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0177, 0.0195, 0.0193, 0.0203, 0.0214, 0.0204, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:33:00,485 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=248988.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:33:08,872 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7960, 2.2588, 1.7814, 1.9627, 2.5952, 2.2257, 2.4241, 2.7920], device='cuda:0'), covar=tensor([0.0209, 0.0514, 0.0722, 0.0577, 0.0321, 0.0474, 0.0268, 0.0323], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0240, 0.0228, 0.0230, 0.0241, 0.0240, 0.0239, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:33:17,539 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 00:33:22,820 INFO [train.py:904] (0/8) Epoch 25, batch 5400, loss[loss=0.2052, simple_loss=0.2912, pruned_loss=0.05958, over 12199.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2651, pruned_loss=0.04162, over 3191354.95 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:34:11,329 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:17,376 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249039.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:20,471 INFO [optim.py:368] (0/8) 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 00:34:29,110 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249047.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:34:40,342 INFO [train.py:904] (0/8) Epoch 25, batch 5450, loss[loss=0.1958, simple_loss=0.2845, pruned_loss=0.05349, over 16877.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.268, pruned_loss=0.04287, over 3183382.18 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:34:46,553 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4854, 3.4313, 2.7455, 2.2327, 2.3260, 2.3366, 3.5931, 3.1651], device='cuda:0'), covar=tensor([0.3003, 0.0605, 0.1765, 0.2745, 0.2620, 0.2203, 0.0488, 0.1289], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0273, 0.0310, 0.0320, 0.0302, 0.0269, 0.0302, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:34:52,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7812, 1.4454, 1.7136, 1.6976, 1.7646, 1.9497, 1.6084, 1.8186], device='cuda:0'), covar=tensor([0.0287, 0.0379, 0.0220, 0.0314, 0.0279, 0.0172, 0.0411, 0.0150], device='cuda:0'), in_proj_covar=tensor([0.0194, 0.0194, 0.0182, 0.0187, 0.0201, 0.0160, 0.0199, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:35:08,687 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:34,361 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249087.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:35:58,373 INFO [train.py:904] (0/8) Epoch 25, batch 5500, loss[loss=0.2027, simple_loss=0.2973, pruned_loss=0.05409, over 16683.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2744, pruned_loss=0.04639, over 3168455.76 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:36:46,403 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5951, 4.1657, 4.1645, 2.7690, 3.8032, 4.2764, 3.8073, 2.4946], device='cuda:0'), covar=tensor([0.0532, 0.0069, 0.0064, 0.0439, 0.0106, 0.0132, 0.0097, 0.0457], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0135, 0.0100, 0.0112, 0.0097, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 00:37:00,872 INFO [optim.py:368] (0/8) 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 00:37:17,262 INFO [train.py:904] (0/8) Epoch 25, batch 5550, loss[loss=0.2803, simple_loss=0.3378, pruned_loss=0.1114, over 11019.00 frames. ], tot_loss[loss=0.1923, simple_loss=0.2817, pruned_loss=0.05142, over 3149608.62 frames. ], batch size: 249, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:38:34,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9947, 2.2721, 2.3094, 2.8248, 1.9191, 3.1507, 1.8536, 2.6177], device='cuda:0'), covar=tensor([0.1349, 0.0705, 0.1205, 0.0245, 0.0148, 0.0441, 0.1744, 0.0826], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0205, 0.0215, 0.0206, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:38:41,408 INFO [train.py:904] (0/8) Epoch 25, batch 5600, loss[loss=0.2677, simple_loss=0.3279, pruned_loss=0.1038, over 11182.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2858, pruned_loss=0.05525, over 3124535.57 frames. ], batch size: 248, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:39:02,221 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:39:32,730 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 00:39:47,616 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.932e+02 3.336e+02 4.056e+02 4.975e+02 8.511e+02, threshold=8.112e+02, percent-clipped=2.0 2023-05-02 00:39:49,286 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0625, 4.1043, 4.4205, 4.3893, 4.4060, 4.1543, 4.1578, 4.1391], device='cuda:0'), covar=tensor([0.0417, 0.0784, 0.0415, 0.0454, 0.0521, 0.0489, 0.0958, 0.0593], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0469, 0.0458, 0.0417, 0.0501, 0.0479, 0.0558, 0.0381], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 00:40:04,588 INFO [train.py:904] (0/8) Epoch 25, batch 5650, loss[loss=0.2089, simple_loss=0.3056, pruned_loss=0.05606, over 16854.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2907, pruned_loss=0.059, over 3105388.12 frames. ], batch size: 42, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:40:41,209 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:40:42,322 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0340, 5.4071, 5.6449, 5.2581, 5.4519, 5.9700, 5.3970, 5.1249], device='cuda:0'), covar=tensor([0.1016, 0.1926, 0.2342, 0.2016, 0.2135, 0.0939, 0.1710, 0.2657], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0618, 0.0676, 0.0502, 0.0669, 0.0707, 0.0526, 0.0675], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:41:21,991 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.59 vs. limit=5.0 2023-05-02 00:41:22,858 INFO [train.py:904] (0/8) Epoch 25, batch 5700, loss[loss=0.2014, simple_loss=0.2893, pruned_loss=0.05674, over 17022.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2916, pruned_loss=0.06039, over 3089286.88 frames. ], batch size: 55, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:42:25,338 INFO [optim.py:368] (0/8) 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 00:42:34,391 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:42:43,883 INFO [train.py:904] (0/8) Epoch 25, batch 5750, loss[loss=0.1926, simple_loss=0.2834, pruned_loss=0.0509, over 16853.00 frames. ], tot_loss[loss=0.2099, simple_loss=0.2949, pruned_loss=0.06244, over 3058785.38 frames. ], batch size: 116, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:43:13,922 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:43:19,339 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9102, 4.1623, 3.9752, 4.0337, 3.7102, 3.8089, 3.8593, 4.1742], device='cuda:0'), covar=tensor([0.1065, 0.0911, 0.1015, 0.0895, 0.0809, 0.1648, 0.0926, 0.1010], device='cuda:0'), in_proj_covar=tensor([0.0692, 0.0834, 0.0690, 0.0645, 0.0531, 0.0532, 0.0701, 0.0653], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:43:53,995 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:44:07,479 INFO [train.py:904] (0/8) Epoch 25, batch 5800, loss[loss=0.1699, simple_loss=0.2671, pruned_loss=0.03636, over 16813.00 frames. ], tot_loss[loss=0.2081, simple_loss=0.2942, pruned_loss=0.06102, over 3057503.84 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:44:24,502 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-05-02 00:44:32,289 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:45:09,431 INFO [optim.py:368] (0/8) 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 00:45:25,862 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.62 vs. limit=2.0 2023-05-02 00:45:26,298 INFO [train.py:904] (0/8) Epoch 25, batch 5850, loss[loss=0.1685, simple_loss=0.2678, pruned_loss=0.0346, over 16677.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2921, pruned_loss=0.0595, over 3049674.28 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:45:37,478 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6278, 2.6106, 2.1175, 2.5090, 3.0260, 2.6850, 3.1704, 3.2640], device='cuda:0'), covar=tensor([0.0126, 0.0505, 0.0657, 0.0469, 0.0305, 0.0448, 0.0276, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0239, 0.0228, 0.0230, 0.0240, 0.0238, 0.0240, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:45:50,724 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0220, 2.3461, 2.3402, 2.8307, 1.7282, 3.2055, 1.8213, 2.6555], device='cuda:0'), covar=tensor([0.1185, 0.0722, 0.1163, 0.0214, 0.0129, 0.0435, 0.1547, 0.0793], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0178, 0.0196, 0.0195, 0.0205, 0.0215, 0.0205, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:46:26,100 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 00:46:46,945 INFO [train.py:904] (0/8) Epoch 25, batch 5900, loss[loss=0.182, simple_loss=0.2723, pruned_loss=0.04583, over 16172.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2916, pruned_loss=0.0595, over 3050917.91 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:46:50,166 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 00:46:54,144 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9074, 4.1517, 3.9794, 4.0236, 3.6858, 3.8199, 3.8286, 4.1672], device='cuda:0'), covar=tensor([0.1001, 0.0839, 0.0995, 0.0864, 0.0820, 0.1398, 0.0884, 0.0877], device='cuda:0'), in_proj_covar=tensor([0.0693, 0.0835, 0.0691, 0.0646, 0.0533, 0.0532, 0.0701, 0.0653], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:47:52,233 INFO [optim.py:368] (0/8) 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 00:48:00,065 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4306, 2.9454, 2.6516, 2.2425, 2.2552, 2.2486, 2.9396, 2.8429], device='cuda:0'), covar=tensor([0.2669, 0.0708, 0.1685, 0.2637, 0.2330, 0.2249, 0.0553, 0.1393], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0273, 0.0311, 0.0321, 0.0303, 0.0270, 0.0303, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:48:08,078 INFO [train.py:904] (0/8) Epoch 25, batch 5950, loss[loss=0.2103, simple_loss=0.2935, pruned_loss=0.06359, over 11772.00 frames. ], tot_loss[loss=0.205, simple_loss=0.293, pruned_loss=0.05847, over 3047708.36 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:48:22,284 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7526, 1.8923, 2.3278, 2.6579, 2.6518, 3.0413, 1.9971, 3.0239], device='cuda:0'), covar=tensor([0.0227, 0.0580, 0.0351, 0.0391, 0.0369, 0.0213, 0.0619, 0.0158], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0194, 0.0182, 0.0186, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:48:32,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8590, 4.6612, 4.7598, 5.0522, 5.2481, 4.6942, 5.3047, 5.2540], device='cuda:0'), covar=tensor([0.2216, 0.1613, 0.2325, 0.1001, 0.0962, 0.0962, 0.0857, 0.0940], device='cuda:0'), in_proj_covar=tensor([0.0662, 0.0815, 0.0937, 0.0820, 0.0630, 0.0647, 0.0675, 0.0788], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:48:36,951 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=249571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:49:28,149 INFO [train.py:904] (0/8) Epoch 25, batch 6000, loss[loss=0.2248, simple_loss=0.3128, pruned_loss=0.06839, over 15504.00 frames. ], tot_loss[loss=0.2047, simple_loss=0.2925, pruned_loss=0.05843, over 3052418.00 frames. ], batch size: 191, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:49:28,150 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 00:49:38,597 INFO [train.py:938] (0/8) Epoch 25, validation: loss=0.1487, simple_loss=0.2613, pruned_loss=0.01811, over 944034.00 frames. 2023-05-02 00:49:38,597 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 00:50:36,547 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.971e+02 2.903e+02 3.391e+02 3.955e+02 5.249e+02, threshold=6.782e+02, percent-clipped=0.0 2023-05-02 00:50:54,744 INFO [train.py:904] (0/8) Epoch 25, batch 6050, loss[loss=0.1889, simple_loss=0.2902, pruned_loss=0.04381, over 16687.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2905, pruned_loss=0.05758, over 3069573.45 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:50:56,011 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1021, 2.2060, 2.2062, 3.7023, 2.0395, 2.5607, 2.3010, 2.3658], device='cuda:0'), covar=tensor([0.1617, 0.3847, 0.3242, 0.0640, 0.4490, 0.2712, 0.3961, 0.3386], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0463, 0.0379, 0.0332, 0.0442, 0.0529, 0.0433, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:51:02,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8961, 2.1859, 2.4578, 3.1203, 2.1863, 2.3765, 2.3324, 2.2823], device='cuda:0'), covar=tensor([0.1439, 0.3261, 0.2520, 0.0762, 0.4064, 0.2417, 0.3218, 0.3310], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0463, 0.0379, 0.0332, 0.0442, 0.0529, 0.0433, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 00:51:33,690 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1297, 3.3353, 3.2220, 5.3504, 4.0230, 4.4531, 2.2081, 3.3883], device='cuda:0'), covar=tensor([0.1260, 0.0714, 0.1055, 0.0160, 0.0422, 0.0443, 0.1449, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0179, 0.0197, 0.0196, 0.0206, 0.0217, 0.0207, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:52:12,400 INFO [train.py:904] (0/8) Epoch 25, batch 6100, loss[loss=0.202, simple_loss=0.2931, pruned_loss=0.05543, over 15260.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2896, pruned_loss=0.0561, over 3087409.19 frames. ], batch size: 190, lr: 2.69e-03, grad_scale: 16.0 2023-05-02 00:52:27,110 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4848, 4.6316, 4.8362, 4.5729, 4.7004, 5.1729, 4.6622, 4.3987], device='cuda:0'), covar=tensor([0.1408, 0.1888, 0.2264, 0.1922, 0.2361, 0.1006, 0.1664, 0.2519], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0614, 0.0675, 0.0501, 0.0666, 0.0703, 0.0522, 0.0671], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:53:15,557 INFO [optim.py:368] (0/8) 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 00:53:29,964 INFO [train.py:904] (0/8) Epoch 25, batch 6150, loss[loss=0.1659, simple_loss=0.2542, pruned_loss=0.03881, over 17236.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2874, pruned_loss=0.05528, over 3097752.29 frames. ], batch size: 52, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:54:13,602 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3595, 4.0159, 3.9700, 2.5338, 3.6007, 4.0336, 3.6194, 2.3450], device='cuda:0'), covar=tensor([0.0592, 0.0053, 0.0058, 0.0458, 0.0107, 0.0115, 0.0104, 0.0453], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0136, 0.0100, 0.0113, 0.0097, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 00:54:51,560 INFO [train.py:904] (0/8) Epoch 25, batch 6200, loss[loss=0.1807, simple_loss=0.2718, pruned_loss=0.04483, over 16666.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.2855, pruned_loss=0.05499, over 3091793.06 frames. ], batch size: 134, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:55:09,955 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:55:37,791 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0628, 5.4753, 5.6705, 5.3781, 5.4772, 5.9937, 5.4443, 5.1807], device='cuda:0'), covar=tensor([0.0937, 0.1734, 0.2237, 0.2061, 0.2428, 0.1016, 0.1584, 0.2588], device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0619, 0.0680, 0.0505, 0.0670, 0.0708, 0.0525, 0.0676], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 00:55:55,238 INFO [optim.py:368] (0/8) 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 00:56:10,151 INFO [train.py:904] (0/8) Epoch 25, batch 6250, loss[loss=0.1665, simple_loss=0.2624, pruned_loss=0.03533, over 16745.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2846, pruned_loss=0.05436, over 3097882.22 frames. ], batch size: 89, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:56:38,381 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=249871.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:56:44,438 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=249875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:57:16,266 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0192, 2.2849, 2.3479, 2.7340, 1.7483, 3.1061, 1.8762, 2.7104], device='cuda:0'), covar=tensor([0.1171, 0.0677, 0.1032, 0.0193, 0.0119, 0.0372, 0.1423, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0179, 0.0197, 0.0196, 0.0206, 0.0217, 0.0207, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 00:57:26,272 INFO [train.py:904] (0/8) Epoch 25, batch 6300, loss[loss=0.1997, simple_loss=0.2897, pruned_loss=0.05491, over 16525.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2855, pruned_loss=0.05475, over 3091493.83 frames. ], batch size: 75, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:57:52,579 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=249919.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:58:10,062 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.90 vs. limit=5.0 2023-05-02 00:58:29,095 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.822e+02 2.746e+02 3.246e+02 3.923e+02 7.422e+02, threshold=6.492e+02, percent-clipped=1.0 2023-05-02 00:58:45,049 INFO [train.py:904] (0/8) Epoch 25, batch 6350, loss[loss=0.2045, simple_loss=0.2956, pruned_loss=0.05667, over 16285.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2871, pruned_loss=0.05649, over 3073705.88 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 00:58:48,553 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249955.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 00:59:20,551 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=249976.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 00:59:55,902 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-250000.pt 2023-05-02 01:00:04,122 INFO [train.py:904] (0/8) Epoch 25, batch 6400, loss[loss=0.1797, simple_loss=0.2707, pruned_loss=0.04437, over 16800.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.287, pruned_loss=0.05713, over 3077370.94 frames. ], batch size: 83, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:00:17,373 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7661, 5.0167, 5.1970, 4.9619, 5.0581, 5.5569, 5.0051, 4.8164], device='cuda:0'), covar=tensor([0.1072, 0.1830, 0.2161, 0.1815, 0.2184, 0.0932, 0.1711, 0.2305], device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0618, 0.0679, 0.0505, 0.0670, 0.0706, 0.0525, 0.0676], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 01:00:23,796 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250016.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 01:00:56,544 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:01:04,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8180, 4.7710, 4.6340, 3.8402, 4.7549, 1.7097, 4.4883, 4.2214], device='cuda:0'), covar=tensor([0.0131, 0.0130, 0.0202, 0.0423, 0.0113, 0.2907, 0.0181, 0.0331], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0167, 0.0207, 0.0184, 0.0184, 0.0213, 0.0196, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:01:05,941 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.084e+02 2.850e+02 3.461e+02 4.247e+02 7.668e+02, threshold=6.921e+02, percent-clipped=1.0 2023-05-02 01:01:20,206 INFO [train.py:904] (0/8) Epoch 25, batch 6450, loss[loss=0.1875, simple_loss=0.2762, pruned_loss=0.04944, over 16229.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2867, pruned_loss=0.05669, over 3060450.77 frames. ], batch size: 165, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:02:01,839 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 01:02:03,862 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0347, 3.2440, 3.4282, 2.0171, 3.0693, 2.1870, 3.5166, 3.5499], device='cuda:0'), covar=tensor([0.0253, 0.0837, 0.0608, 0.2188, 0.0859, 0.1028, 0.0613, 0.0983], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0167, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:02:36,770 INFO [train.py:904] (0/8) Epoch 25, batch 6500, loss[loss=0.2378, simple_loss=0.298, pruned_loss=0.08884, over 11425.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2853, pruned_loss=0.05628, over 3062892.16 frames. ], batch size: 246, lr: 2.69e-03, grad_scale: 8.0 2023-05-02 01:03:19,069 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3589, 2.7727, 3.0278, 1.9587, 2.7939, 2.0517, 3.0625, 3.1180], device='cuda:0'), covar=tensor([0.0281, 0.0881, 0.0630, 0.2178, 0.0861, 0.1100, 0.0635, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0166, 0.0169, 0.0154, 0.0146, 0.0131, 0.0144, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:03:40,566 INFO [optim.py:368] (0/8) 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 01:03:52,690 INFO [train.py:904] (0/8) Epoch 25, batch 6550, loss[loss=0.2515, simple_loss=0.3223, pruned_loss=0.09033, over 11664.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2877, pruned_loss=0.05708, over 3067992.82 frames. ], batch size: 247, lr: 2.69e-03, grad_scale: 4.0 2023-05-02 01:04:18,261 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:04:36,942 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5178, 2.1017, 1.6033, 1.7702, 2.3873, 2.0558, 2.2619, 2.6134], device='cuda:0'), covar=tensor([0.0294, 0.0571, 0.0788, 0.0695, 0.0371, 0.0532, 0.0265, 0.0387], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0238, 0.0228, 0.0230, 0.0239, 0.0238, 0.0239, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:05:05,614 INFO [train.py:904] (0/8) Epoch 25, batch 6600, loss[loss=0.2028, simple_loss=0.2904, pruned_loss=0.05758, over 16746.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2897, pruned_loss=0.05758, over 3054309.05 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:06:08,502 INFO [optim.py:368] (0/8) 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 01:06:13,867 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6874, 4.7336, 4.5896, 4.2340, 4.2377, 4.6746, 4.4756, 4.3675], device='cuda:0'), covar=tensor([0.0605, 0.0515, 0.0317, 0.0343, 0.1021, 0.0485, 0.0444, 0.0691], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0451, 0.0349, 0.0354, 0.0355, 0.0409, 0.0240, 0.0424], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:06:18,498 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 01:06:21,908 INFO [train.py:904] (0/8) Epoch 25, batch 6650, loss[loss=0.1721, simple_loss=0.2603, pruned_loss=0.04199, over 16683.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2897, pruned_loss=0.05758, over 3080654.01 frames. ], batch size: 76, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:23,344 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5517, 3.2477, 3.7975, 1.8944, 3.9048, 3.9421, 2.9961, 2.8968], device='cuda:0'), covar=tensor([0.0794, 0.0318, 0.0170, 0.1241, 0.0079, 0.0155, 0.0449, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0111, 0.0101, 0.0140, 0.0085, 0.0131, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:07:36,440 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 01:07:37,556 INFO [train.py:904] (0/8) Epoch 25, batch 6700, loss[loss=0.1956, simple_loss=0.2836, pruned_loss=0.05378, over 16866.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.2887, pruned_loss=0.05811, over 3085994.96 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:07:50,681 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250311.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 01:08:22,567 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=250332.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:08:41,137 INFO [optim.py:368] (0/8) 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 01:08:54,019 INFO [train.py:904] (0/8) Epoch 25, batch 6750, loss[loss=0.2698, simple_loss=0.3325, pruned_loss=0.1035, over 12117.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2883, pruned_loss=0.05848, over 3077643.73 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:09:24,745 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 01:10:10,579 INFO [train.py:904] (0/8) Epoch 25, batch 6800, loss[loss=0.1834, simple_loss=0.2744, pruned_loss=0.04622, over 16738.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2874, pruned_loss=0.05786, over 3085742.40 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:16,511 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 25, batch 6850, loss[loss=0.2071, simple_loss=0.3246, pruned_loss=0.04479, over 17155.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2882, pruned_loss=0.05762, over 3102863.81 frames. ], batch size: 46, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:11:28,066 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7802, 2.3120, 1.9259, 2.0674, 2.6882, 2.3661, 2.5191, 2.8662], device='cuda:0'), covar=tensor([0.0244, 0.0536, 0.0655, 0.0603, 0.0303, 0.0440, 0.0289, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0218, 0.0238, 0.0227, 0.0228, 0.0238, 0.0237, 0.0237, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:11:43,237 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7623, 3.9666, 4.0950, 2.3359, 3.6282, 3.0431, 4.2474, 4.2621], device='cuda:0'), covar=tensor([0.0202, 0.0723, 0.0600, 0.1935, 0.0688, 0.0879, 0.0426, 0.0705], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0168, 0.0170, 0.0155, 0.0148, 0.0131, 0.0145, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:11:52,838 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:12:21,976 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7564, 4.7432, 5.1643, 5.1536, 5.1188, 4.8613, 4.7880, 4.6747], device='cuda:0'), covar=tensor([0.0424, 0.0815, 0.0556, 0.0475, 0.0535, 0.0640, 0.1081, 0.0587], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0470, 0.0456, 0.0418, 0.0501, 0.0478, 0.0555, 0.0382], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 01:12:43,351 INFO [train.py:904] (0/8) Epoch 25, batch 6900, loss[loss=0.1871, simple_loss=0.2813, pruned_loss=0.04643, over 16691.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2901, pruned_loss=0.05662, over 3117912.52 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:13:06,144 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250518.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:13:47,068 INFO [optim.py:368] (0/8) 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 01:13:51,487 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9276, 2.1833, 2.2554, 3.4189, 2.0636, 2.4434, 2.2652, 2.2730], device='cuda:0'), covar=tensor([0.1445, 0.3299, 0.2805, 0.0637, 0.4064, 0.2287, 0.3345, 0.3092], device='cuda:0'), in_proj_covar=tensor([0.0410, 0.0459, 0.0377, 0.0330, 0.0440, 0.0526, 0.0431, 0.0537], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:14:00,193 INFO [train.py:904] (0/8) Epoch 25, batch 6950, loss[loss=0.2641, simple_loss=0.3323, pruned_loss=0.09789, over 11136.00 frames. ], tot_loss[loss=0.2051, simple_loss=0.2924, pruned_loss=0.05894, over 3091035.66 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:14:04,153 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 01:14:41,254 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0003, 3.0267, 1.8798, 3.2358, 2.2653, 3.2893, 2.1119, 2.5726], device='cuda:0'), covar=tensor([0.0352, 0.0451, 0.1767, 0.0264, 0.0910, 0.0634, 0.1581, 0.0838], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0180, 0.0195, 0.0168, 0.0177, 0.0218, 0.0203, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:15:14,730 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1620, 2.4564, 2.0267, 2.1678, 2.8532, 2.4936, 2.7628, 3.0542], device='cuda:0'), covar=tensor([0.0200, 0.0555, 0.0675, 0.0577, 0.0314, 0.0448, 0.0281, 0.0334], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0238, 0.0228, 0.0229, 0.0239, 0.0238, 0.0238, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:15:18,387 INFO [train.py:904] (0/8) Epoch 25, batch 7000, loss[loss=0.2078, simple_loss=0.2838, pruned_loss=0.06593, over 11624.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2917, pruned_loss=0.05772, over 3104395.29 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:15:26,840 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3462, 5.2818, 5.1162, 4.3010, 5.1645, 1.5984, 4.8791, 4.7403], device='cuda:0'), covar=tensor([0.0116, 0.0136, 0.0217, 0.0459, 0.0118, 0.3207, 0.0220, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0166, 0.0206, 0.0183, 0.0183, 0.0213, 0.0195, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:15:31,400 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250611.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 01:15:48,801 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4738, 2.9507, 2.7193, 2.3016, 2.3059, 2.3519, 2.9661, 2.8655], device='cuda:0'), covar=tensor([0.2361, 0.0711, 0.1553, 0.2567, 0.2392, 0.2166, 0.0501, 0.1390], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0273, 0.0311, 0.0320, 0.0303, 0.0269, 0.0301, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 01:16:03,513 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=250632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:16:18,608 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 01:16:22,013 INFO [optim.py:368] (0/8) 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 01:16:25,615 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.88 vs. limit=5.0 2023-05-02 01:16:35,099 INFO [train.py:904] (0/8) Epoch 25, batch 7050, loss[loss=0.2084, simple_loss=0.2981, pruned_loss=0.05937, over 16912.00 frames. ], tot_loss[loss=0.2044, simple_loss=0.2934, pruned_loss=0.05771, over 3110959.87 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:16:44,700 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250659.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 01:16:49,616 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250662.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:06,041 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8821, 2.7211, 2.6012, 1.9468, 2.5521, 2.6945, 2.5635, 1.9871], device='cuda:0'), covar=tensor([0.0432, 0.0094, 0.0094, 0.0376, 0.0148, 0.0136, 0.0144, 0.0382], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0087, 0.0087, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 01:17:14,032 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2897, 6.0224, 6.2275, 5.8209, 5.9097, 6.4143, 5.9443, 5.7314], device='cuda:0'), covar=tensor([0.0904, 0.1617, 0.2123, 0.1777, 0.2458, 0.0879, 0.1462, 0.2217], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0619, 0.0680, 0.0507, 0.0672, 0.0708, 0.0526, 0.0678], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 01:17:17,437 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=250680.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:17:51,105 INFO [train.py:904] (0/8) Epoch 25, batch 7100, loss[loss=0.1904, simple_loss=0.2836, pruned_loss=0.0486, over 16900.00 frames. ], tot_loss[loss=0.2035, simple_loss=0.2915, pruned_loss=0.05771, over 3096577.84 frames. ], batch size: 109, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:18:05,469 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5581, 2.2439, 1.9166, 1.9556, 2.5093, 2.1620, 2.3559, 2.6884], device='cuda:0'), covar=tensor([0.0220, 0.0399, 0.0506, 0.0474, 0.0268, 0.0399, 0.0189, 0.0247], device='cuda:0'), in_proj_covar=tensor([0.0217, 0.0237, 0.0227, 0.0228, 0.0238, 0.0236, 0.0236, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:18:23,468 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250723.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:18:56,842 INFO [optim.py:368] (0/8) 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 01:19:09,291 INFO [train.py:904] (0/8) Epoch 25, batch 7150, loss[loss=0.2082, simple_loss=0.2906, pruned_loss=0.06294, over 16379.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2903, pruned_loss=0.05817, over 3079397.04 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:20:22,682 INFO [train.py:904] (0/8) Epoch 25, batch 7200, loss[loss=0.1569, simple_loss=0.258, pruned_loss=0.02788, over 16823.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2879, pruned_loss=0.05627, over 3085929.05 frames. ], batch size: 102, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:21:01,711 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 01:21:08,268 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,137 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=250845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:21:28,840 INFO [optim.py:368] (0/8) 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 01:21:41,050 INFO [train.py:904] (0/8) Epoch 25, batch 7250, loss[loss=0.197, simple_loss=0.277, pruned_loss=0.05846, over 16446.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2859, pruned_loss=0.0553, over 3080257.88 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:22:12,949 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6288, 2.5769, 1.9181, 2.6755, 2.0842, 2.7758, 2.1842, 2.4036], device='cuda:0'), covar=tensor([0.0328, 0.0368, 0.1243, 0.0263, 0.0672, 0.0517, 0.1176, 0.0594], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0180, 0.0196, 0.0168, 0.0178, 0.0219, 0.0204, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:22:42,102 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:22:55,038 INFO [train.py:904] (0/8) Epoch 25, batch 7300, loss[loss=0.1981, simple_loss=0.2853, pruned_loss=0.05545, over 17047.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2851, pruned_loss=0.05512, over 3081482.32 frames. ], batch size: 53, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:23:00,830 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=250906.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 01:23:59,654 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.919e+02 2.893e+02 3.388e+02 3.918e+02 7.629e+02, threshold=6.776e+02, percent-clipped=7.0 2023-05-02 01:24:09,702 INFO [train.py:904] (0/8) Epoch 25, batch 7350, loss[loss=0.2428, simple_loss=0.3056, pruned_loss=0.08995, over 11106.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2863, pruned_loss=0.05667, over 3048128.96 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:21,493 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1811, 3.4640, 3.5338, 2.3137, 3.2801, 3.5294, 3.2874, 2.1481], device='cuda:0'), covar=tensor([0.0518, 0.0067, 0.0064, 0.0410, 0.0109, 0.0138, 0.0109, 0.0441], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0087, 0.0087, 0.0134, 0.0099, 0.0111, 0.0096, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 01:25:27,073 INFO [train.py:904] (0/8) Epoch 25, batch 7400, loss[loss=0.2176, simple_loss=0.3142, pruned_loss=0.06045, over 16405.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2876, pruned_loss=0.05706, over 3059371.27 frames. ], batch size: 35, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:25:50,973 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:26:35,985 INFO [optim.py:368] (0/8) 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 01:26:46,738 INFO [train.py:904] (0/8) Epoch 25, batch 7450, loss[loss=0.1784, simple_loss=0.2715, pruned_loss=0.0427, over 16771.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2886, pruned_loss=0.05818, over 3057669.87 frames. ], batch size: 83, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:28:06,271 INFO [train.py:904] (0/8) Epoch 25, batch 7500, loss[loss=0.2448, simple_loss=0.3159, pruned_loss=0.0869, over 11341.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2881, pruned_loss=0.05722, over 3044040.84 frames. ], batch size: 248, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:28:12,276 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0859, 2.4077, 2.5263, 1.9326, 2.6546, 2.7455, 2.3932, 2.3371], device='cuda:0'), covar=tensor([0.0752, 0.0285, 0.0297, 0.1041, 0.0157, 0.0377, 0.0477, 0.0497], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0139, 0.0084, 0.0129, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:29:12,778 INFO [optim.py:368] (0/8) 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 01:29:23,962 INFO [train.py:904] (0/8) Epoch 25, batch 7550, loss[loss=0.1859, simple_loss=0.2734, pruned_loss=0.04918, over 16827.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2869, pruned_loss=0.05706, over 3055802.52 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 4.0 2023-05-02 01:29:27,022 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4646, 3.2938, 2.6632, 2.2267, 2.2869, 2.3185, 3.3857, 3.0725], device='cuda:0'), covar=tensor([0.3053, 0.0711, 0.1868, 0.2644, 0.2533, 0.2218, 0.0548, 0.1401], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0273, 0.0311, 0.0321, 0.0304, 0.0270, 0.0301, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 01:30:19,159 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:32,681 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:30:37,934 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 25, batch 7600, loss[loss=0.1852, simple_loss=0.275, pruned_loss=0.04772, over 16994.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2867, pruned_loss=0.05781, over 3040373.50 frames. ], batch size: 41, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:31:43,659 INFO [optim.py:368] (0/8) 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 01:31:53,117 INFO [train.py:904] (0/8) Epoch 25, batch 7650, loss[loss=0.248, simple_loss=0.3115, pruned_loss=0.09221, over 11325.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2873, pruned_loss=0.05818, over 3054287.32 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:31:54,104 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 01:32:02,532 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251259.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:32:34,094 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2921, 3.2095, 3.4817, 1.7148, 3.5949, 3.6866, 2.8915, 2.6815], device='cuda:0'), covar=tensor([0.0892, 0.0304, 0.0213, 0.1369, 0.0099, 0.0205, 0.0468, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0100, 0.0139, 0.0084, 0.0130, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:32:35,788 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.75 vs. limit=2.0 2023-05-02 01:33:06,989 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:33:08,942 INFO [train.py:904] (0/8) Epoch 25, batch 7700, loss[loss=0.1914, simple_loss=0.2756, pruned_loss=0.0536, over 16686.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2872, pruned_loss=0.05848, over 3064423.63 frames. ], batch size: 57, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:33:15,468 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 01:33:31,791 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:34:14,485 INFO [optim.py:368] (0/8) 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 01:34:25,532 INFO [train.py:904] (0/8) Epoch 25, batch 7750, loss[loss=0.2137, simple_loss=0.3031, pruned_loss=0.06215, over 16955.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2876, pruned_loss=0.05767, over 3099675.16 frames. ], batch size: 41, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:34:39,718 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:34:45,581 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:35:40,757 INFO [train.py:904] (0/8) Epoch 25, batch 7800, loss[loss=0.239, simple_loss=0.3072, pruned_loss=0.08538, over 11140.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2882, pruned_loss=0.05846, over 3085532.30 frames. ], batch size: 247, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:35:42,342 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251404.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:36:45,160 INFO [optim.py:368] (0/8) 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 01:36:55,311 INFO [train.py:904] (0/8) Epoch 25, batch 7850, loss[loss=0.1897, simple_loss=0.2834, pruned_loss=0.04796, over 16391.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2889, pruned_loss=0.05833, over 3072070.00 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:37:13,945 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251465.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:37:50,403 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:37:55,461 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8263, 4.2024, 3.1645, 2.3801, 2.8226, 2.6716, 4.7302, 3.6366], device='cuda:0'), covar=tensor([0.3187, 0.0666, 0.1854, 0.2805, 0.2885, 0.2142, 0.0394, 0.1412], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0274, 0.0313, 0.0322, 0.0305, 0.0271, 0.0301, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 01:38:06,791 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251501.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 01:38:08,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.54 vs. limit=5.0 2023-05-02 01:38:09,248 INFO [train.py:904] (0/8) Epoch 25, batch 7900, loss[loss=0.1962, simple_loss=0.2858, pruned_loss=0.05336, over 16813.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2882, pruned_loss=0.05799, over 3083641.29 frames. ], batch size: 116, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:38:17,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3311, 3.5303, 3.7655, 2.0883, 3.2576, 2.4422, 3.6899, 3.8672], device='cuda:0'), covar=tensor([0.0271, 0.0804, 0.0576, 0.2283, 0.0829, 0.1033, 0.0653, 0.0980], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0167, 0.0169, 0.0155, 0.0147, 0.0131, 0.0144, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:38:37,417 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4845, 3.4865, 3.4451, 2.7110, 3.3205, 2.1751, 3.1439, 2.7800], device='cuda:0'), covar=tensor([0.0186, 0.0156, 0.0204, 0.0230, 0.0112, 0.2315, 0.0145, 0.0246], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0163, 0.0203, 0.0180, 0.0179, 0.0209, 0.0191, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:39:03,593 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:08,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0259, 3.0628, 1.6832, 3.2354, 2.2767, 3.2751, 1.8665, 2.4720], device='cuda:0'), covar=tensor([0.0378, 0.0442, 0.2130, 0.0303, 0.0946, 0.0595, 0.2010, 0.0959], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0180, 0.0197, 0.0168, 0.0178, 0.0220, 0.0205, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:39:17,280 INFO [optim.py:368] (0/8) 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 01:39:22,130 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:39:28,561 INFO [train.py:904] (0/8) Epoch 25, batch 7950, loss[loss=0.1881, simple_loss=0.2752, pruned_loss=0.05051, over 16408.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2882, pruned_loss=0.05772, over 3089931.23 frames. ], batch size: 146, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:39:30,112 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:40:27,550 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0038, 2.0531, 2.5981, 2.9341, 2.7688, 3.4726, 2.2570, 3.4422], device='cuda:0'), covar=tensor([0.0266, 0.0543, 0.0379, 0.0337, 0.0358, 0.0157, 0.0546, 0.0141], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0196, 0.0183, 0.0186, 0.0202, 0.0161, 0.0199, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:40:46,608 INFO [train.py:904] (0/8) Epoch 25, batch 8000, loss[loss=0.2019, simple_loss=0.2987, pruned_loss=0.05251, over 16713.00 frames. ], tot_loss[loss=0.2022, simple_loss=0.2885, pruned_loss=0.05796, over 3089798.80 frames. ], batch size: 124, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:41:43,333 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6597, 4.6955, 5.0378, 4.9937, 5.0289, 4.6990, 4.7091, 4.5510], device='cuda:0'), covar=tensor([0.0336, 0.0572, 0.0377, 0.0435, 0.0465, 0.0446, 0.0943, 0.0512], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0477, 0.0461, 0.0424, 0.0507, 0.0487, 0.0562, 0.0388], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 01:41:44,837 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 01:42:01,900 INFO [train.py:904] (0/8) Epoch 25, batch 8050, loss[loss=0.2086, simple_loss=0.2928, pruned_loss=0.06219, over 16670.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2894, pruned_loss=0.05897, over 3067081.22 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:42:08,492 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:42:08,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6035, 4.4934, 4.7066, 4.8259, 4.9985, 4.5032, 4.9644, 5.0063], device='cuda:0'), covar=tensor([0.2000, 0.1220, 0.1485, 0.0699, 0.0598, 0.0948, 0.0668, 0.0715], device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0793, 0.0914, 0.0802, 0.0616, 0.0634, 0.0669, 0.0774], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:42:22,419 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1280, 2.2234, 2.6993, 3.0649, 2.8879, 3.6195, 2.3067, 3.6400], device='cuda:0'), covar=tensor([0.0230, 0.0491, 0.0347, 0.0309, 0.0341, 0.0160, 0.0525, 0.0153], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0195, 0.0182, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:43:16,985 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:43:17,645 INFO [train.py:904] (0/8) Epoch 25, batch 8100, loss[loss=0.1883, simple_loss=0.2791, pruned_loss=0.04872, over 16487.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.288, pruned_loss=0.05762, over 3084839.02 frames. ], batch size: 75, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:43:27,352 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=251709.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:44:16,506 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1441, 2.4319, 2.6030, 1.9584, 2.6948, 2.7976, 2.4338, 2.3678], device='cuda:0'), covar=tensor([0.0663, 0.0281, 0.0242, 0.0924, 0.0131, 0.0302, 0.0426, 0.0447], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0100, 0.0139, 0.0084, 0.0129, 0.0129, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:44:22,960 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.895e+02 2.808e+02 3.404e+02 4.395e+02 1.324e+03, threshold=6.808e+02, percent-clipped=5.0 2023-05-02 01:44:33,804 INFO [train.py:904] (0/8) Epoch 25, batch 8150, loss[loss=0.16, simple_loss=0.2496, pruned_loss=0.0352, over 16692.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2857, pruned_loss=0.05666, over 3097645.29 frames. ], batch size: 76, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:44:44,523 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251760.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:00,646 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=251770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:45:50,915 INFO [train.py:904] (0/8) Epoch 25, batch 8200, loss[loss=0.1932, simple_loss=0.2832, pruned_loss=0.05161, over 15241.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2829, pruned_loss=0.05567, over 3109650.56 frames. ], batch size: 191, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:46:59,192 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.400e+02 2.618e+02 3.043e+02 3.668e+02 8.118e+02, threshold=6.086e+02, percent-clipped=2.0 2023-05-02 01:47:04,913 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3811, 1.7013, 2.1699, 2.4318, 2.4618, 2.7537, 2.0097, 2.7062], device='cuda:0'), covar=tensor([0.0257, 0.0574, 0.0370, 0.0370, 0.0376, 0.0227, 0.0515, 0.0178], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0194, 0.0181, 0.0184, 0.0200, 0.0159, 0.0197, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:47:10,673 INFO [train.py:904] (0/8) Epoch 25, batch 8250, loss[loss=0.1842, simple_loss=0.2749, pruned_loss=0.04675, over 16640.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2817, pruned_loss=0.05352, over 3086494.56 frames. ], batch size: 57, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:47:12,519 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:29,884 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=251902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:48:30,711 INFO [train.py:904] (0/8) Epoch 25, batch 8300, loss[loss=0.1748, simple_loss=0.2734, pruned_loss=0.0381, over 16731.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2794, pruned_loss=0.05084, over 3073049.15 frames. ], batch size: 134, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:41,166 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 2.151e+02 2.560e+02 3.132e+02 5.384e+02, threshold=5.121e+02, percent-clipped=0.0 2023-05-02 01:49:52,646 INFO [train.py:904] (0/8) Epoch 25, batch 8350, loss[loss=0.2156, simple_loss=0.29, pruned_loss=0.07058, over 11941.00 frames. ], tot_loss[loss=0.1886, simple_loss=0.2791, pruned_loss=0.04908, over 3068462.84 frames. ], batch size: 246, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:49:59,403 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=251957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:50:55,687 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9311, 3.6730, 4.1000, 2.1894, 4.2515, 4.2918, 3.3521, 3.3744], device='cuda:0'), covar=tensor([0.0665, 0.0250, 0.0164, 0.1099, 0.0073, 0.0161, 0.0343, 0.0367], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0109, 0.0098, 0.0137, 0.0083, 0.0127, 0.0127, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 01:51:05,357 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=251997.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:51:10,796 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-252000.pt 2023-05-02 01:51:17,320 INFO [train.py:904] (0/8) Epoch 25, batch 8400, loss[loss=0.1638, simple_loss=0.259, pruned_loss=0.03433, over 16277.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2766, pruned_loss=0.04681, over 3079612.32 frames. ], batch size: 165, lr: 2.68e-03, grad_scale: 8.0 2023-05-02 01:51:22,061 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252005.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:51:48,667 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8150, 3.5362, 3.8848, 2.1961, 4.0485, 4.0681, 3.2144, 3.2599], device='cuda:0'), covar=tensor([0.0648, 0.0243, 0.0207, 0.1073, 0.0075, 0.0189, 0.0365, 0.0366], device='cuda:0'), in_proj_covar=tensor([0.0146, 0.0109, 0.0098, 0.0137, 0.0083, 0.0127, 0.0127, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 01:51:54,262 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 01:52:03,679 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9508, 2.6934, 2.9396, 2.1002, 2.7585, 2.1672, 2.8401, 2.9038], device='cuda:0'), covar=tensor([0.0317, 0.1026, 0.0489, 0.1951, 0.0768, 0.1021, 0.0602, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0154, 0.0164, 0.0166, 0.0152, 0.0144, 0.0128, 0.0142, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:52:27,819 INFO [optim.py:368] (0/8) 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 01:52:40,187 INFO [train.py:904] (0/8) Epoch 25, batch 8450, loss[loss=0.1784, simple_loss=0.2625, pruned_loss=0.04717, over 12194.00 frames. ], tot_loss[loss=0.1829, simple_loss=0.2749, pruned_loss=0.04546, over 3067110.42 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:52:48,131 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1937, 2.5154, 2.6641, 1.9741, 2.7947, 2.8405, 2.5204, 2.5341], device='cuda:0'), covar=tensor([0.0618, 0.0270, 0.0217, 0.0965, 0.0124, 0.0299, 0.0449, 0.0389], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0108, 0.0097, 0.0136, 0.0082, 0.0126, 0.0126, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 01:52:51,816 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252060.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:52:59,912 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:53:58,692 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8972, 5.1780, 5.3570, 5.1186, 5.1564, 5.7028, 5.1866, 4.8957], device='cuda:0'), covar=tensor([0.0956, 0.1797, 0.2098, 0.1796, 0.2366, 0.0837, 0.1569, 0.2517], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0600, 0.0665, 0.0493, 0.0654, 0.0689, 0.0517, 0.0661], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 01:54:01,932 INFO [train.py:904] (0/8) Epoch 25, batch 8500, loss[loss=0.1578, simple_loss=0.2506, pruned_loss=0.03251, over 15317.00 frames. ], tot_loss[loss=0.1791, simple_loss=0.2714, pruned_loss=0.04341, over 3056046.48 frames. ], batch size: 190, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:54:11,541 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252108.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 01:55:15,324 INFO [optim.py:368] (0/8) 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 01:55:28,848 INFO [train.py:904] (0/8) Epoch 25, batch 8550, loss[loss=0.2044, simple_loss=0.2961, pruned_loss=0.05638, over 16618.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.2688, pruned_loss=0.04232, over 3047566.85 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:56:18,085 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7552, 3.0806, 3.4927, 1.9884, 2.9225, 2.1996, 3.3053, 3.3091], device='cuda:0'), covar=tensor([0.0274, 0.0820, 0.0517, 0.2139, 0.0809, 0.1012, 0.0656, 0.0976], device='cuda:0'), in_proj_covar=tensor([0.0155, 0.0164, 0.0166, 0.0152, 0.0144, 0.0129, 0.0142, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 01:57:09,042 INFO [train.py:904] (0/8) Epoch 25, batch 8600, loss[loss=0.1771, simple_loss=0.2764, pruned_loss=0.03893, over 16906.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2687, pruned_loss=0.04128, over 3053414.30 frames. ], batch size: 125, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 01:57:24,009 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4112, 2.5331, 2.2237, 2.3387, 2.8827, 2.5268, 2.8654, 3.0928], device='cuda:0'), covar=tensor([0.0152, 0.0481, 0.0542, 0.0506, 0.0306, 0.0475, 0.0265, 0.0296], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0233, 0.0223, 0.0224, 0.0233, 0.0232, 0.0230, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:57:29,971 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0963, 4.0887, 4.0070, 3.2504, 4.0304, 1.7788, 3.8413, 3.6298], device='cuda:0'), covar=tensor([0.0108, 0.0095, 0.0156, 0.0230, 0.0091, 0.2745, 0.0124, 0.0273], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0163, 0.0202, 0.0179, 0.0179, 0.0210, 0.0191, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 01:58:34,360 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.652e+02 2.200e+02 2.678e+02 3.278e+02 9.763e+02, threshold=5.356e+02, percent-clipped=4.0 2023-05-02 01:58:37,313 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4032, 2.9657, 2.7038, 2.1821, 2.1453, 2.3059, 2.9787, 2.7951], device='cuda:0'), covar=tensor([0.2969, 0.0709, 0.1799, 0.3073, 0.2840, 0.2205, 0.0505, 0.1524], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0267, 0.0305, 0.0314, 0.0297, 0.0266, 0.0295, 0.0338], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 01:58:41,606 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 01:58:48,816 INFO [train.py:904] (0/8) Epoch 25, batch 8650, loss[loss=0.1829, simple_loss=0.2811, pruned_loss=0.04231, over 16400.00 frames. ], tot_loss[loss=0.1732, simple_loss=0.2669, pruned_loss=0.03975, over 3043620.26 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:00:23,226 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252297.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:00:33,698 INFO [train.py:904] (0/8) Epoch 25, batch 8700, loss[loss=0.1702, simple_loss=0.2508, pruned_loss=0.0448, over 12262.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2644, pruned_loss=0.03871, over 3059502.66 frames. ], batch size: 248, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:01:02,635 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4580, 3.5328, 2.1756, 3.8872, 2.6703, 3.8505, 2.3403, 2.8784], device='cuda:0'), covar=tensor([0.0298, 0.0361, 0.1598, 0.0224, 0.0849, 0.0493, 0.1599, 0.0778], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0175, 0.0192, 0.0163, 0.0174, 0.0213, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 02:01:53,005 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=252345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:01:54,401 INFO [optim.py:368] (0/8) 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 02:02:09,366 INFO [train.py:904] (0/8) Epoch 25, batch 8750, loss[loss=0.1711, simple_loss=0.2839, pruned_loss=0.02916, over 16914.00 frames. ], tot_loss[loss=0.1706, simple_loss=0.2647, pruned_loss=0.03827, over 3070808.83 frames. ], batch size: 102, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:02:41,278 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=252365.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:04:01,520 INFO [train.py:904] (0/8) Epoch 25, batch 8800, loss[loss=0.1647, simple_loss=0.269, pruned_loss=0.03024, over 16882.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.263, pruned_loss=0.03729, over 3071705.76 frames. ], batch size: 96, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:04:13,662 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2448, 1.6165, 1.9842, 2.2209, 2.3055, 2.4879, 1.7221, 2.4701], device='cuda:0'), covar=tensor([0.0332, 0.0599, 0.0423, 0.0393, 0.0429, 0.0275, 0.0615, 0.0183], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0192, 0.0181, 0.0182, 0.0199, 0.0158, 0.0196, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 02:04:21,278 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252417.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:06,545 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:05:17,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9801, 4.2566, 4.1344, 4.1489, 3.8211, 3.8718, 3.8920, 4.2798], device='cuda:0'), covar=tensor([0.1081, 0.0890, 0.0870, 0.0741, 0.0753, 0.1729, 0.0958, 0.0950], device='cuda:0'), in_proj_covar=tensor([0.0676, 0.0817, 0.0675, 0.0631, 0.0518, 0.0526, 0.0685, 0.0638], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 02:05:28,910 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 02:05:31,486 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.474e+02 2.215e+02 2.676e+02 3.061e+02 6.379e+02, threshold=5.352e+02, percent-clipped=4.0 2023-05-02 02:05:46,046 INFO [train.py:904] (0/8) Epoch 25, batch 8850, loss[loss=0.1523, simple_loss=0.2467, pruned_loss=0.02896, over 12415.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2652, pruned_loss=0.03666, over 3055079.06 frames. ], batch size: 250, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:06:16,209 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 02:06:35,642 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9691, 2.2365, 2.3533, 2.9827, 1.8620, 3.2732, 1.7042, 2.7858], device='cuda:0'), covar=tensor([0.1226, 0.0724, 0.1061, 0.0166, 0.0082, 0.0360, 0.1548, 0.0681], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0176, 0.0194, 0.0192, 0.0201, 0.0213, 0.0204, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 02:06:39,745 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252478.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:07,343 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3283, 3.7643, 3.7690, 2.5862, 3.3720, 3.7832, 3.5219, 2.3043], device='cuda:0'), covar=tensor([0.0520, 0.0049, 0.0043, 0.0377, 0.0103, 0.0078, 0.0083, 0.0444], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0084, 0.0086, 0.0132, 0.0098, 0.0109, 0.0094, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 02:07:16,277 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:07:16,648 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 02:07:31,608 INFO [train.py:904] (0/8) Epoch 25, batch 8900, loss[loss=0.1703, simple_loss=0.2571, pruned_loss=0.0418, over 12889.00 frames. ], tot_loss[loss=0.1699, simple_loss=0.2667, pruned_loss=0.03654, over 3072561.31 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:09:18,928 INFO [optim.py:368] (0/8) 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 02:09:34,581 INFO [train.py:904] (0/8) Epoch 25, batch 8950, loss[loss=0.1555, simple_loss=0.2507, pruned_loss=0.03014, over 12341.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2661, pruned_loss=0.03662, over 3072190.59 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:10:59,636 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252592.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:11:21,240 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1954, 3.2513, 1.9552, 3.5259, 2.4694, 3.4785, 2.1921, 2.6993], device='cuda:0'), covar=tensor([0.0359, 0.0428, 0.1817, 0.0212, 0.0918, 0.0670, 0.1602, 0.0799], device='cuda:0'), in_proj_covar=tensor([0.0168, 0.0173, 0.0190, 0.0160, 0.0172, 0.0209, 0.0198, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 02:11:21,910 INFO [train.py:904] (0/8) Epoch 25, batch 9000, loss[loss=0.1628, simple_loss=0.256, pruned_loss=0.03481, over 16270.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2626, pruned_loss=0.03548, over 3072804.41 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:11:21,911 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 02:11:31,592 INFO [train.py:938] (0/8) Epoch 25, validation: loss=0.1442, simple_loss=0.248, pruned_loss=0.02014, over 944034.00 frames. 2023-05-02 02:11:31,593 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 02:11:53,594 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:01,732 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 25, batch 9050, loss[loss=0.169, simple_loss=0.2591, pruned_loss=0.03951, over 16345.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2629, pruned_loss=0.03566, over 3082030.77 frames. ], batch size: 146, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:13:15,920 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252653.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:18,385 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4941, 3.1511, 3.3205, 1.8284, 3.4647, 3.5469, 2.9439, 2.8109], device='cuda:0'), covar=tensor([0.0698, 0.0270, 0.0209, 0.1299, 0.0103, 0.0201, 0.0432, 0.0461], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0107, 0.0095, 0.0135, 0.0081, 0.0124, 0.0125, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 02:13:46,497 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:13:56,522 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:14:58,391 INFO [train.py:904] (0/8) Epoch 25, batch 9100, loss[loss=0.1679, simple_loss=0.2651, pruned_loss=0.03535, over 16878.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2622, pruned_loss=0.03597, over 3084988.32 frames. ], batch size: 96, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:15:50,535 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2379, 2.2543, 2.2174, 3.9422, 2.1931, 2.5573, 2.3711, 2.3930], device='cuda:0'), covar=tensor([0.1346, 0.3850, 0.3405, 0.0552, 0.4376, 0.2767, 0.3665, 0.3668], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0454, 0.0373, 0.0324, 0.0434, 0.0517, 0.0425, 0.0528], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 02:16:00,776 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:16:41,929 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.565e+02 2.125e+02 2.555e+02 2.905e+02 5.383e+02, threshold=5.110e+02, percent-clipped=1.0 2023-05-02 02:16:54,461 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.67 vs. limit=2.0 2023-05-02 02:16:56,628 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 02:16:57,786 INFO [train.py:904] (0/8) Epoch 25, batch 9150, loss[loss=0.165, simple_loss=0.258, pruned_loss=0.03606, over 16102.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2625, pruned_loss=0.03581, over 3072436.48 frames. ], batch size: 165, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:17:42,144 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:20,264 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:18:20,650 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 02:18:43,296 INFO [train.py:904] (0/8) Epoch 25, batch 9200, loss[loss=0.152, simple_loss=0.249, pruned_loss=0.02745, over 16522.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2584, pruned_loss=0.03465, over 3067286.47 frames. ], batch size: 68, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:20:05,212 INFO [optim.py:368] (0/8) 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 02:20:19,106 INFO [train.py:904] (0/8) Epoch 25, batch 9250, loss[loss=0.1378, simple_loss=0.2393, pruned_loss=0.01818, over 16888.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2582, pruned_loss=0.03483, over 3065513.23 frames. ], batch size: 96, lr: 2.67e-03, grad_scale: 16.0 2023-05-02 02:20:28,208 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252857.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:20:37,642 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 02:22:11,610 INFO [train.py:904] (0/8) Epoch 25, batch 9300, loss[loss=0.1598, simple_loss=0.2527, pruned_loss=0.03344, over 16736.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2569, pruned_loss=0.03446, over 3074950.66 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:22:12,587 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5485, 3.6815, 3.6905, 2.6682, 3.3747, 3.7688, 3.4829, 2.3203], device='cuda:0'), covar=tensor([0.0489, 0.0052, 0.0053, 0.0369, 0.0111, 0.0083, 0.0081, 0.0468], device='cuda:0'), in_proj_covar=tensor([0.0133, 0.0085, 0.0085, 0.0132, 0.0098, 0.0108, 0.0094, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 02:22:42,769 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 02:22:47,954 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=252918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:23:19,131 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.98 vs. limit=5.0 2023-05-02 02:23:45,079 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.307e+02 1.953e+02 2.173e+02 2.787e+02 5.118e+02, threshold=4.346e+02, percent-clipped=0.0 2023-05-02 02:23:47,292 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252948.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:23:55,324 INFO [train.py:904] (0/8) Epoch 25, batch 9350, loss[loss=0.1734, simple_loss=0.2623, pruned_loss=0.04224, over 16513.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2575, pruned_loss=0.03472, over 3095335.57 frames. ], batch size: 62, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:24:29,899 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=252969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:25:16,699 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=252993.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:25:36,661 INFO [train.py:904] (0/8) Epoch 25, batch 9400, loss[loss=0.1557, simple_loss=0.2366, pruned_loss=0.03741, over 12494.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2576, pruned_loss=0.03456, over 3088196.74 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:26:10,935 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7437, 4.0196, 2.9525, 2.2649, 2.4730, 2.5601, 4.3294, 3.3135], device='cuda:0'), covar=tensor([0.2967, 0.0568, 0.1906, 0.3027, 0.3080, 0.2175, 0.0341, 0.1486], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0264, 0.0303, 0.0311, 0.0292, 0.0263, 0.0292, 0.0336], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 02:26:19,764 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:27:05,789 INFO [optim.py:368] (0/8) 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 02:27:17,036 INFO [train.py:904] (0/8) Epoch 25, batch 9450, loss[loss=0.1825, simple_loss=0.2716, pruned_loss=0.04668, over 12855.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2597, pruned_loss=0.03505, over 3082385.10 frames. ], batch size: 250, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:27:19,830 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:27:57,660 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:31,274 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:28:54,671 INFO [train.py:904] (0/8) Epoch 25, batch 9500, loss[loss=0.141, simple_loss=0.2281, pruned_loss=0.02692, over 12973.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2582, pruned_loss=0.03443, over 3076478.63 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:29:33,160 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253121.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:06,155 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:30:26,029 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0283, 4.0467, 4.3148, 4.3102, 4.3240, 4.0953, 4.0857, 4.0964], device='cuda:0'), covar=tensor([0.0384, 0.0994, 0.0555, 0.0460, 0.0541, 0.0500, 0.0923, 0.0471], device='cuda:0'), in_proj_covar=tensor([0.0405, 0.0455, 0.0444, 0.0407, 0.0489, 0.0467, 0.0537, 0.0373], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 02:30:26,578 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.79 vs. limit=2.0 2023-05-02 02:30:26,755 INFO [optim.py:368] (0/8) 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 02:30:40,905 INFO [train.py:904] (0/8) Epoch 25, batch 9550, loss[loss=0.1554, simple_loss=0.2488, pruned_loss=0.03104, over 12525.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2584, pruned_loss=0.03455, over 3074464.09 frames. ], batch size: 250, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:22,177 INFO [train.py:904] (0/8) Epoch 25, batch 9600, loss[loss=0.163, simple_loss=0.2512, pruned_loss=0.03742, over 12210.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2596, pruned_loss=0.03518, over 3063285.61 frames. ], batch size: 247, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:32:42,589 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253213.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:33:55,686 INFO [optim.py:368] (0/8) 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 02:33:59,616 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253248.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:34:10,812 INFO [train.py:904] (0/8) Epoch 25, batch 9650, loss[loss=0.1706, simple_loss=0.2628, pruned_loss=0.03918, over 16864.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2617, pruned_loss=0.03558, over 3058098.15 frames. ], batch size: 116, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:34:52,654 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:35:17,424 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.64 vs. limit=5.0 2023-05-02 02:35:46,091 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253296.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:35:50,941 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4970, 4.6341, 4.7861, 4.6071, 4.7202, 5.1968, 4.7052, 4.3878], device='cuda:0'), covar=tensor([0.1287, 0.1846, 0.2148, 0.1797, 0.2293, 0.0849, 0.1440, 0.2192], device='cuda:0'), in_proj_covar=tensor([0.0396, 0.0581, 0.0648, 0.0480, 0.0633, 0.0670, 0.0503, 0.0638], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 02:36:00,303 INFO [train.py:904] (0/8) Epoch 25, batch 9700, loss[loss=0.1577, simple_loss=0.2466, pruned_loss=0.03438, over 12222.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.26, pruned_loss=0.03481, over 3061259.24 frames. ], batch size: 246, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:36:27,596 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:42,613 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:36:56,096 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253329.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:37:34,483 INFO [optim.py:368] (0/8) 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 02:37:37,172 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253349.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:37:44,447 INFO [train.py:904] (0/8) Epoch 25, batch 9750, loss[loss=0.154, simple_loss=0.2522, pruned_loss=0.02788, over 16776.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2588, pruned_loss=0.035, over 3063008.47 frames. ], batch size: 124, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:38:21,280 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:38:38,970 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253380.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:00,992 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253390.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:23,686 INFO [train.py:904] (0/8) Epoch 25, batch 9800, loss[loss=0.15, simple_loss=0.2569, pruned_loss=0.02161, over 16857.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2589, pruned_loss=0.03435, over 3064883.16 frames. ], batch size: 96, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:39:31,990 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1349, 3.6125, 3.5138, 2.3911, 3.2672, 3.6084, 3.3734, 2.0872], device='cuda:0'), covar=tensor([0.0623, 0.0053, 0.0067, 0.0436, 0.0128, 0.0097, 0.0100, 0.0536], device='cuda:0'), in_proj_covar=tensor([0.0132, 0.0084, 0.0085, 0.0132, 0.0098, 0.0107, 0.0094, 0.0128], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 02:39:53,017 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:39:58,176 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6108, 2.6527, 1.8428, 2.8012, 2.0294, 2.7832, 2.1463, 2.3565], device='cuda:0'), covar=tensor([0.0283, 0.0332, 0.1420, 0.0302, 0.0753, 0.0484, 0.1239, 0.0645], device='cuda:0'), in_proj_covar=tensor([0.0167, 0.0172, 0.0191, 0.0160, 0.0172, 0.0208, 0.0198, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0003, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 02:40:07,703 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 02:40:37,919 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:40:55,715 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.418e+02 1.999e+02 2.298e+02 2.745e+02 1.319e+03, threshold=4.597e+02, percent-clipped=1.0 2023-05-02 02:41:06,303 INFO [train.py:904] (0/8) Epoch 25, batch 9850, loss[loss=0.1617, simple_loss=0.2593, pruned_loss=0.03204, over 16267.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2598, pruned_loss=0.03394, over 3081755.16 frames. ], batch size: 166, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:41:45,672 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-05-02 02:42:03,938 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:00,298 INFO [train.py:904] (0/8) Epoch 25, batch 9900, loss[loss=0.1762, simple_loss=0.279, pruned_loss=0.03667, over 16566.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2602, pruned_loss=0.03402, over 3064140.40 frames. ], batch size: 62, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:43:25,466 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253513.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:43:41,987 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:44:46,260 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.406e+02 2.031e+02 2.358e+02 2.986e+02 6.654e+02, threshold=4.715e+02, percent-clipped=4.0 2023-05-02 02:44:59,719 INFO [train.py:904] (0/8) Epoch 25, batch 9950, loss[loss=0.1618, simple_loss=0.2581, pruned_loss=0.03272, over 16702.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2617, pruned_loss=0.03428, over 3049197.82 frames. ], batch size: 76, lr: 2.67e-03, grad_scale: 4.0 2023-05-02 02:45:18,705 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253561.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:45:34,735 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:46:13,998 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:46:30,499 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 02:47:02,500 INFO [train.py:904] (0/8) Epoch 25, batch 10000, loss[loss=0.1352, simple_loss=0.2286, pruned_loss=0.0209, over 17197.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2611, pruned_loss=0.0344, over 3057123.41 frames. ], batch size: 46, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:47:03,355 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4489, 2.0493, 1.7672, 1.8131, 2.2844, 1.9503, 1.8501, 2.3296], device='cuda:0'), covar=tensor([0.0191, 0.0453, 0.0580, 0.0512, 0.0292, 0.0439, 0.0241, 0.0316], device='cuda:0'), in_proj_covar=tensor([0.0212, 0.0236, 0.0225, 0.0225, 0.0235, 0.0235, 0.0228, 0.0230], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 02:47:53,034 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253628.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:36,798 INFO [optim.py:368] (0/8) 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 02:48:39,160 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:41,230 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=253650.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:48:46,083 INFO [train.py:904] (0/8) Epoch 25, batch 10050, loss[loss=0.1647, simple_loss=0.262, pruned_loss=0.03368, over 16659.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2615, pruned_loss=0.03427, over 3065322.90 frames. ], batch size: 134, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:48:47,290 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 02:49:35,982 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 02:49:48,464 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:11,407 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=253697.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:20,897 INFO [train.py:904] (0/8) Epoch 25, batch 10100, loss[loss=0.1564, simple_loss=0.2458, pruned_loss=0.03349, over 16977.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2618, pruned_loss=0.03437, over 3069233.81 frames. ], batch size: 109, lr: 2.67e-03, grad_scale: 8.0 2023-05-02 02:50:35,689 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=253711.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:50:55,054 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2604, 3.5776, 3.5736, 2.4286, 3.2261, 3.6278, 3.4029, 2.1272], device='cuda:0'), covar=tensor([0.0535, 0.0057, 0.0058, 0.0406, 0.0135, 0.0083, 0.0098, 0.0516], device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0083, 0.0084, 0.0131, 0.0097, 0.0106, 0.0093, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 02:51:21,524 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253736.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:51:33,768 INFO [optim.py:368] (0/8) 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 02:51:40,509 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-25.pt 2023-05-02 02:52:07,297 INFO [train.py:904] (0/8) Epoch 26, batch 0, loss[loss=0.2315, simple_loss=0.2938, pruned_loss=0.08459, over 16444.00 frames. ], tot_loss[loss=0.2315, simple_loss=0.2938, pruned_loss=0.08459, over 16444.00 frames. ], batch size: 75, lr: 2.61e-03, grad_scale: 8.0 2023-05-02 02:52:07,298 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 02:52:14,678 INFO [train.py:938] (0/8) Epoch 26, validation: loss=0.1437, simple_loss=0.2472, pruned_loss=0.02009, over 944034.00 frames. 2023-05-02 02:52:14,679 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 02:52:42,132 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4475, 3.9940, 4.5440, 2.2830, 4.7330, 4.8024, 3.4480, 3.6914], device='cuda:0'), covar=tensor([0.0588, 0.0254, 0.0194, 0.1122, 0.0063, 0.0138, 0.0418, 0.0367], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0107, 0.0094, 0.0135, 0.0080, 0.0123, 0.0125, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 02:52:44,971 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253775.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:53:23,656 INFO [train.py:904] (0/8) Epoch 26, batch 50, loss[loss=0.1608, simple_loss=0.2662, pruned_loss=0.02764, over 17253.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2702, pruned_loss=0.04752, over 747365.27 frames. ], batch size: 52, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:53:34,897 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5363, 4.3459, 4.5451, 4.6913, 4.8235, 4.4013, 4.7075, 4.7763], device='cuda:0'), covar=tensor([0.2195, 0.1809, 0.1844, 0.1053, 0.0776, 0.1170, 0.2029, 0.1792], device='cuda:0'), in_proj_covar=tensor([0.0625, 0.0763, 0.0879, 0.0779, 0.0597, 0.0616, 0.0647, 0.0750], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 02:53:47,711 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 02:54:28,066 INFO [optim.py:368] (0/8) 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 02:54:30,994 INFO [train.py:904] (0/8) Epoch 26, batch 100, loss[loss=0.1676, simple_loss=0.2625, pruned_loss=0.03638, over 16402.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2651, pruned_loss=0.04535, over 1317364.24 frames. ], batch size: 68, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:55:03,720 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:55:39,326 INFO [train.py:904] (0/8) Epoch 26, batch 150, loss[loss=0.1632, simple_loss=0.2584, pruned_loss=0.03405, over 17051.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2634, pruned_loss=0.04535, over 1764887.24 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:07,764 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=253923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:56:11,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7025, 3.7746, 2.3269, 3.9476, 2.9877, 3.8753, 2.4091, 3.0494], device='cuda:0'), covar=tensor([0.0276, 0.0401, 0.1582, 0.0434, 0.0754, 0.0821, 0.1445, 0.0689], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0175, 0.0194, 0.0165, 0.0176, 0.0213, 0.0202, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 02:56:46,628 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.138e+02 2.556e+02 3.055e+02 5.694e+02, threshold=5.112e+02, percent-clipped=1.0 2023-05-02 02:56:49,075 INFO [train.py:904] (0/8) Epoch 26, batch 200, loss[loss=0.1965, simple_loss=0.2697, pruned_loss=0.06166, over 16843.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2628, pruned_loss=0.04445, over 2103736.28 frames. ], batch size: 83, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:56:55,782 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9725, 4.8438, 4.8367, 4.4162, 4.5485, 4.8858, 4.7711, 4.5548], device='cuda:0'), covar=tensor([0.0626, 0.0911, 0.0388, 0.0399, 0.1016, 0.0553, 0.0421, 0.0814], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0438, 0.0342, 0.0345, 0.0344, 0.0394, 0.0235, 0.0409], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 02:57:33,688 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=253985.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:57:55,718 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-254000.pt 2023-05-02 02:58:03,910 INFO [train.py:904] (0/8) Epoch 26, batch 250, loss[loss=0.1457, simple_loss=0.2354, pruned_loss=0.02802, over 16830.00 frames. ], tot_loss[loss=0.174, simple_loss=0.26, pruned_loss=0.04401, over 2367286.13 frames. ], batch size: 42, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:58:07,784 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:42,517 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7242, 4.4956, 4.7527, 4.9028, 5.0644, 4.5519, 4.9935, 5.0428], device='cuda:0'), covar=tensor([0.1849, 0.1346, 0.1544, 0.0785, 0.0585, 0.1016, 0.1079, 0.0749], device='cuda:0'), in_proj_covar=tensor([0.0637, 0.0779, 0.0898, 0.0793, 0.0606, 0.0628, 0.0662, 0.0763], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 02:58:47,157 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:58:50,875 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254036.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:10,749 INFO [optim.py:368] (0/8) 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 02:59:14,077 INFO [train.py:904] (0/8) Epoch 26, batch 300, loss[loss=0.1455, simple_loss=0.2417, pruned_loss=0.02464, over 17269.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2576, pruned_loss=0.04241, over 2569069.57 frames. ], batch size: 52, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 02:59:45,488 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254075.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 02:59:57,063 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:00:23,587 INFO [train.py:904] (0/8) Epoch 26, batch 350, loss[loss=0.1449, simple_loss=0.2284, pruned_loss=0.0307, over 12557.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2556, pruned_loss=0.04062, over 2738666.81 frames. ], batch size: 246, lr: 2.61e-03, grad_scale: 1.0 2023-05-02 03:00:24,078 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1971, 2.2463, 2.7261, 3.1819, 2.9153, 3.6471, 2.1834, 3.5953], device='cuda:0'), covar=tensor([0.0259, 0.0618, 0.0390, 0.0348, 0.0380, 0.0195, 0.0648, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0195, 0.0183, 0.0185, 0.0202, 0.0159, 0.0198, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:00:51,987 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:01:30,800 INFO [optim.py:368] (0/8) 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 03:01:33,894 INFO [train.py:904] (0/8) Epoch 26, batch 400, loss[loss=0.1587, simple_loss=0.2534, pruned_loss=0.03202, over 16724.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2541, pruned_loss=0.04033, over 2873058.18 frames. ], batch size: 57, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:02:07,842 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:02:14,810 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4067, 5.3437, 5.2952, 4.7505, 4.8934, 5.3491, 5.2921, 4.9658], device='cuda:0'), covar=tensor([0.0604, 0.0534, 0.0332, 0.0397, 0.1184, 0.0448, 0.0284, 0.0793], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0446, 0.0349, 0.0352, 0.0350, 0.0403, 0.0239, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:02:44,716 INFO [train.py:904] (0/8) Epoch 26, batch 450, loss[loss=0.155, simple_loss=0.2387, pruned_loss=0.03559, over 15668.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2522, pruned_loss=0.03984, over 2973688.54 frames. ], batch size: 191, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:03:12,882 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:03:15,097 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.473e+02 2.133e+02 2.439e+02 2.881e+02 4.883e+02, threshold=4.879e+02, percent-clipped=0.0 2023-05-02 03:03:53,001 INFO [train.py:904] (0/8) Epoch 26, batch 500, loss[loss=0.1544, simple_loss=0.2504, pruned_loss=0.02917, over 17142.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2509, pruned_loss=0.03885, over 3056331.50 frames. ], batch size: 47, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:04:18,989 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254271.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:05:01,750 INFO [train.py:904] (0/8) Epoch 26, batch 550, loss[loss=0.1499, simple_loss=0.2491, pruned_loss=0.02531, over 17169.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2505, pruned_loss=0.03847, over 3115460.19 frames. ], batch size: 46, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:05:05,533 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=254306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:05:47,921 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 03:06:08,400 INFO [optim.py:368] (0/8) 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 03:06:11,650 INFO [train.py:904] (0/8) Epoch 26, batch 600, loss[loss=0.1599, simple_loss=0.2361, pruned_loss=0.04192, over 16917.00 frames. ], tot_loss[loss=0.164, simple_loss=0.25, pruned_loss=0.03897, over 3160292.31 frames. ], batch size: 109, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:06:13,036 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=254354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:07:21,745 INFO [train.py:904] (0/8) Epoch 26, batch 650, loss[loss=0.1559, simple_loss=0.2477, pruned_loss=0.032, over 17049.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2497, pruned_loss=0.03864, over 3198586.94 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:07:22,590 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 03:07:34,422 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.55 vs. limit=5.0 2023-05-02 03:08:28,771 INFO [optim.py:368] (0/8) 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 03:08:30,997 INFO [train.py:904] (0/8) Epoch 26, batch 700, loss[loss=0.1572, simple_loss=0.2568, pruned_loss=0.02884, over 17119.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2497, pruned_loss=0.03811, over 3232852.73 frames. ], batch size: 49, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:08:33,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8416, 2.8531, 2.5238, 2.7157, 3.2236, 2.8351, 3.4138, 3.2541], device='cuda:0'), covar=tensor([0.0176, 0.0467, 0.0535, 0.0498, 0.0312, 0.0442, 0.0303, 0.0344], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0246, 0.0234, 0.0234, 0.0245, 0.0244, 0.0241, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:09:41,503 INFO [train.py:904] (0/8) Epoch 26, batch 750, loss[loss=0.1546, simple_loss=0.2366, pruned_loss=0.03627, over 15963.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2499, pruned_loss=0.03807, over 3257363.14 frames. ], batch size: 35, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:10:27,490 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:10:46,997 INFO [optim.py:368] (0/8) 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 03:10:50,651 INFO [train.py:904] (0/8) Epoch 26, batch 800, loss[loss=0.1586, simple_loss=0.2384, pruned_loss=0.03943, over 16704.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2493, pruned_loss=0.03801, over 3278949.63 frames. ], batch size: 89, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:10:50,935 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8192, 3.7503, 3.8992, 3.6537, 3.8564, 4.2901, 3.9054, 3.5255], device='cuda:0'), covar=tensor([0.2131, 0.2524, 0.2735, 0.2782, 0.2793, 0.2038, 0.1677, 0.2820], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0622, 0.0691, 0.0512, 0.0677, 0.0713, 0.0535, 0.0681], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 03:11:19,898 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254574.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:51,481 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4267, 2.9737, 3.3217, 1.8588, 3.4027, 3.4471, 2.8107, 2.6725], device='cuda:0'), covar=tensor([0.0764, 0.0317, 0.0265, 0.1162, 0.0160, 0.0261, 0.0466, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0110, 0.0099, 0.0139, 0.0084, 0.0128, 0.0128, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 03:11:51,514 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:11:59,177 INFO [train.py:904] (0/8) Epoch 26, batch 850, loss[loss=0.1546, simple_loss=0.2472, pruned_loss=0.03101, over 17180.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2483, pruned_loss=0.03775, over 3285437.48 frames. ], batch size: 46, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:12:45,141 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:13:08,054 INFO [optim.py:368] (0/8) 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 03:13:10,211 INFO [train.py:904] (0/8) Epoch 26, batch 900, loss[loss=0.1751, simple_loss=0.2494, pruned_loss=0.05044, over 16822.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2475, pruned_loss=0.03732, over 3292825.41 frames. ], batch size: 102, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:13:44,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7992, 4.0480, 4.3161, 2.1443, 4.5984, 4.8087, 3.3674, 3.3532], device='cuda:0'), covar=tensor([0.1077, 0.0218, 0.0241, 0.1286, 0.0087, 0.0133, 0.0449, 0.0589], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0111, 0.0099, 0.0139, 0.0085, 0.0129, 0.0129, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 03:13:58,899 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3967, 3.5528, 4.0302, 2.2689, 3.1022, 2.4695, 3.7461, 3.7607], device='cuda:0'), covar=tensor([0.0288, 0.0977, 0.0499, 0.2104, 0.0923, 0.1021, 0.0669, 0.1080], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0168, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 03:14:19,306 INFO [train.py:904] (0/8) Epoch 26, batch 950, loss[loss=0.1388, simple_loss=0.2226, pruned_loss=0.02746, over 15815.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2476, pruned_loss=0.03764, over 3300616.08 frames. ], batch size: 35, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:14:34,073 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4426, 3.9736, 4.4171, 2.4603, 4.6479, 4.7294, 3.4405, 3.8158], device='cuda:0'), covar=tensor([0.0573, 0.0301, 0.0282, 0.1046, 0.0092, 0.0177, 0.0435, 0.0348], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0111, 0.0099, 0.0140, 0.0085, 0.0129, 0.0129, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 03:14:49,034 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.93 vs. limit=2.0 2023-05-02 03:15:24,095 INFO [optim.py:368] (0/8) 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 03:15:27,115 INFO [train.py:904] (0/8) Epoch 26, batch 1000, loss[loss=0.1478, simple_loss=0.2454, pruned_loss=0.02514, over 17007.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2467, pruned_loss=0.03732, over 3306861.51 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:15:31,172 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:16:28,396 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 26, batch 1050, loss[loss=0.1874, simple_loss=0.2616, pruned_loss=0.0566, over 16874.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2472, pruned_loss=0.03766, over 3315146.12 frames. ], batch size: 109, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:16:47,336 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8828, 4.9761, 5.3680, 5.3262, 5.3480, 5.0054, 4.9437, 4.7801], device='cuda:0'), covar=tensor([0.0377, 0.0580, 0.0394, 0.0397, 0.0490, 0.0410, 0.0953, 0.0487], device='cuda:0'), in_proj_covar=tensor([0.0429, 0.0481, 0.0469, 0.0430, 0.0515, 0.0494, 0.0569, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 03:16:55,071 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:42,912 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=254851.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:17:43,592 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.253e+02 2.171e+02 2.721e+02 3.139e+02 5.089e+02, threshold=5.442e+02, percent-clipped=1.0 2023-05-02 03:17:44,777 INFO [train.py:904] (0/8) Epoch 26, batch 1100, loss[loss=0.1562, simple_loss=0.2387, pruned_loss=0.03686, over 12113.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2463, pruned_loss=0.03747, over 3304387.77 frames. ], batch size: 246, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:17:53,555 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:17,257 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1756, 4.0120, 4.2404, 4.3512, 4.4139, 3.9764, 4.2537, 4.4078], device='cuda:0'), covar=tensor([0.1402, 0.1158, 0.1179, 0.0597, 0.0504, 0.1355, 0.1799, 0.0666], device='cuda:0'), in_proj_covar=tensor([0.0673, 0.0824, 0.0953, 0.0839, 0.0639, 0.0661, 0.0696, 0.0806], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:18:36,866 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254892.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:18:53,031 INFO [train.py:904] (0/8) Epoch 26, batch 1150, loss[loss=0.1439, simple_loss=0.2324, pruned_loss=0.02766, over 17228.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2459, pruned_loss=0.03694, over 3309772.31 frames. ], batch size: 44, lr: 2.61e-03, grad_scale: 2.0 2023-05-02 03:19:04,309 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=254912.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:29,220 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=254930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:19:46,562 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.61 vs. limit=5.0 2023-05-02 03:19:52,788 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9919, 5.0463, 5.4381, 5.4001, 5.4105, 5.0927, 5.0322, 4.9409], device='cuda:0'), covar=tensor([0.0377, 0.0626, 0.0417, 0.0460, 0.0507, 0.0437, 0.0957, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0482, 0.0470, 0.0430, 0.0516, 0.0494, 0.0569, 0.0393], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 03:19:59,615 INFO [optim.py:368] (0/8) 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 03:20:00,674 INFO [train.py:904] (0/8) Epoch 26, batch 1200, loss[loss=0.1764, simple_loss=0.2507, pruned_loss=0.05104, over 16834.00 frames. ], tot_loss[loss=0.1591, simple_loss=0.2451, pruned_loss=0.03655, over 3314281.27 frames. ], batch size: 116, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:10,676 INFO [train.py:904] (0/8) Epoch 26, batch 1250, loss[loss=0.1593, simple_loss=0.255, pruned_loss=0.03179, over 16759.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2452, pruned_loss=0.03623, over 3324114.89 frames. ], batch size: 57, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:21:26,783 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 03:22:19,977 INFO [optim.py:368] (0/8) 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 03:22:21,130 INFO [train.py:904] (0/8) Epoch 26, batch 1300, loss[loss=0.16, simple_loss=0.2424, pruned_loss=0.03883, over 16521.00 frames. ], tot_loss[loss=0.1592, simple_loss=0.2456, pruned_loss=0.03636, over 3313117.48 frames. ], batch size: 146, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:13,023 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 03:23:30,846 INFO [train.py:904] (0/8) Epoch 26, batch 1350, loss[loss=0.1536, simple_loss=0.2523, pruned_loss=0.0275, over 17057.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.2466, pruned_loss=0.03648, over 3316062.76 frames. ], batch size: 50, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:23:42,226 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:24:38,903 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 26, batch 1400, loss[loss=0.1686, simple_loss=0.2445, pruned_loss=0.04635, over 16836.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2458, pruned_loss=0.03659, over 3322166.83 frames. ], batch size: 109, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:24:41,097 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:12,150 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6136, 3.2244, 3.6316, 1.9216, 3.7246, 3.7912, 3.0845, 2.8152], device='cuda:0'), covar=tensor([0.0844, 0.0324, 0.0250, 0.1358, 0.0151, 0.0210, 0.0447, 0.0513], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0113, 0.0101, 0.0141, 0.0086, 0.0131, 0.0131, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 03:25:29,802 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 03:25:34,162 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:25:37,601 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4936, 3.0255, 3.3190, 1.9098, 3.4432, 3.4692, 2.8582, 2.6763], device='cuda:0'), covar=tensor([0.0748, 0.0309, 0.0294, 0.1196, 0.0150, 0.0233, 0.0522, 0.0488], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0113, 0.0101, 0.0141, 0.0086, 0.0131, 0.0131, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 03:25:43,454 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7567, 4.7642, 5.0797, 5.0934, 5.1328, 4.8406, 4.8002, 4.6279], device='cuda:0'), covar=tensor([0.0407, 0.0961, 0.0629, 0.0631, 0.0575, 0.0557, 0.0969, 0.0634], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0486, 0.0474, 0.0431, 0.0518, 0.0496, 0.0572, 0.0395], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 03:25:48,426 INFO [train.py:904] (0/8) Epoch 26, batch 1450, loss[loss=0.1589, simple_loss=0.2346, pruned_loss=0.0416, over 16768.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2448, pruned_loss=0.03641, over 3328671.01 frames. ], batch size: 83, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:25:54,152 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:26,914 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255230.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:40,697 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:26:48,812 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8514, 4.6704, 4.8653, 5.1097, 5.2472, 4.6659, 5.2439, 5.2797], device='cuda:0'), covar=tensor([0.2056, 0.1641, 0.2282, 0.0979, 0.0750, 0.1019, 0.0825, 0.0781], device='cuda:0'), in_proj_covar=tensor([0.0682, 0.0835, 0.0963, 0.0850, 0.0645, 0.0670, 0.0702, 0.0816], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:26:56,777 INFO [optim.py:368] (0/8) 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 03:26:57,924 INFO [train.py:904] (0/8) Epoch 26, batch 1500, loss[loss=0.1431, simple_loss=0.2335, pruned_loss=0.02633, over 16828.00 frames. ], tot_loss[loss=0.1588, simple_loss=0.2444, pruned_loss=0.03655, over 3326588.33 frames. ], batch size: 42, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:27:30,894 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:28:04,988 INFO [train.py:904] (0/8) Epoch 26, batch 1550, loss[loss=0.1621, simple_loss=0.2428, pruned_loss=0.04067, over 16726.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2462, pruned_loss=0.03723, over 3330592.17 frames. ], batch size: 134, lr: 2.61e-03, grad_scale: 4.0 2023-05-02 03:29:12,932 INFO [optim.py:368] (0/8) 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 03:29:14,111 INFO [train.py:904] (0/8) Epoch 26, batch 1600, loss[loss=0.1623, simple_loss=0.2611, pruned_loss=0.03174, over 17070.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2475, pruned_loss=0.03774, over 3330597.91 frames. ], batch size: 55, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:29:14,919 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 03:30:03,523 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:30:23,429 INFO [train.py:904] (0/8) Epoch 26, batch 1650, loss[loss=0.1524, simple_loss=0.2398, pruned_loss=0.03253, over 17161.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2492, pruned_loss=0.03848, over 3332189.77 frames. ], batch size: 46, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:30:35,818 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:28,977 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=255449.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:32,752 INFO [optim.py:368] (0/8) 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 03:31:34,045 INFO [train.py:904] (0/8) Epoch 26, batch 1700, loss[loss=0.1825, simple_loss=0.2631, pruned_loss=0.05089, over 16934.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2509, pruned_loss=0.03907, over 3335282.37 frames. ], batch size: 109, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:31:34,413 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:43,181 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255459.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:31:47,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7326, 4.7636, 5.1020, 5.0853, 5.1238, 4.8243, 4.7621, 4.6049], device='cuda:0'), covar=tensor([0.0365, 0.0608, 0.0418, 0.0425, 0.0470, 0.0416, 0.0921, 0.0568], device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0490, 0.0476, 0.0435, 0.0523, 0.0500, 0.0576, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 03:31:59,945 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 03:32:29,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9524, 3.0524, 3.2703, 2.0711, 2.8603, 2.2811, 3.4457, 3.4136], device='cuda:0'), covar=tensor([0.0227, 0.0928, 0.0652, 0.2022, 0.0896, 0.1067, 0.0555, 0.0850], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0169, 0.0170, 0.0157, 0.0148, 0.0132, 0.0146, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 03:32:37,274 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 03:32:40,141 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:32:42,743 INFO [train.py:904] (0/8) Epoch 26, batch 1750, loss[loss=0.1531, simple_loss=0.2486, pruned_loss=0.02882, over 17135.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2523, pruned_loss=0.03919, over 3332716.58 frames. ], batch size: 49, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:32:47,860 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=255507.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:32:54,906 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4079, 5.3368, 5.2964, 4.7843, 4.9226, 5.2950, 5.2628, 4.9393], device='cuda:0'), covar=tensor([0.0665, 0.0644, 0.0322, 0.0385, 0.1183, 0.0555, 0.0334, 0.0830], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0473, 0.0369, 0.0373, 0.0372, 0.0427, 0.0254, 0.0445], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 03:33:39,236 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 03:33:49,125 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 26, batch 1800, loss[loss=0.1511, simple_loss=0.2402, pruned_loss=0.03094, over 16982.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2532, pruned_loss=0.03938, over 3326064.88 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:33:54,881 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=255555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:34:58,649 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 03:34:59,049 INFO [train.py:904] (0/8) Epoch 26, batch 1850, loss[loss=0.168, simple_loss=0.2741, pruned_loss=0.03099, over 17141.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2542, pruned_loss=0.03971, over 3315543.93 frames. ], batch size: 48, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:36:05,027 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.481e+02 2.077e+02 2.353e+02 2.960e+02 6.152e+02, threshold=4.707e+02, percent-clipped=2.0 2023-05-02 03:36:06,158 INFO [train.py:904] (0/8) Epoch 26, batch 1900, loss[loss=0.1635, simple_loss=0.2634, pruned_loss=0.03187, over 17105.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2538, pruned_loss=0.03909, over 3310194.46 frames. ], batch size: 47, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:37:16,470 INFO [train.py:904] (0/8) Epoch 26, batch 1950, loss[loss=0.16, simple_loss=0.2587, pruned_loss=0.03064, over 17300.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2535, pruned_loss=0.03889, over 3303980.77 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:37:35,511 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 03:38:12,963 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=255744.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:38:24,269 INFO [optim.py:368] (0/8) 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 03:38:25,475 INFO [train.py:904] (0/8) Epoch 26, batch 2000, loss[loss=0.1542, simple_loss=0.2446, pruned_loss=0.03189, over 16849.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.253, pruned_loss=0.03857, over 3305156.39 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:38:54,412 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.79 vs. limit=5.0 2023-05-02 03:39:35,318 INFO [train.py:904] (0/8) Epoch 26, batch 2050, loss[loss=0.1564, simple_loss=0.2637, pruned_loss=0.0246, over 17305.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2537, pruned_loss=0.03924, over 3308148.74 frames. ], batch size: 52, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:39:37,555 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8512, 2.0314, 2.5636, 2.8425, 2.7237, 3.3514, 2.2992, 3.3206], device='cuda:0'), covar=tensor([0.0330, 0.0607, 0.0431, 0.0400, 0.0428, 0.0267, 0.0552, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0200, 0.0188, 0.0191, 0.0207, 0.0165, 0.0202, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:40:14,338 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6963, 3.7217, 2.8695, 2.2497, 2.4144, 2.3642, 3.8725, 3.2652], device='cuda:0'), covar=tensor([0.2700, 0.0651, 0.1727, 0.2989, 0.2706, 0.2157, 0.0492, 0.1561], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0275, 0.0312, 0.0323, 0.0305, 0.0274, 0.0303, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 03:40:44,483 INFO [optim.py:368] (0/8) 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 03:40:45,704 INFO [train.py:904] (0/8) Epoch 26, batch 2100, loss[loss=0.1788, simple_loss=0.2606, pruned_loss=0.04844, over 16476.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2544, pruned_loss=0.04009, over 3305818.98 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:41:01,314 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8430, 1.3668, 1.7662, 1.7013, 1.7302, 1.9491, 1.6252, 1.8545], device='cuda:0'), covar=tensor([0.0245, 0.0435, 0.0243, 0.0303, 0.0298, 0.0215, 0.0453, 0.0168], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0199, 0.0188, 0.0191, 0.0207, 0.0164, 0.0202, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:41:23,594 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9309, 2.7531, 2.8493, 2.1214, 2.6491, 2.1177, 2.6691, 2.9196], device='cuda:0'), covar=tensor([0.0329, 0.0899, 0.0547, 0.1779, 0.0884, 0.0917, 0.0661, 0.0761], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0170, 0.0171, 0.0158, 0.0149, 0.0133, 0.0147, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 03:41:54,472 INFO [train.py:904] (0/8) Epoch 26, batch 2150, loss[loss=0.1657, simple_loss=0.2417, pruned_loss=0.0448, over 16762.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2547, pruned_loss=0.04055, over 3295084.93 frames. ], batch size: 124, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:42:55,220 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5131, 3.6580, 3.8030, 2.6374, 3.5213, 3.9337, 3.6212, 2.2174], device='cuda:0'), covar=tensor([0.0567, 0.0189, 0.0070, 0.0439, 0.0129, 0.0103, 0.0107, 0.0540], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0089, 0.0090, 0.0137, 0.0103, 0.0114, 0.0098, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 03:43:04,641 INFO [optim.py:368] (0/8) 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 03:43:05,926 INFO [train.py:904] (0/8) Epoch 26, batch 2200, loss[loss=0.172, simple_loss=0.2596, pruned_loss=0.04218, over 16614.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2559, pruned_loss=0.04136, over 3295096.70 frames. ], batch size: 75, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:43:10,578 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=255956.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:43:19,111 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6295, 4.7154, 4.8598, 4.6745, 4.6963, 5.3204, 4.8679, 4.5749], device='cuda:0'), covar=tensor([0.1492, 0.2123, 0.2509, 0.2344, 0.2893, 0.1092, 0.1726, 0.2399], device='cuda:0'), in_proj_covar=tensor([0.0434, 0.0640, 0.0708, 0.0525, 0.0695, 0.0731, 0.0549, 0.0698], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 03:43:49,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2958, 5.9193, 5.9812, 5.7229, 5.8253, 6.3344, 5.8431, 5.5520], device='cuda:0'), covar=tensor([0.0835, 0.1833, 0.2315, 0.2152, 0.2298, 0.0977, 0.1580, 0.2331], device='cuda:0'), in_proj_covar=tensor([0.0434, 0.0639, 0.0707, 0.0525, 0.0693, 0.0730, 0.0549, 0.0697], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 03:44:07,677 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8783, 2.6315, 2.4687, 1.8998, 2.5631, 2.7235, 2.6137, 1.9797], device='cuda:0'), covar=tensor([0.0476, 0.0144, 0.0112, 0.0426, 0.0168, 0.0153, 0.0146, 0.0406], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0089, 0.0090, 0.0138, 0.0103, 0.0114, 0.0099, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 03:44:12,777 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-256000.pt 2023-05-02 03:44:19,345 INFO [train.py:904] (0/8) Epoch 26, batch 2250, loss[loss=0.1747, simple_loss=0.2658, pruned_loss=0.04181, over 16660.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2561, pruned_loss=0.04124, over 3301599.46 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:44:31,556 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6649, 1.8633, 2.3959, 2.5548, 2.6307, 2.6748, 1.9281, 2.8323], device='cuda:0'), covar=tensor([0.0234, 0.0544, 0.0341, 0.0332, 0.0329, 0.0322, 0.0599, 0.0219], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0199, 0.0188, 0.0191, 0.0206, 0.0164, 0.0202, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:44:39,407 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:09,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4842, 3.5116, 3.6650, 2.6302, 3.4288, 3.8197, 3.5606, 2.2855], device='cuda:0'), covar=tensor([0.0549, 0.0166, 0.0063, 0.0413, 0.0118, 0.0086, 0.0097, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0089, 0.0090, 0.0138, 0.0103, 0.0114, 0.0098, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 03:45:09,594 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256038.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:17,781 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:45:27,440 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.334e+02 2.041e+02 2.398e+02 2.796e+02 4.429e+02, threshold=4.795e+02, percent-clipped=0.0 2023-05-02 03:45:29,164 INFO [train.py:904] (0/8) Epoch 26, batch 2300, loss[loss=0.1456, simple_loss=0.2372, pruned_loss=0.02697, over 17240.00 frames. ], tot_loss[loss=0.169, simple_loss=0.256, pruned_loss=0.04101, over 3288922.38 frames. ], batch size: 45, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:45:55,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8957, 5.2674, 5.0328, 5.0332, 4.7370, 4.7043, 4.7088, 5.3676], device='cuda:0'), covar=tensor([0.1357, 0.0922, 0.1069, 0.0949, 0.0918, 0.1117, 0.1259, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0718, 0.0871, 0.0712, 0.0672, 0.0554, 0.0553, 0.0729, 0.0679], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:46:09,878 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:14,126 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256084.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:23,798 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256092.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:34,570 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:46:39,732 INFO [train.py:904] (0/8) Epoch 26, batch 2350, loss[loss=0.1833, simple_loss=0.2739, pruned_loss=0.04638, over 16487.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2558, pruned_loss=0.04095, over 3300626.36 frames. ], batch size: 68, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:47:19,451 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:32,061 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:36,995 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:47:45,353 INFO [optim.py:368] (0/8) 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 03:47:46,521 INFO [train.py:904] (0/8) Epoch 26, batch 2400, loss[loss=0.1777, simple_loss=0.2569, pruned_loss=0.04924, over 16482.00 frames. ], tot_loss[loss=0.1691, simple_loss=0.2567, pruned_loss=0.04079, over 3306284.16 frames. ], batch size: 75, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:47:55,519 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 03:48:11,622 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 03:48:42,773 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256193.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 03:48:55,680 INFO [train.py:904] (0/8) Epoch 26, batch 2450, loss[loss=0.1778, simple_loss=0.2767, pruned_loss=0.0395, over 16711.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2581, pruned_loss=0.0407, over 3315460.11 frames. ], batch size: 57, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:01,729 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.511e+02 2.193e+02 2.630e+02 3.165e+02 5.747e+02, threshold=5.261e+02, percent-clipped=2.0 2023-05-02 03:50:03,666 INFO [train.py:904] (0/8) Epoch 26, batch 2500, loss[loss=0.172, simple_loss=0.27, pruned_loss=0.03699, over 16616.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2577, pruned_loss=0.04057, over 3317305.96 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:50:48,308 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256285.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:50:49,339 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6625, 4.7445, 4.8985, 4.7495, 4.7462, 5.3366, 4.8530, 4.5849], device='cuda:0'), covar=tensor([0.1453, 0.2015, 0.2700, 0.2126, 0.2900, 0.1058, 0.1670, 0.2535], device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0639, 0.0707, 0.0523, 0.0694, 0.0728, 0.0548, 0.0699], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 03:51:13,231 INFO [train.py:904] (0/8) Epoch 26, batch 2550, loss[loss=0.1807, simple_loss=0.2729, pruned_loss=0.04423, over 17088.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2578, pruned_loss=0.04078, over 3315804.44 frames. ], batch size: 55, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:51:26,227 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256312.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:52:14,136 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 03:52:23,674 INFO [train.py:904] (0/8) Epoch 26, batch 2600, loss[loss=0.1472, simple_loss=0.2384, pruned_loss=0.028, over 16806.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2574, pruned_loss=0.04022, over 3316885.36 frames. ], batch size: 42, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:52:50,728 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:20,841 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256394.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:53:33,352 INFO [train.py:904] (0/8) Epoch 26, batch 2650, loss[loss=0.1909, simple_loss=0.2677, pruned_loss=0.05711, over 12517.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.258, pruned_loss=0.0399, over 3315039.27 frames. ], batch size: 246, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:54:14,476 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256433.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:20,224 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:24,159 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256440.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:54:24,692 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 03:54:40,366 INFO [optim.py:368] (0/8) 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 03:54:41,416 INFO [train.py:904] (0/8) Epoch 26, batch 2700, loss[loss=0.1721, simple_loss=0.2744, pruned_loss=0.03492, over 17058.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2575, pruned_loss=0.03943, over 3316086.79 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:55:06,518 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:55:28,876 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256488.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 03:55:48,618 INFO [train.py:904] (0/8) Epoch 26, batch 2750, loss[loss=0.1642, simple_loss=0.2695, pruned_loss=0.02939, over 17045.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2578, pruned_loss=0.03884, over 3323463.59 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:56:30,011 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256532.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:56:33,721 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3032, 2.4157, 2.4326, 4.1105, 2.3293, 2.7173, 2.4876, 2.5306], device='cuda:0'), covar=tensor([0.1471, 0.3608, 0.3070, 0.0636, 0.4072, 0.2635, 0.3638, 0.3497], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0471, 0.0386, 0.0339, 0.0447, 0.0538, 0.0442, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:56:56,720 INFO [optim.py:368] (0/8) 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 03:56:58,723 INFO [train.py:904] (0/8) Epoch 26, batch 2800, loss[loss=0.1743, simple_loss=0.2535, pruned_loss=0.04753, over 16736.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2569, pruned_loss=0.03836, over 3333920.28 frames. ], batch size: 124, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:57:02,638 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9343, 4.5510, 4.4970, 3.2181, 3.8556, 4.4972, 4.0445, 2.6128], device='cuda:0'), covar=tensor([0.0544, 0.0088, 0.0059, 0.0426, 0.0157, 0.0120, 0.0105, 0.0541], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0136, 0.0101, 0.0113, 0.0098, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 03:58:07,644 INFO [train.py:904] (0/8) Epoch 26, batch 2850, loss[loss=0.2023, simple_loss=0.2734, pruned_loss=0.0656, over 15979.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2557, pruned_loss=0.03844, over 3331682.65 frames. ], batch size: 35, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:58:21,205 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256612.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:00,646 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256641.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:09,188 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8584, 2.0337, 2.5988, 2.8950, 2.7590, 3.3949, 2.3651, 3.3975], device='cuda:0'), covar=tensor([0.0300, 0.0597, 0.0385, 0.0369, 0.0386, 0.0217, 0.0503, 0.0196], device='cuda:0'), in_proj_covar=tensor([0.0197, 0.0199, 0.0188, 0.0191, 0.0206, 0.0166, 0.0203, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 03:59:15,040 INFO [optim.py:368] (0/8) 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 03:59:16,871 INFO [train.py:904] (0/8) Epoch 26, batch 2900, loss[loss=0.151, simple_loss=0.2393, pruned_loss=0.03138, over 17113.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2552, pruned_loss=0.03884, over 3336673.45 frames. ], batch size: 47, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 03:59:27,184 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256660.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 03:59:29,157 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2432, 5.6294, 5.3579, 5.4289, 5.1103, 5.0422, 5.1233, 5.7351], device='cuda:0'), covar=tensor([0.1386, 0.0887, 0.1270, 0.0900, 0.0836, 0.0875, 0.1229, 0.0914], device='cuda:0'), in_proj_covar=tensor([0.0716, 0.0870, 0.0713, 0.0671, 0.0553, 0.0551, 0.0730, 0.0677], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:00:13,252 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 26, batch 2950, loss[loss=0.1596, simple_loss=0.2615, pruned_loss=0.02883, over 17052.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2546, pruned_loss=0.03922, over 3344273.43 frames. ], batch size: 50, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:00:33,220 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8216, 2.8931, 2.8254, 5.0120, 4.0497, 4.4185, 1.7230, 3.0761], device='cuda:0'), covar=tensor([0.1365, 0.0798, 0.1169, 0.0212, 0.0175, 0.0368, 0.1631, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0180, 0.0199, 0.0200, 0.0206, 0.0219, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 04:01:01,189 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256728.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:14,636 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:18,084 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256740.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:20,858 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256742.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:01:35,077 INFO [optim.py:368] (0/8) 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 04:01:35,098 INFO [train.py:904] (0/8) Epoch 26, batch 3000, loss[loss=0.1571, simple_loss=0.2552, pruned_loss=0.02949, over 17138.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2549, pruned_loss=0.03999, over 3341612.84 frames. ], batch size: 47, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:01:35,098 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 04:01:44,001 INFO [train.py:938] (0/8) Epoch 26, validation: loss=0.1339, simple_loss=0.2392, pruned_loss=0.01435, over 944034.00 frames. 2023-05-02 04:01:44,002 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 04:02:27,417 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,515 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:31,591 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256788.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:02:35,209 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5132, 4.8384, 4.6506, 4.6515, 4.3892, 4.3184, 4.3378, 4.8976], device='cuda:0'), covar=tensor([0.1264, 0.0902, 0.0989, 0.0841, 0.0815, 0.1357, 0.1216, 0.0887], device='cuda:0'), in_proj_covar=tensor([0.0714, 0.0868, 0.0710, 0.0669, 0.0553, 0.0549, 0.0728, 0.0677], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:02:53,053 INFO [train.py:904] (0/8) Epoch 26, batch 3050, loss[loss=0.1973, simple_loss=0.2649, pruned_loss=0.06488, over 16463.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2549, pruned_loss=0.03984, over 3337942.03 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:03:16,762 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6330, 5.0502, 4.5447, 4.9013, 4.6071, 4.5170, 4.5883, 5.1159], device='cuda:0'), covar=tensor([0.2553, 0.1909, 0.2908, 0.1757, 0.1786, 0.1961, 0.2709, 0.2076], device='cuda:0'), in_proj_covar=tensor([0.0716, 0.0870, 0.0712, 0.0671, 0.0554, 0.0550, 0.0730, 0.0678], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:03:27,120 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=256827.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:03:27,321 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256827.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:03:38,260 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:04:02,823 INFO [optim.py:368] (0/8) 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 04:04:02,843 INFO [train.py:904] (0/8) Epoch 26, batch 3100, loss[loss=0.1583, simple_loss=0.2382, pruned_loss=0.03918, over 16505.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2548, pruned_loss=0.0398, over 3341767.97 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:04:47,252 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=256883.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:04:53,604 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256888.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:05:13,365 INFO [train.py:904] (0/8) Epoch 26, batch 3150, loss[loss=0.1734, simple_loss=0.2566, pruned_loss=0.04511, over 16759.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2543, pruned_loss=0.03966, over 3334529.54 frames. ], batch size: 83, lr: 2.60e-03, grad_scale: 4.0 2023-05-02 04:05:13,768 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1205, 5.1222, 5.5583, 5.5249, 5.5641, 5.2236, 5.1487, 5.0350], device='cuda:0'), covar=tensor([0.0347, 0.0624, 0.0379, 0.0471, 0.0462, 0.0375, 0.0983, 0.0457], device='cuda:0'), in_proj_covar=tensor([0.0446, 0.0502, 0.0486, 0.0445, 0.0534, 0.0513, 0.0594, 0.0411], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 04:05:50,901 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8043, 2.4318, 2.4424, 3.7622, 2.8758, 3.8362, 1.5843, 2.7634], device='cuda:0'), covar=tensor([0.1453, 0.0841, 0.1295, 0.0235, 0.0159, 0.0405, 0.1787, 0.0931], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0180, 0.0199, 0.0201, 0.0207, 0.0220, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 04:06:05,120 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=256941.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:06:09,352 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=256944.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:06:21,404 INFO [optim.py:368] (0/8) 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 04:06:21,419 INFO [train.py:904] (0/8) Epoch 26, batch 3200, loss[loss=0.1772, simple_loss=0.2438, pruned_loss=0.05531, over 16991.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2531, pruned_loss=0.03904, over 3336992.81 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:06:30,603 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0249, 4.4283, 3.2001, 2.4744, 2.8367, 2.8136, 4.8636, 3.7124], device='cuda:0'), covar=tensor([0.2715, 0.0646, 0.1769, 0.2948, 0.2930, 0.1962, 0.0368, 0.1488], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0277, 0.0313, 0.0325, 0.0307, 0.0275, 0.0304, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 04:07:11,957 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=256989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:07:29,908 INFO [train.py:904] (0/8) Epoch 26, batch 3250, loss[loss=0.1342, simple_loss=0.2173, pruned_loss=0.02553, over 16977.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2527, pruned_loss=0.03922, over 3334999.01 frames. ], batch size: 41, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:07:43,756 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:07:43,892 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5456, 2.4864, 2.6102, 4.4700, 2.4770, 2.8675, 2.5491, 2.6771], device='cuda:0'), covar=tensor([0.1450, 0.3831, 0.3188, 0.0555, 0.4278, 0.2564, 0.3747, 0.3615], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0471, 0.0385, 0.0339, 0.0446, 0.0538, 0.0441, 0.0550], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:08:03,355 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:08:04,938 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4816, 2.4131, 2.4717, 4.3575, 2.4000, 2.7954, 2.4803, 2.5968], device='cuda:0'), covar=tensor([0.1335, 0.3679, 0.3170, 0.0542, 0.4118, 0.2656, 0.3674, 0.3742], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0471, 0.0385, 0.0339, 0.0446, 0.0538, 0.0441, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:08:38,572 INFO [optim.py:368] (0/8) 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 04:08:38,593 INFO [train.py:904] (0/8) Epoch 26, batch 3300, loss[loss=0.1611, simple_loss=0.2486, pruned_loss=0.03681, over 16809.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2534, pruned_loss=0.03954, over 3336889.06 frames. ], batch size: 96, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:09:07,494 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257074.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:09,487 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257076.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:09:48,093 INFO [train.py:904] (0/8) Epoch 26, batch 3350, loss[loss=0.1898, simple_loss=0.2722, pruned_loss=0.05363, over 15567.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2542, pruned_loss=0.03995, over 3319375.20 frames. ], batch size: 190, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:10:20,797 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:10:26,944 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 04:10:46,910 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6025, 3.5831, 4.2029, 2.2405, 3.2451, 2.6095, 4.1006, 3.8311], device='cuda:0'), covar=tensor([0.0299, 0.1085, 0.0475, 0.2143, 0.0859, 0.1041, 0.0592, 0.1134], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0157, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 04:10:56,577 INFO [optim.py:368] (0/8) 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 04:10:56,611 INFO [train.py:904] (0/8) Epoch 26, batch 3400, loss[loss=0.1797, simple_loss=0.261, pruned_loss=0.04917, over 11843.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2535, pruned_loss=0.03942, over 3324589.57 frames. ], batch size: 246, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:11:13,594 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257165.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:27,521 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257175.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:11:39,238 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257183.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 04:12:06,429 INFO [train.py:904] (0/8) Epoch 26, batch 3450, loss[loss=0.1834, simple_loss=0.2679, pruned_loss=0.04944, over 16743.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2521, pruned_loss=0.03871, over 3329704.93 frames. ], batch size: 62, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:12:33,394 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:12:39,881 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257239.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:12:59,752 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:13:16,459 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 26, batch 3500, loss[loss=0.1894, simple_loss=0.2702, pruned_loss=0.05434, over 16494.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2511, pruned_loss=0.03865, over 3328709.68 frames. ], batch size: 146, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:13:58,938 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:25,683 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:14:26,377 INFO [train.py:904] (0/8) Epoch 26, batch 3550, loss[loss=0.1548, simple_loss=0.2425, pruned_loss=0.03359, over 16503.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2504, pruned_loss=0.03823, over 3329750.59 frames. ], batch size: 75, lr: 2.60e-03, grad_scale: 8.0 2023-05-02 04:15:13,183 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7045, 6.1236, 5.8208, 5.9228, 5.4833, 5.5421, 5.4528, 6.2854], device='cuda:0'), covar=tensor([0.1453, 0.1018, 0.1041, 0.0840, 0.0909, 0.0672, 0.1444, 0.0820], device='cuda:0'), in_proj_covar=tensor([0.0725, 0.0883, 0.0720, 0.0681, 0.0559, 0.0558, 0.0739, 0.0687], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:15:34,875 INFO [train.py:904] (0/8) Epoch 26, batch 3600, loss[loss=0.1593, simple_loss=0.2503, pruned_loss=0.03416, over 17073.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2501, pruned_loss=0.03819, over 3333300.57 frames. ], batch size: 55, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:15:35,980 INFO [optim.py:368] (0/8) 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 04:15:58,495 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257369.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:16:47,815 INFO [train.py:904] (0/8) Epoch 26, batch 3650, loss[loss=0.1852, simple_loss=0.2545, pruned_loss=0.05797, over 16518.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2498, pruned_loss=0.03814, over 3323019.27 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:58,062 INFO [train.py:904] (0/8) Epoch 26, batch 3700, loss[loss=0.1702, simple_loss=0.2441, pruned_loss=0.04817, over 11221.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2482, pruned_loss=0.03987, over 3298876.28 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:17:59,893 INFO [optim.py:368] (0/8) 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 04:18:18,537 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:18:41,373 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257483.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:19:08,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4014, 2.9311, 2.7183, 2.3361, 2.2844, 2.3539, 2.9249, 2.8260], device='cuda:0'), covar=tensor([0.2425, 0.0598, 0.1488, 0.2155, 0.2342, 0.2128, 0.0497, 0.1292], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0274, 0.0310, 0.0322, 0.0305, 0.0273, 0.0302, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 04:19:09,138 INFO [train.py:904] (0/8) Epoch 26, batch 3750, loss[loss=0.1827, simple_loss=0.2663, pruned_loss=0.04957, over 11443.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2483, pruned_loss=0.04087, over 3279555.97 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:19:34,006 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257521.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:19:44,810 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257528.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:19:48,977 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257531.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:20:01,214 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257539.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:20:20,034 INFO [train.py:904] (0/8) Epoch 26, batch 3800, loss[loss=0.1757, simple_loss=0.2507, pruned_loss=0.05042, over 16786.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2502, pruned_loss=0.04223, over 3267197.20 frames. ], batch size: 124, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:20:22,164 INFO [optim.py:368] (0/8) 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 04:20:55,724 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:20:55,824 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257578.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:21:09,655 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257587.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:21:22,626 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:21:30,549 INFO [train.py:904] (0/8) Epoch 26, batch 3850, loss[loss=0.1757, simple_loss=0.2474, pruned_loss=0.05204, over 16348.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2507, pruned_loss=0.04331, over 3269336.48 frames. ], batch size: 165, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:15,678 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8234, 3.7605, 3.9050, 3.6326, 3.8504, 4.2918, 3.9118, 3.5191], device='cuda:0'), covar=tensor([0.2170, 0.2328, 0.2376, 0.2418, 0.2708, 0.1673, 0.1555, 0.2452], device='cuda:0'), in_proj_covar=tensor([0.0433, 0.0638, 0.0704, 0.0522, 0.0692, 0.0728, 0.0546, 0.0698], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 04:22:20,359 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257639.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:22:32,923 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9913, 5.4207, 5.5887, 5.2058, 5.2876, 5.9503, 5.4391, 5.1579], device='cuda:0'), covar=tensor([0.0974, 0.1657, 0.1710, 0.1991, 0.2648, 0.0810, 0.1324, 0.2176], device='cuda:0'), in_proj_covar=tensor([0.0432, 0.0637, 0.0702, 0.0521, 0.0691, 0.0726, 0.0545, 0.0696], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 04:22:41,241 INFO [train.py:904] (0/8) Epoch 26, batch 3900, loss[loss=0.1667, simple_loss=0.2419, pruned_loss=0.04578, over 16475.00 frames. ], tot_loss[loss=0.1692, simple_loss=0.2506, pruned_loss=0.04389, over 3273568.39 frames. ], batch size: 146, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:22:42,474 INFO [optim.py:368] (0/8) 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 04:23:04,497 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:23:04,619 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2818, 3.2819, 2.1783, 3.4657, 2.5911, 3.4964, 2.3266, 2.6873], device='cuda:0'), covar=tensor([0.0297, 0.0434, 0.1404, 0.0293, 0.0817, 0.0635, 0.1281, 0.0762], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0183, 0.0199, 0.0176, 0.0181, 0.0223, 0.0206, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 04:23:51,422 INFO [train.py:904] (0/8) Epoch 26, batch 3950, loss[loss=0.1723, simple_loss=0.2567, pruned_loss=0.04391, over 15589.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2498, pruned_loss=0.0444, over 3280929.16 frames. ], batch size: 191, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:24:12,554 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:25:02,913 INFO [train.py:904] (0/8) Epoch 26, batch 4000, loss[loss=0.1899, simple_loss=0.2636, pruned_loss=0.05809, over 16673.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.25, pruned_loss=0.04474, over 3271225.05 frames. ], batch size: 134, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:25:03,990 INFO [optim.py:368] (0/8) 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 04:26:01,749 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.10 vs. limit=2.0 2023-05-02 04:26:10,143 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5959, 3.6888, 2.2377, 4.2641, 2.8278, 4.2982, 2.5301, 2.9641], device='cuda:0'), covar=tensor([0.0318, 0.0375, 0.1832, 0.0143, 0.0911, 0.0295, 0.1590, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0182, 0.0199, 0.0175, 0.0181, 0.0223, 0.0206, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 04:26:13,228 INFO [train.py:904] (0/8) Epoch 26, batch 4050, loss[loss=0.1567, simple_loss=0.241, pruned_loss=0.03615, over 16393.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2507, pruned_loss=0.04416, over 3278051.05 frames. ], batch size: 75, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:26:39,473 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:26:42,425 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257823.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:27:25,367 INFO [train.py:904] (0/8) Epoch 26, batch 4100, loss[loss=0.2112, simple_loss=0.2905, pruned_loss=0.06595, over 12084.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2521, pruned_loss=0.04357, over 3260502.21 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:27:26,544 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.240e+02 1.750e+02 2.045e+02 2.393e+02 4.321e+02, threshold=4.090e+02, percent-clipped=1.0 2023-05-02 04:27:35,108 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257859.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:27:49,113 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257869.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:02,805 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:10,433 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257883.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:30,699 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=257897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:28:39,718 INFO [train.py:904] (0/8) Epoch 26, batch 4150, loss[loss=0.2291, simple_loss=0.3068, pruned_loss=0.07573, over 11416.00 frames. ], tot_loss[loss=0.1745, simple_loss=0.2584, pruned_loss=0.04531, over 3230380.96 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:07,262 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:15,914 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:28,838 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=257934.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 04:29:44,249 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=257944.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 04:29:45,301 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=257945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:52,850 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=257950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:29:56,815 INFO [train.py:904] (0/8) Epoch 26, batch 4200, loss[loss=0.1892, simple_loss=0.2809, pruned_loss=0.04878, over 16651.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2653, pruned_loss=0.04663, over 3224108.29 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:29:58,476 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.624e+02 2.083e+02 2.573e+02 2.968e+02 5.903e+02, threshold=5.147e+02, percent-clipped=5.0 2023-05-02 04:30:25,253 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3483, 4.3950, 4.6829, 4.6487, 4.7078, 4.4005, 4.3248, 4.2680], device='cuda:0'), covar=tensor([0.0336, 0.0588, 0.0448, 0.0452, 0.0444, 0.0435, 0.1248, 0.0608], device='cuda:0'), in_proj_covar=tensor([0.0433, 0.0488, 0.0474, 0.0434, 0.0522, 0.0499, 0.0578, 0.0401], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 04:30:52,164 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7969, 4.9794, 5.1928, 4.9037, 5.0750, 5.6061, 4.9478, 4.6391], device='cuda:0'), covar=tensor([0.0864, 0.1727, 0.1514, 0.1643, 0.1910, 0.0693, 0.1311, 0.2325], device='cuda:0'), in_proj_covar=tensor([0.0427, 0.0630, 0.0694, 0.0514, 0.0680, 0.0719, 0.0539, 0.0686], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 04:31:08,529 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-258000.pt 2023-05-02 04:31:15,320 INFO [train.py:904] (0/8) Epoch 26, batch 4250, loss[loss=0.1821, simple_loss=0.272, pruned_loss=0.04615, over 17201.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2677, pruned_loss=0.04577, over 3218332.91 frames. ], batch size: 44, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:31:29,177 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 26, batch 4300, loss[loss=0.1771, simple_loss=0.2697, pruned_loss=0.04221, over 16708.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2692, pruned_loss=0.04515, over 3217692.33 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:32:31,423 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.134e+02 2.518e+02 2.964e+02 4.474e+02, threshold=5.035e+02, percent-clipped=0.0 2023-05-02 04:33:45,792 INFO [train.py:904] (0/8) Epoch 26, batch 4350, loss[loss=0.1875, simple_loss=0.276, pruned_loss=0.04946, over 16938.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2727, pruned_loss=0.04643, over 3212745.33 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:34:15,492 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3936, 5.6741, 5.4365, 5.4667, 5.1387, 5.0130, 5.0457, 5.7704], device='cuda:0'), covar=tensor([0.1058, 0.0718, 0.0983, 0.0791, 0.0756, 0.0774, 0.1110, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0707, 0.0858, 0.0701, 0.0662, 0.0546, 0.0545, 0.0721, 0.0670], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:34:15,534 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258123.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:34:58,990 INFO [train.py:904] (0/8) Epoch 26, batch 4400, loss[loss=0.1875, simple_loss=0.2806, pruned_loss=0.04725, over 16784.00 frames. ], tot_loss[loss=0.1858, simple_loss=0.2753, pruned_loss=0.04809, over 3199997.49 frames. ], batch size: 83, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:35:00,107 INFO [optim.py:368] (0/8) 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 04:35:00,903 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.92 vs. limit=2.0 2023-05-02 04:35:06,249 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2345, 4.3012, 4.5390, 4.4916, 4.5343, 4.3114, 4.2788, 4.1835], device='cuda:0'), covar=tensor([0.0329, 0.0544, 0.0377, 0.0431, 0.0432, 0.0417, 0.0771, 0.0565], device='cuda:0'), in_proj_covar=tensor([0.0429, 0.0482, 0.0468, 0.0429, 0.0516, 0.0494, 0.0571, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 04:35:13,115 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5367, 3.6525, 2.6627, 2.2100, 2.4870, 2.4120, 3.9171, 3.3572], device='cuda:0'), covar=tensor([0.2887, 0.0622, 0.1910, 0.2557, 0.2539, 0.2076, 0.0412, 0.1150], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0274, 0.0311, 0.0322, 0.0306, 0.0271, 0.0302, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 04:35:25,198 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258171.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:35:25,247 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1090, 5.7107, 5.8890, 5.5096, 5.6660, 6.1950, 5.6081, 5.3051], device='cuda:0'), covar=tensor([0.0804, 0.1506, 0.1948, 0.1635, 0.2011, 0.0747, 0.1303, 0.2169], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0624, 0.0686, 0.0508, 0.0673, 0.0711, 0.0533, 0.0679], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 04:36:06,260 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0046, 2.9653, 1.9473, 3.2680, 2.2877, 3.3263, 2.1355, 2.5307], device='cuda:0'), covar=tensor([0.0351, 0.0416, 0.1710, 0.0187, 0.0882, 0.0458, 0.1488, 0.0746], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0181, 0.0196, 0.0172, 0.0179, 0.0220, 0.0204, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 04:36:11,040 INFO [train.py:904] (0/8) Epoch 26, batch 4450, loss[loss=0.1969, simple_loss=0.2784, pruned_loss=0.05767, over 12050.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2793, pruned_loss=0.0497, over 3204442.72 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:36:28,132 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:36:32,988 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.47 vs. limit=5.0 2023-05-02 04:36:46,369 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 04:36:55,356 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258234.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:37:03,330 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258239.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:37:22,670 INFO [train.py:904] (0/8) Epoch 26, batch 4500, loss[loss=0.199, simple_loss=0.2847, pruned_loss=0.05666, over 15404.00 frames. ], tot_loss[loss=0.1905, simple_loss=0.2801, pruned_loss=0.05044, over 3224403.58 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:37:23,850 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.439e+02 1.900e+02 2.188e+02 2.602e+02 4.588e+02, threshold=4.376e+02, percent-clipped=0.0 2023-05-02 04:38:05,643 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258282.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 04:38:35,281 INFO [train.py:904] (0/8) Epoch 26, batch 4550, loss[loss=0.2013, simple_loss=0.2865, pruned_loss=0.05806, over 16698.00 frames. ], tot_loss[loss=0.1918, simple_loss=0.2806, pruned_loss=0.05149, over 3227654.54 frames. ], batch size: 57, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:38:39,775 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258306.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:39:48,617 INFO [train.py:904] (0/8) Epoch 26, batch 4600, loss[loss=0.1842, simple_loss=0.2793, pruned_loss=0.04461, over 16575.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2817, pruned_loss=0.05196, over 3218117.39 frames. ], batch size: 62, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:39:50,256 INFO [optim.py:368] (0/8) 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 04:41:03,114 INFO [train.py:904] (0/8) Epoch 26, batch 4650, loss[loss=0.2418, simple_loss=0.312, pruned_loss=0.08584, over 12004.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2809, pruned_loss=0.05249, over 3202768.50 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:41:07,925 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7736, 1.9782, 2.1330, 3.2050, 1.9678, 2.1960, 2.1346, 2.0881], device='cuda:0'), covar=tensor([0.1899, 0.4297, 0.3298, 0.0916, 0.5187, 0.3012, 0.3936, 0.4181], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0467, 0.0381, 0.0335, 0.0444, 0.0535, 0.0439, 0.0546], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:42:14,233 INFO [train.py:904] (0/8) Epoch 26, batch 4700, loss[loss=0.1643, simple_loss=0.251, pruned_loss=0.03882, over 16510.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2778, pruned_loss=0.05142, over 3195317.70 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:42:16,008 INFO [optim.py:368] (0/8) 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 04:42:16,455 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:42:22,436 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2019, 3.6246, 3.8966, 1.7758, 4.1772, 4.2435, 3.0802, 2.7937], device='cuda:0'), covar=tensor([0.1323, 0.0243, 0.0220, 0.1529, 0.0081, 0.0131, 0.0462, 0.0702], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0100, 0.0139, 0.0086, 0.0130, 0.0130, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 04:42:45,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9177, 4.9239, 4.6872, 3.3833, 4.0071, 4.7433, 3.9981, 2.8264], device='cuda:0'), covar=tensor([0.0546, 0.0031, 0.0038, 0.0367, 0.0106, 0.0077, 0.0108, 0.0438], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0088, 0.0090, 0.0136, 0.0102, 0.0114, 0.0098, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 04:43:23,114 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5048, 3.4990, 3.4717, 2.7287, 3.2513, 2.0634, 3.0720, 2.6841], device='cuda:0'), covar=tensor([0.0235, 0.0233, 0.0205, 0.0302, 0.0164, 0.2590, 0.0173, 0.0333], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0172, 0.0210, 0.0184, 0.0186, 0.0215, 0.0199, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:43:26,794 INFO [train.py:904] (0/8) Epoch 26, batch 4750, loss[loss=0.1697, simple_loss=0.2572, pruned_loss=0.04116, over 16732.00 frames. ], tot_loss[loss=0.1859, simple_loss=0.2734, pruned_loss=0.04921, over 3201796.88 frames. ], batch size: 124, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:43:38,918 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,002 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:43:45,156 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258515.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:21,177 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258539.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:41,654 INFO [train.py:904] (0/8) Epoch 26, batch 4800, loss[loss=0.1731, simple_loss=0.2676, pruned_loss=0.0393, over 16901.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2706, pruned_loss=0.0475, over 3209814.66 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:44:43,332 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 1.748e+02 2.158e+02 2.549e+02 7.061e+02, threshold=4.316e+02, percent-clipped=3.0 2023-05-02 04:44:56,482 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258563.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:44:58,727 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 04:45:11,377 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=258572.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:34,303 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258587.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:45:57,768 INFO [train.py:904] (0/8) Epoch 26, batch 4850, loss[loss=0.1798, simple_loss=0.2757, pruned_loss=0.04199, over 16938.00 frames. ], tot_loss[loss=0.1819, simple_loss=0.2711, pruned_loss=0.04636, over 3192081.64 frames. ], batch size: 109, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:46:03,359 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=258606.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:46:44,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2301, 2.3818, 2.4911, 3.9736, 2.2474, 2.7719, 2.4252, 2.5113], device='cuda:0'), covar=tensor([0.1435, 0.3512, 0.2954, 0.0548, 0.4046, 0.2386, 0.3609, 0.3176], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0467, 0.0380, 0.0334, 0.0443, 0.0534, 0.0438, 0.0545], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:46:48,180 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 04:47:14,832 INFO [train.py:904] (0/8) Epoch 26, batch 4900, loss[loss=0.1708, simple_loss=0.2553, pruned_loss=0.04311, over 12091.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2695, pruned_loss=0.04487, over 3178959.56 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:47:16,712 INFO [optim.py:368] (0/8) 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 04:47:17,074 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=258654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:48:29,895 INFO [train.py:904] (0/8) Epoch 26, batch 4950, loss[loss=0.182, simple_loss=0.2756, pruned_loss=0.04427, over 16901.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2695, pruned_loss=0.04448, over 3170498.05 frames. ], batch size: 116, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:48:31,525 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9367, 3.9228, 3.8805, 2.7525, 3.8949, 1.7384, 3.6106, 3.2952], device='cuda:0'), covar=tensor([0.0212, 0.0195, 0.0239, 0.0641, 0.0149, 0.3714, 0.0224, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0170, 0.0208, 0.0183, 0.0184, 0.0214, 0.0198, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 04:49:41,062 INFO [train.py:904] (0/8) Epoch 26, batch 5000, loss[loss=0.1661, simple_loss=0.2632, pruned_loss=0.03449, over 16873.00 frames. ], tot_loss[loss=0.18, simple_loss=0.2712, pruned_loss=0.04444, over 3182678.27 frames. ], batch size: 102, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:49:42,191 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.524e+02 2.044e+02 2.396e+02 2.809e+02 5.347e+02, threshold=4.792e+02, percent-clipped=3.0 2023-05-02 04:50:54,493 INFO [train.py:904] (0/8) Epoch 26, batch 5050, loss[loss=0.1786, simple_loss=0.2839, pruned_loss=0.03664, over 16827.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2717, pruned_loss=0.04422, over 3201254.36 frames. ], batch size: 96, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:51:05,014 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:52:00,774 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2050, 3.3199, 3.6006, 2.0925, 3.0702, 2.3703, 3.5715, 3.5318], device='cuda:0'), covar=tensor([0.0223, 0.0836, 0.0608, 0.2101, 0.0845, 0.0929, 0.0591, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0170, 0.0171, 0.0157, 0.0148, 0.0132, 0.0147, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 04:52:07,190 INFO [train.py:904] (0/8) Epoch 26, batch 5100, loss[loss=0.1828, simple_loss=0.2687, pruned_loss=0.04842, over 11631.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.27, pruned_loss=0.04344, over 3208817.66 frames. ], batch size: 247, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:52:08,938 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.295e+02 1.999e+02 2.336e+02 2.714e+02 4.372e+02, threshold=4.672e+02, percent-clipped=0.0 2023-05-02 04:52:27,835 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=258867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:53:21,036 INFO [train.py:904] (0/8) Epoch 26, batch 5150, loss[loss=0.1972, simple_loss=0.2918, pruned_loss=0.05131, over 15353.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2697, pruned_loss=0.04274, over 3221174.06 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:53:33,578 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.57 vs. limit=2.0 2023-05-02 04:54:35,101 INFO [train.py:904] (0/8) Epoch 26, batch 5200, loss[loss=0.1658, simple_loss=0.2607, pruned_loss=0.03544, over 15434.00 frames. ], tot_loss[loss=0.1768, simple_loss=0.2687, pruned_loss=0.04241, over 3219086.97 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:54:36,752 INFO [optim.py:368] (0/8) 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 04:54:53,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=258965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:55:48,701 INFO [train.py:904] (0/8) Epoch 26, batch 5250, loss[loss=0.1772, simple_loss=0.2775, pruned_loss=0.0384, over 16573.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2665, pruned_loss=0.04177, over 3229869.45 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:56:23,023 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259026.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:57:03,190 INFO [train.py:904] (0/8) Epoch 26, batch 5300, loss[loss=0.1838, simple_loss=0.2768, pruned_loss=0.04544, over 15353.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2639, pruned_loss=0.04104, over 3223070.63 frames. ], batch size: 190, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:57:04,414 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.339e+02 1.887e+02 2.180e+02 2.580e+02 4.921e+02, threshold=4.360e+02, percent-clipped=4.0 2023-05-02 04:57:56,235 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 04:58:18,028 INFO [train.py:904] (0/8) Epoch 26, batch 5350, loss[loss=0.1832, simple_loss=0.2704, pruned_loss=0.04803, over 16511.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2618, pruned_loss=0.04067, over 3207451.52 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:58:27,198 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:31,438 INFO [train.py:904] (0/8) Epoch 26, batch 5400, loss[loss=0.1643, simple_loss=0.2571, pruned_loss=0.03578, over 16456.00 frames. ], tot_loss[loss=0.1724, simple_loss=0.2633, pruned_loss=0.04076, over 3220873.21 frames. ], batch size: 68, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 04:59:32,576 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.458e+02 1.944e+02 2.298e+02 2.661e+02 4.475e+02, threshold=4.595e+02, percent-clipped=1.0 2023-05-02 04:59:38,343 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 04:59:52,978 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:00:17,399 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 05:00:48,645 INFO [train.py:904] (0/8) Epoch 26, batch 5450, loss[loss=0.1894, simple_loss=0.2747, pruned_loss=0.05208, over 11783.00 frames. ], tot_loss[loss=0.1759, simple_loss=0.2669, pruned_loss=0.04249, over 3205758.05 frames. ], batch size: 248, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:01:07,882 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:01:25,078 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:05,216 INFO [train.py:904] (0/8) Epoch 26, batch 5500, loss[loss=0.2475, simple_loss=0.3289, pruned_loss=0.08309, over 11625.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2741, pruned_loss=0.04663, over 3181231.12 frames. ], batch size: 246, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:02:07,105 INFO [optim.py:368] (0/8) 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 05:02:23,201 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259264.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:02:57,575 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:22,301 INFO [train.py:904] (0/8) Epoch 26, batch 5550, loss[loss=0.272, simple_loss=0.3322, pruned_loss=0.1059, over 11063.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2815, pruned_loss=0.05205, over 3142700.25 frames. ], batch size: 250, lr: 2.59e-03, grad_scale: 8.0 2023-05-02 05:03:50,536 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:03:57,668 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259325.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:04:40,422 INFO [train.py:904] (0/8) Epoch 26, batch 5600, loss[loss=0.2514, simple_loss=0.3309, pruned_loss=0.086, over 15278.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2856, pruned_loss=0.05531, over 3124690.36 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 16.0 2023-05-02 05:04:41,810 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.957e+02 3.012e+02 3.705e+02 4.647e+02 1.087e+03, threshold=7.410e+02, percent-clipped=8.0 2023-05-02 05:05:12,442 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1100, 3.1377, 3.4750, 1.6707, 3.6193, 3.7311, 2.8875, 2.5368], device='cuda:0'), covar=tensor([0.1203, 0.0311, 0.0256, 0.1432, 0.0120, 0.0176, 0.0487, 0.0662], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0101, 0.0140, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 05:05:29,355 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.21 vs. limit=5.0 2023-05-02 05:06:04,642 INFO [train.py:904] (0/8) Epoch 26, batch 5650, loss[loss=0.2546, simple_loss=0.3207, pruned_loss=0.09431, over 11505.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2905, pruned_loss=0.05953, over 3076076.45 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:06:22,268 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259414.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:07:22,192 INFO [train.py:904] (0/8) Epoch 26, batch 5700, loss[loss=0.2672, simple_loss=0.328, pruned_loss=0.1032, over 11287.00 frames. ], tot_loss[loss=0.2076, simple_loss=0.2924, pruned_loss=0.06139, over 3053110.81 frames. ], batch size: 246, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:07:25,093 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.978e+02 3.137e+02 3.722e+02 4.714e+02 8.155e+02, threshold=7.444e+02, percent-clipped=2.0 2023-05-02 05:07:56,972 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259475.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:08:39,971 INFO [train.py:904] (0/8) Epoch 26, batch 5750, loss[loss=0.1818, simple_loss=0.2747, pruned_loss=0.04441, over 16428.00 frames. ], tot_loss[loss=0.2112, simple_loss=0.2957, pruned_loss=0.06336, over 3029300.92 frames. ], batch size: 75, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:02,530 INFO [train.py:904] (0/8) Epoch 26, batch 5800, loss[loss=0.1753, simple_loss=0.2674, pruned_loss=0.04157, over 16514.00 frames. ], tot_loss[loss=0.2091, simple_loss=0.2947, pruned_loss=0.06179, over 3030728.56 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:10:05,679 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.039e+02 2.780e+02 3.397e+02 4.110e+02 5.922e+02, threshold=6.793e+02, percent-clipped=0.0 2023-05-02 05:10:47,027 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259582.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:11:19,041 INFO [train.py:904] (0/8) Epoch 26, batch 5850, loss[loss=0.2048, simple_loss=0.2903, pruned_loss=0.05971, over 17209.00 frames. ], tot_loss[loss=0.2061, simple_loss=0.2923, pruned_loss=0.05993, over 3055314.70 frames. ], batch size: 45, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:11:44,444 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259620.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:11:46,476 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:12:39,970 INFO [train.py:904] (0/8) Epoch 26, batch 5900, loss[loss=0.1645, simple_loss=0.2619, pruned_loss=0.03357, over 16898.00 frames. ], tot_loss[loss=0.2048, simple_loss=0.2914, pruned_loss=0.05907, over 3064304.48 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:12:43,698 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.729e+02 2.720e+02 3.451e+02 4.224e+02 7.867e+02, threshold=6.903e+02, percent-clipped=3.0 2023-05-02 05:13:00,394 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.02 vs. limit=2.0 2023-05-02 05:13:06,354 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9797, 5.4260, 5.6097, 5.2961, 5.4034, 5.9806, 5.4799, 5.2364], device='cuda:0'), covar=tensor([0.1003, 0.1665, 0.2706, 0.1960, 0.2394, 0.0829, 0.1453, 0.2268], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0623, 0.0684, 0.0510, 0.0673, 0.0712, 0.0534, 0.0680], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 05:13:08,194 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259669.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:13:21,859 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7356, 4.7995, 4.6094, 4.2764, 4.2505, 4.7025, 4.4839, 4.4235], device='cuda:0'), covar=tensor([0.0693, 0.0822, 0.0352, 0.0387, 0.1017, 0.0606, 0.0621, 0.0793], device='cuda:0'), in_proj_covar=tensor([0.0305, 0.0455, 0.0355, 0.0359, 0.0357, 0.0413, 0.0242, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:13:26,902 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5909, 4.4706, 4.6120, 4.7824, 4.9450, 4.4865, 4.9483, 4.9597], device='cuda:0'), covar=tensor([0.1973, 0.1277, 0.1730, 0.0808, 0.0753, 0.1107, 0.0816, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0666, 0.0816, 0.0941, 0.0826, 0.0629, 0.0655, 0.0682, 0.0795], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:14:01,054 INFO [train.py:904] (0/8) Epoch 26, batch 5950, loss[loss=0.1957, simple_loss=0.2915, pruned_loss=0.04996, over 16756.00 frames. ], tot_loss[loss=0.2041, simple_loss=0.2922, pruned_loss=0.05799, over 3078177.17 frames. ], batch size: 124, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:14:03,277 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7327, 4.7328, 5.0984, 5.0531, 5.0665, 4.7776, 4.7290, 4.6103], device='cuda:0'), covar=tensor([0.0300, 0.0594, 0.0315, 0.0351, 0.0402, 0.0392, 0.0901, 0.0500], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0474, 0.0461, 0.0423, 0.0509, 0.0485, 0.0564, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 05:14:23,644 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0016, 4.0420, 4.3220, 4.2937, 4.3119, 4.0740, 4.0857, 4.0801], device='cuda:0'), covar=tensor([0.0345, 0.0671, 0.0408, 0.0430, 0.0425, 0.0483, 0.0804, 0.0517], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0473, 0.0460, 0.0422, 0.0508, 0.0484, 0.0562, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 05:14:29,214 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 05:15:12,644 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259749.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:15:18,043 INFO [train.py:904] (0/8) Epoch 26, batch 6000, loss[loss=0.2105, simple_loss=0.2911, pruned_loss=0.06493, over 11386.00 frames. ], tot_loss[loss=0.2031, simple_loss=0.2911, pruned_loss=0.05754, over 3061401.75 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:15:18,044 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 05:15:28,182 INFO [train.py:938] (0/8) Epoch 26, validation: loss=0.1485, simple_loss=0.2607, pruned_loss=0.01818, over 944034.00 frames. 2023-05-02 05:15:28,182 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 05:15:30,552 INFO [optim.py:368] (0/8) 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 05:15:55,089 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=259770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:13,060 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:16:33,788 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 05:16:46,434 INFO [train.py:904] (0/8) Epoch 26, batch 6050, loss[loss=0.1941, simple_loss=0.2995, pruned_loss=0.04439, over 16765.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2889, pruned_loss=0.05661, over 3072152.30 frames. ], batch size: 76, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:16:59,616 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259810.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:17:03,784 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0752, 5.3395, 5.1682, 5.1547, 4.9036, 4.8116, 4.6764, 5.4628], device='cuda:0'), covar=tensor([0.1281, 0.0910, 0.0998, 0.0913, 0.0845, 0.0946, 0.1401, 0.0876], device='cuda:0'), in_proj_covar=tensor([0.0699, 0.0843, 0.0694, 0.0650, 0.0534, 0.0534, 0.0711, 0.0659], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:17:51,009 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259843.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:18:05,776 INFO [train.py:904] (0/8) Epoch 26, batch 6100, loss[loss=0.1725, simple_loss=0.2626, pruned_loss=0.04113, over 17239.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2882, pruned_loss=0.05555, over 3077395.28 frames. ], batch size: 45, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:18:08,751 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3055, 2.9817, 2.7593, 2.2707, 2.2805, 2.3883, 2.9927, 2.9039], device='cuda:0'), covar=tensor([0.2504, 0.0651, 0.1479, 0.2468, 0.2090, 0.1974, 0.0487, 0.1187], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0274, 0.0309, 0.0322, 0.0303, 0.0271, 0.0301, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 05:18:09,302 INFO [optim.py:368] (0/8) 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 05:18:27,200 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=259866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:18:38,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6853, 1.7983, 1.5882, 1.5282, 1.9439, 1.6338, 1.5882, 1.9318], device='cuda:0'), covar=tensor([0.0251, 0.0335, 0.0455, 0.0403, 0.0258, 0.0318, 0.0201, 0.0238], device='cuda:0'), in_proj_covar=tensor([0.0225, 0.0240, 0.0230, 0.0231, 0.0240, 0.0239, 0.0239, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:18:47,329 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-05-02 05:18:52,421 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=259882.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:19:23,809 INFO [train.py:904] (0/8) Epoch 26, batch 6150, loss[loss=0.1871, simple_loss=0.2758, pruned_loss=0.04921, over 16566.00 frames. ], tot_loss[loss=0.1976, simple_loss=0.2861, pruned_loss=0.05452, over 3079879.31 frames. ], batch size: 75, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:19:48,993 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=259927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:05,848 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259930.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:20:42,393 INFO [train.py:904] (0/8) Epoch 26, batch 6200, loss[loss=0.2189, simple_loss=0.3082, pruned_loss=0.06475, over 15454.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2844, pruned_loss=0.05398, over 3095747.76 frames. ], batch size: 191, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:20:44,632 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.064e+02 2.750e+02 3.295e+02 3.920e+02 8.789e+02, threshold=6.590e+02, percent-clipped=2.0 2023-05-02 05:21:04,923 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=259968.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:21:36,998 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9362, 4.1826, 4.0171, 4.0646, 3.7439, 3.8371, 3.8457, 4.1767], device='cuda:0'), covar=tensor([0.1086, 0.0888, 0.0973, 0.0860, 0.0764, 0.1454, 0.0977, 0.0943], device='cuda:0'), in_proj_covar=tensor([0.0705, 0.0850, 0.0700, 0.0657, 0.0540, 0.0540, 0.0717, 0.0666], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:21:54,208 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-260000.pt 2023-05-02 05:22:00,555 INFO [train.py:904] (0/8) Epoch 26, batch 6250, loss[loss=0.2449, simple_loss=0.3116, pruned_loss=0.08909, over 11706.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2846, pruned_loss=0.05436, over 3107646.72 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:22:01,182 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.01 vs. limit=5.0 2023-05-02 05:22:41,218 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:23:15,630 INFO [train.py:904] (0/8) Epoch 26, batch 6300, loss[loss=0.1637, simple_loss=0.2572, pruned_loss=0.03509, over 16835.00 frames. ], tot_loss[loss=0.1961, simple_loss=0.2845, pruned_loss=0.05383, over 3105727.00 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:23:19,604 INFO [optim.py:368] (0/8) 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 05:23:41,796 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4869, 4.1346, 3.9375, 2.5857, 3.6585, 4.0797, 3.7064, 1.9495], device='cuda:0'), covar=tensor([0.0676, 0.0072, 0.0098, 0.0551, 0.0141, 0.0168, 0.0142, 0.0724], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0087, 0.0089, 0.0134, 0.0100, 0.0113, 0.0097, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 05:23:44,822 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260070.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:17,424 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:24:35,610 INFO [train.py:904] (0/8) Epoch 26, batch 6350, loss[loss=0.228, simple_loss=0.2973, pruned_loss=0.07941, over 11774.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2856, pruned_loss=0.05541, over 3082600.60 frames. ], batch size: 246, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:24:39,481 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260105.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:24:59,058 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260118.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:25:29,706 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260138.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:25:52,110 INFO [train.py:904] (0/8) Epoch 26, batch 6400, loss[loss=0.1852, simple_loss=0.274, pruned_loss=0.04817, over 16277.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2855, pruned_loss=0.05573, over 3098812.36 frames. ], batch size: 165, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:25:54,619 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.816e+02 2.921e+02 3.424e+02 4.125e+02 9.297e+02, threshold=6.848e+02, percent-clipped=3.0 2023-05-02 05:27:08,578 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 05:27:08,968 INFO [train.py:904] (0/8) Epoch 26, batch 6450, loss[loss=0.1844, simple_loss=0.2774, pruned_loss=0.04567, over 16826.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2855, pruned_loss=0.05534, over 3094921.03 frames. ], batch size: 102, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:27:39,887 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:27:43,229 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260224.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:28:23,376 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4225, 3.3995, 3.4476, 3.5260, 3.5587, 3.3132, 3.5295, 3.6089], device='cuda:0'), covar=tensor([0.1340, 0.0999, 0.1014, 0.0629, 0.0734, 0.2353, 0.1146, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0663, 0.0814, 0.0938, 0.0822, 0.0629, 0.0652, 0.0682, 0.0796], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:28:28,525 INFO [train.py:904] (0/8) Epoch 26, batch 6500, loss[loss=0.2196, simple_loss=0.2878, pruned_loss=0.07574, over 11985.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2839, pruned_loss=0.05517, over 3082693.67 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:28:31,535 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.633e+02 2.830e+02 3.427e+02 3.864e+02 5.518e+02, threshold=6.855e+02, percent-clipped=0.0 2023-05-02 05:29:03,728 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3078, 4.3595, 4.6576, 4.6389, 4.6711, 4.3684, 4.3604, 4.3214], device='cuda:0'), covar=tensor([0.0370, 0.0667, 0.0408, 0.0429, 0.0486, 0.0452, 0.1030, 0.0546], device='cuda:0'), in_proj_covar=tensor([0.0424, 0.0479, 0.0465, 0.0426, 0.0512, 0.0488, 0.0566, 0.0391], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 05:29:20,129 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260285.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:29:21,312 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7764, 3.5797, 3.5829, 3.8998, 3.9728, 3.6490, 3.8997, 3.9834], device='cuda:0'), covar=tensor([0.1607, 0.1493, 0.2097, 0.1056, 0.0990, 0.2529, 0.1358, 0.1217], device='cuda:0'), in_proj_covar=tensor([0.0659, 0.0811, 0.0933, 0.0818, 0.0626, 0.0648, 0.0678, 0.0791], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:29:49,658 INFO [train.py:904] (0/8) Epoch 26, batch 6550, loss[loss=0.2075, simple_loss=0.3083, pruned_loss=0.05335, over 16586.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2869, pruned_loss=0.05577, over 3091700.34 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:30:21,543 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1685, 5.5057, 5.2266, 5.2597, 5.0059, 4.9120, 4.8695, 5.5913], device='cuda:0'), covar=tensor([0.1324, 0.0807, 0.0985, 0.0921, 0.0827, 0.0873, 0.1312, 0.0832], device='cuda:0'), in_proj_covar=tensor([0.0700, 0.0843, 0.0695, 0.0651, 0.0536, 0.0536, 0.0713, 0.0659], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:30:52,435 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5057, 1.7311, 2.1658, 2.3429, 2.3874, 2.6817, 1.8280, 2.6366], device='cuda:0'), covar=tensor([0.0219, 0.0523, 0.0316, 0.0348, 0.0357, 0.0213, 0.0568, 0.0146], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0195, 0.0183, 0.0188, 0.0203, 0.0161, 0.0200, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:31:07,553 INFO [train.py:904] (0/8) Epoch 26, batch 6600, loss[loss=0.2733, simple_loss=0.334, pruned_loss=0.1063, over 11304.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2892, pruned_loss=0.05617, over 3111173.62 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:31:09,949 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.723e+02 2.646e+02 3.224e+02 4.276e+02 9.076e+02, threshold=6.447e+02, percent-clipped=2.0 2023-05-02 05:31:59,497 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:32:26,397 INFO [train.py:904] (0/8) Epoch 26, batch 6650, loss[loss=0.1912, simple_loss=0.2759, pruned_loss=0.05322, over 16697.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.289, pruned_loss=0.05687, over 3084285.52 frames. ], batch size: 134, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:32:29,802 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260405.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:33:21,086 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:33:43,218 INFO [train.py:904] (0/8) Epoch 26, batch 6700, loss[loss=0.2024, simple_loss=0.2826, pruned_loss=0.06113, over 16942.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2875, pruned_loss=0.05703, over 3096833.20 frames. ], batch size: 109, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:33:43,551 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260453.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:33:45,912 INFO [optim.py:368] (0/8) 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 05:34:34,791 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260486.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:35:01,134 INFO [train.py:904] (0/8) Epoch 26, batch 6750, loss[loss=0.1723, simple_loss=0.2665, pruned_loss=0.03907, over 16712.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2864, pruned_loss=0.05695, over 3100205.66 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:35:31,804 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260522.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:35:41,971 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3247, 4.4160, 4.2414, 3.9581, 3.9593, 4.3463, 4.0621, 4.0699], device='cuda:0'), covar=tensor([0.0608, 0.0486, 0.0281, 0.0301, 0.0770, 0.0426, 0.0689, 0.0613], device='cuda:0'), in_proj_covar=tensor([0.0303, 0.0452, 0.0351, 0.0355, 0.0353, 0.0409, 0.0241, 0.0426], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:36:18,141 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7573, 3.9868, 2.9832, 2.3746, 2.6369, 2.5329, 4.3652, 3.4909], device='cuda:0'), covar=tensor([0.2850, 0.0620, 0.1778, 0.2560, 0.2647, 0.2047, 0.0369, 0.1279], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0273, 0.0309, 0.0322, 0.0303, 0.0271, 0.0300, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 05:36:19,269 INFO [train.py:904] (0/8) Epoch 26, batch 6800, loss[loss=0.1906, simple_loss=0.2796, pruned_loss=0.05075, over 16439.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2868, pruned_loss=0.05698, over 3096249.07 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 8.0 2023-05-02 05:36:21,474 INFO [optim.py:368] (0/8) 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 05:36:44,739 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:37:01,680 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=260580.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 05:37:35,992 INFO [train.py:904] (0/8) Epoch 26, batch 6850, loss[loss=0.2494, simple_loss=0.312, pruned_loss=0.09343, over 11541.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2871, pruned_loss=0.05714, over 3102632.59 frames. ], batch size: 248, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:37:57,017 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5123, 4.6906, 4.8824, 4.5828, 4.6652, 5.2558, 4.6919, 4.4031], device='cuda:0'), covar=tensor([0.1421, 0.1948, 0.2624, 0.2124, 0.2629, 0.1022, 0.1773, 0.2545], device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0630, 0.0690, 0.0513, 0.0680, 0.0720, 0.0540, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 05:38:10,016 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.10 vs. limit=5.0 2023-05-02 05:38:50,108 INFO [train.py:904] (0/8) Epoch 26, batch 6900, loss[loss=0.175, simple_loss=0.2673, pruned_loss=0.04136, over 16672.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2888, pruned_loss=0.05594, over 3125023.63 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:38:53,863 INFO [optim.py:368] (0/8) 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 05:39:40,145 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260686.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:40:05,864 INFO [train.py:904] (0/8) Epoch 26, batch 6950, loss[loss=0.2069, simple_loss=0.285, pruned_loss=0.06439, over 15310.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2904, pruned_loss=0.05782, over 3120011.54 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:40:48,884 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:40:54,431 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:41:23,883 INFO [train.py:904] (0/8) Epoch 26, batch 7000, loss[loss=0.1921, simple_loss=0.2917, pruned_loss=0.04622, over 16591.00 frames. ], tot_loss[loss=0.202, simple_loss=0.2899, pruned_loss=0.05703, over 3117206.51 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:41:29,421 INFO [optim.py:368] (0/8) 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 05:41:56,087 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5268, 3.6005, 3.3692, 3.0220, 3.2225, 3.5024, 3.3255, 3.3300], device='cuda:0'), covar=tensor([0.0646, 0.0762, 0.0299, 0.0314, 0.0512, 0.0526, 0.1484, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0451, 0.0351, 0.0354, 0.0352, 0.0407, 0.0241, 0.0425], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:42:23,692 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:42:29,007 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 05:42:39,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.61 vs. limit=2.0 2023-05-02 05:42:42,011 INFO [train.py:904] (0/8) Epoch 26, batch 7050, loss[loss=0.1872, simple_loss=0.2866, pruned_loss=0.04389, over 16758.00 frames. ], tot_loss[loss=0.2024, simple_loss=0.291, pruned_loss=0.05695, over 3116431.61 frames. ], batch size: 89, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:42:46,659 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3448, 2.1367, 1.7358, 1.9460, 2.3840, 2.1101, 2.1159, 2.5198], device='cuda:0'), covar=tensor([0.0247, 0.0422, 0.0623, 0.0523, 0.0286, 0.0393, 0.0272, 0.0321], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0239, 0.0231, 0.0231, 0.0241, 0.0239, 0.0238, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:42:55,726 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:43:59,445 INFO [train.py:904] (0/8) Epoch 26, batch 7100, loss[loss=0.2223, simple_loss=0.2944, pruned_loss=0.07515, over 11546.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2899, pruned_loss=0.05691, over 3101564.98 frames. ], batch size: 250, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:44:05,369 INFO [optim.py:368] (0/8) 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 05:44:42,714 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=260880.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:44:44,296 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=260881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:45:17,439 INFO [train.py:904] (0/8) Epoch 26, batch 7150, loss[loss=0.262, simple_loss=0.3197, pruned_loss=0.1021, over 11694.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.288, pruned_loss=0.05651, over 3109328.24 frames. ], batch size: 250, lr: 2.58e-03, grad_scale: 2.0 2023-05-02 05:45:30,827 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-02 05:45:53,428 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=260928.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 05:46:13,579 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=260942.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:46:29,366 INFO [train.py:904] (0/8) Epoch 26, batch 7200, loss[loss=0.2091, simple_loss=0.2973, pruned_loss=0.06041, over 11822.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2865, pruned_loss=0.05557, over 3090568.20 frames. ], batch size: 247, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:46:35,555 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.782e+02 3.336e+02 4.472e+02 6.403e+02, threshold=6.673e+02, percent-clipped=2.0 2023-05-02 05:47:53,036 INFO [train.py:904] (0/8) Epoch 26, batch 7250, loss[loss=0.1867, simple_loss=0.2612, pruned_loss=0.05606, over 16665.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2842, pruned_loss=0.05441, over 3099579.58 frames. ], batch size: 62, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:47:57,647 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.78 vs. limit=2.0 2023-05-02 05:47:59,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6041, 2.5861, 2.3671, 4.0560, 2.9714, 3.9337, 1.4810, 2.8226], device='cuda:0'), covar=tensor([0.1538, 0.0893, 0.1445, 0.0168, 0.0242, 0.0406, 0.1825, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0180, 0.0199, 0.0198, 0.0208, 0.0218, 0.0208, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 05:48:45,064 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6472, 3.9051, 2.9543, 2.3087, 2.5735, 2.5618, 4.2732, 3.4550], device='cuda:0'), covar=tensor([0.3113, 0.0638, 0.1860, 0.2754, 0.2699, 0.2127, 0.0396, 0.1335], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0275, 0.0310, 0.0323, 0.0304, 0.0272, 0.0301, 0.0349], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 05:48:46,283 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4127, 3.3738, 3.4367, 3.5131, 3.5369, 3.3191, 3.5212, 3.5846], device='cuda:0'), covar=tensor([0.1211, 0.0917, 0.0924, 0.0553, 0.0664, 0.2239, 0.0991, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0650, 0.0802, 0.0925, 0.0811, 0.0622, 0.0644, 0.0674, 0.0785], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 05:49:09,263 INFO [train.py:904] (0/8) Epoch 26, batch 7300, loss[loss=0.2047, simple_loss=0.2986, pruned_loss=0.05541, over 16831.00 frames. ], tot_loss[loss=0.1951, simple_loss=0.283, pruned_loss=0.05364, over 3107278.04 frames. ], batch size: 116, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:49:15,975 INFO [optim.py:368] (0/8) 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 05:49:47,240 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9597, 2.9196, 1.9928, 3.2412, 2.3236, 3.2959, 2.1472, 2.5165], device='cuda:0'), covar=tensor([0.0353, 0.0431, 0.1609, 0.0179, 0.0891, 0.0417, 0.1576, 0.0895], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0170, 0.0179, 0.0219, 0.0206, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 05:49:53,995 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 05:50:00,845 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:50:26,314 INFO [train.py:904] (0/8) Epoch 26, batch 7350, loss[loss=0.2263, simple_loss=0.299, pruned_loss=0.07681, over 11261.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2844, pruned_loss=0.05489, over 3074263.28 frames. ], batch size: 246, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:50:32,675 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261107.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:50:59,374 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9118, 2.9484, 2.5582, 4.8698, 3.7451, 4.1691, 1.6940, 3.0160], device='cuda:0'), covar=tensor([0.1366, 0.0806, 0.1309, 0.0177, 0.0332, 0.0412, 0.1670, 0.0878], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0198, 0.0208, 0.0217, 0.0208, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 05:51:18,069 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261136.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 05:51:44,998 INFO [train.py:904] (0/8) Epoch 26, batch 7400, loss[loss=0.2071, simple_loss=0.2911, pruned_loss=0.06157, over 15447.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2861, pruned_loss=0.05582, over 3054583.98 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:51:50,756 INFO [optim.py:368] (0/8) 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 05:52:53,537 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 05:52:55,031 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261197.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 05:53:04,391 INFO [train.py:904] (0/8) Epoch 26, batch 7450, loss[loss=0.2119, simple_loss=0.3037, pruned_loss=0.06007, over 16500.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2873, pruned_loss=0.05717, over 3038167.75 frames. ], batch size: 68, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:53:48,000 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8836, 3.1679, 3.3571, 1.9869, 3.0105, 2.2308, 3.3619, 3.4294], device='cuda:0'), covar=tensor([0.0292, 0.0887, 0.0643, 0.2338, 0.0858, 0.1060, 0.0749, 0.1004], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0169, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 05:54:01,627 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:54:27,105 INFO [train.py:904] (0/8) Epoch 26, batch 7500, loss[loss=0.2106, simple_loss=0.2957, pruned_loss=0.06279, over 15326.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.287, pruned_loss=0.05601, over 3048239.86 frames. ], batch size: 190, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:54:33,504 INFO [optim.py:368] (0/8) 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 05:54:42,417 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 05:54:48,433 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3607, 2.9585, 2.6285, 2.2969, 2.2429, 2.2873, 2.9513, 2.8475], device='cuda:0'), covar=tensor([0.2587, 0.0748, 0.1702, 0.2584, 0.2417, 0.2270, 0.0531, 0.1506], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0272, 0.0308, 0.0321, 0.0302, 0.0270, 0.0299, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 05:55:45,648 INFO [train.py:904] (0/8) Epoch 26, batch 7550, loss[loss=0.203, simple_loss=0.2887, pruned_loss=0.05864, over 15533.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.2869, pruned_loss=0.05702, over 3026824.58 frames. ], batch size: 191, lr: 2.58e-03, grad_scale: 4.0 2023-05-02 05:57:01,733 INFO [train.py:904] (0/8) Epoch 26, batch 7600, loss[loss=0.1773, simple_loss=0.2765, pruned_loss=0.03904, over 16824.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2862, pruned_loss=0.0568, over 3063227.51 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 05:57:07,520 INFO [optim.py:368] (0/8) 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 05:57:17,396 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1734, 4.2615, 4.5334, 4.5217, 4.5190, 4.2278, 4.2497, 4.1844], device='cuda:0'), covar=tensor([0.0369, 0.0625, 0.0419, 0.0408, 0.0486, 0.0476, 0.1001, 0.0570], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0471, 0.0456, 0.0419, 0.0505, 0.0481, 0.0557, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 05:57:54,037 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261386.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:58:20,648 INFO [train.py:904] (0/8) Epoch 26, batch 7650, loss[loss=0.2019, simple_loss=0.2952, pruned_loss=0.05428, over 16839.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2867, pruned_loss=0.05744, over 3067876.98 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 05:58:26,587 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:08,008 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:28,038 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3298, 3.8782, 3.8566, 2.3586, 3.4603, 3.9163, 3.4584, 2.1560], device='cuda:0'), covar=tensor([0.0597, 0.0061, 0.0067, 0.0498, 0.0118, 0.0126, 0.0123, 0.0511], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0087, 0.0088, 0.0134, 0.0100, 0.0113, 0.0097, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 05:59:36,576 INFO [train.py:904] (0/8) Epoch 26, batch 7700, loss[loss=0.2036, simple_loss=0.2899, pruned_loss=0.05867, over 17247.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2859, pruned_loss=0.05728, over 3082905.57 frames. ], batch size: 52, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 05:59:40,111 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261455.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:42,091 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 05:59:46,569 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.047e+02 2.911e+02 3.600e+02 4.177e+02 7.928e+02, threshold=7.200e+02, percent-clipped=2.0 2023-05-02 06:00:36,068 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261492.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 06:00:52,215 INFO [train.py:904] (0/8) Epoch 26, batch 7750, loss[loss=0.2242, simple_loss=0.2934, pruned_loss=0.07752, over 11751.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2863, pruned_loss=0.05706, over 3080724.07 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:01:13,791 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:01:45,037 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:02:09,790 INFO [train.py:904] (0/8) Epoch 26, batch 7800, loss[loss=0.2151, simple_loss=0.2953, pruned_loss=0.06749, over 15426.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2866, pruned_loss=0.05709, over 3103703.55 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:02:19,358 INFO [optim.py:368] (0/8) 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 06:03:00,433 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261585.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:03:28,319 INFO [train.py:904] (0/8) Epoch 26, batch 7850, loss[loss=0.18, simple_loss=0.271, pruned_loss=0.0445, over 16238.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2874, pruned_loss=0.0572, over 3086314.08 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:26,798 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7813, 4.6180, 4.8403, 4.9718, 5.1384, 4.6284, 5.1629, 5.1424], device='cuda:0'), covar=tensor([0.1968, 0.1393, 0.1567, 0.0729, 0.0597, 0.0950, 0.0533, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0651, 0.0804, 0.0926, 0.0813, 0.0622, 0.0647, 0.0675, 0.0787], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:04:44,281 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 06:04:46,524 INFO [train.py:904] (0/8) Epoch 26, batch 7900, loss[loss=0.1808, simple_loss=0.267, pruned_loss=0.04735, over 16855.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2868, pruned_loss=0.05689, over 3096373.10 frames. ], batch size: 39, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:04:55,565 INFO [optim.py:368] (0/8) 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 06:04:56,171 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9241, 3.1285, 3.2960, 2.0192, 3.0063, 2.1940, 3.3977, 3.4506], device='cuda:0'), covar=tensor([0.0266, 0.0885, 0.0655, 0.2262, 0.0843, 0.1078, 0.0669, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0168, 0.0169, 0.0155, 0.0146, 0.0131, 0.0146, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 06:05:31,892 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 06:06:05,693 INFO [train.py:904] (0/8) Epoch 26, batch 7950, loss[loss=0.2128, simple_loss=0.296, pruned_loss=0.06475, over 17000.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2867, pruned_loss=0.05663, over 3106963.42 frames. ], batch size: 55, lr: 2.57e-03, grad_scale: 2.0 2023-05-02 06:06:59,445 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2029, 2.1229, 1.7519, 1.8094, 2.3070, 1.9894, 1.9519, 2.4005], device='cuda:0'), covar=tensor([0.0246, 0.0446, 0.0587, 0.0518, 0.0281, 0.0410, 0.0248, 0.0289], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0238, 0.0229, 0.0230, 0.0240, 0.0238, 0.0238, 0.0236], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:07:06,733 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7623, 4.0253, 3.0223, 2.3987, 2.6885, 2.6316, 4.4260, 3.5737], device='cuda:0'), covar=tensor([0.3115, 0.0664, 0.1857, 0.2884, 0.3108, 0.2128, 0.0427, 0.1410], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0274, 0.0310, 0.0323, 0.0305, 0.0273, 0.0302, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 06:07:23,919 INFO [train.py:904] (0/8) Epoch 26, batch 8000, loss[loss=0.2159, simple_loss=0.3042, pruned_loss=0.06379, over 16808.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2877, pruned_loss=0.05718, over 3106850.16 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:07:32,670 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.840e+02 2.545e+02 3.130e+02 3.720e+02 6.505e+02, threshold=6.260e+02, percent-clipped=1.0 2023-05-02 06:08:04,503 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 06:08:23,675 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=261792.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 06:08:40,304 INFO [train.py:904] (0/8) Epoch 26, batch 8050, loss[loss=0.2008, simple_loss=0.2762, pruned_loss=0.06269, over 11537.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2875, pruned_loss=0.05717, over 3091932.85 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:08:54,196 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=261812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:13,108 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261824.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:27,156 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:09:37,848 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=261840.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 06:09:57,228 INFO [train.py:904] (0/8) Epoch 26, batch 8100, loss[loss=0.194, simple_loss=0.2831, pruned_loss=0.0524, over 17012.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.287, pruned_loss=0.05638, over 3098921.00 frames. ], batch size: 55, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:10:06,904 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.718e+02 2.493e+02 2.989e+02 3.536e+02 5.601e+02, threshold=5.979e+02, percent-clipped=0.0 2023-05-02 06:10:46,791 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:10:59,165 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8771, 2.0990, 2.4863, 2.7754, 2.6931, 3.2954, 2.1294, 3.2244], device='cuda:0'), covar=tensor([0.0241, 0.0524, 0.0361, 0.0395, 0.0354, 0.0183, 0.0584, 0.0166], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0196, 0.0183, 0.0189, 0.0204, 0.0162, 0.0201, 0.0161], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:11:00,408 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=261894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:11:14,755 INFO [train.py:904] (0/8) Epoch 26, batch 8150, loss[loss=0.1642, simple_loss=0.2502, pruned_loss=0.03912, over 16573.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2844, pruned_loss=0.05524, over 3107324.25 frames. ], batch size: 62, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:33,066 INFO [train.py:904] (0/8) Epoch 26, batch 8200, loss[loss=0.1869, simple_loss=0.2757, pruned_loss=0.04899, over 16897.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2821, pruned_loss=0.05474, over 3102100.33 frames. ], batch size: 109, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:12:43,206 INFO [optim.py:368] (0/8) 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 06:12:56,730 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9804, 5.3173, 5.0730, 5.0571, 4.7625, 4.7621, 4.6491, 5.3824], device='cuda:0'), covar=tensor([0.1230, 0.0812, 0.0919, 0.0897, 0.0912, 0.0955, 0.1317, 0.0859], device='cuda:0'), in_proj_covar=tensor([0.0694, 0.0832, 0.0686, 0.0641, 0.0532, 0.0532, 0.0701, 0.0653], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:13:12,970 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=261977.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:13:18,080 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 06:13:51,111 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-262000.pt 2023-05-02 06:13:58,699 INFO [train.py:904] (0/8) Epoch 26, batch 8250, loss[loss=0.2139, simple_loss=0.3065, pruned_loss=0.06065, over 16301.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2816, pruned_loss=0.05247, over 3090919.55 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:14:25,664 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-02 06:14:57,046 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262038.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:15:03,104 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8596, 2.6009, 2.5244, 1.9866, 2.6081, 2.7160, 2.6239, 1.7492], device='cuda:0'), covar=tensor([0.0466, 0.0154, 0.0127, 0.0411, 0.0150, 0.0163, 0.0147, 0.0611], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0087, 0.0088, 0.0134, 0.0100, 0.0113, 0.0097, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 06:15:22,690 INFO [train.py:904] (0/8) Epoch 26, batch 8300, loss[loss=0.1735, simple_loss=0.2705, pruned_loss=0.03828, over 16500.00 frames. ], tot_loss[loss=0.1889, simple_loss=0.2788, pruned_loss=0.04948, over 3075279.34 frames. ], batch size: 68, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:15:32,831 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.264e+02 2.806e+02 3.280e+02 5.175e+02, threshold=5.611e+02, percent-clipped=0.0 2023-05-02 06:15:42,006 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 06:16:07,316 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.21 vs. limit=5.0 2023-05-02 06:16:44,228 INFO [train.py:904] (0/8) Epoch 26, batch 8350, loss[loss=0.1964, simple_loss=0.2821, pruned_loss=0.05533, over 12062.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2779, pruned_loss=0.04785, over 3063566.54 frames. ], batch size: 247, lr: 2.57e-03, grad_scale: 4.0 2023-05-02 06:16:59,003 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:17:19,728 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-05-02 06:17:22,481 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.81 vs. limit=2.0 2023-05-02 06:17:24,225 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8476, 3.8813, 4.1634, 4.1265, 4.1305, 3.9303, 3.9020, 3.9683], device='cuda:0'), covar=tensor([0.0418, 0.0889, 0.0531, 0.0503, 0.0564, 0.0610, 0.1046, 0.0628], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0475, 0.0459, 0.0422, 0.0508, 0.0484, 0.0558, 0.0387], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 06:17:51,703 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0995, 1.5823, 1.9554, 2.0347, 2.2078, 2.3295, 1.7943, 2.3216], device='cuda:0'), covar=tensor([0.0352, 0.0565, 0.0368, 0.0448, 0.0409, 0.0281, 0.0632, 0.0188], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0195, 0.0182, 0.0187, 0.0203, 0.0160, 0.0199, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:17:53,462 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 06:17:59,476 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4712, 3.0986, 2.7570, 2.2820, 2.2502, 2.3397, 3.0149, 2.8998], device='cuda:0'), covar=tensor([0.2609, 0.0731, 0.1635, 0.2998, 0.2919, 0.2435, 0.0489, 0.1490], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0271, 0.0307, 0.0320, 0.0301, 0.0270, 0.0299, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 06:18:04,038 INFO [train.py:904] (0/8) Epoch 26, batch 8400, loss[loss=0.1625, simple_loss=0.265, pruned_loss=0.03002, over 16822.00 frames. ], tot_loss[loss=0.1834, simple_loss=0.2754, pruned_loss=0.04571, over 3062841.52 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:18:05,962 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8630, 1.3089, 1.7843, 1.7141, 1.8537, 1.9894, 1.6332, 1.9005], device='cuda:0'), covar=tensor([0.0299, 0.0548, 0.0290, 0.0347, 0.0355, 0.0227, 0.0617, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0195, 0.0182, 0.0187, 0.0203, 0.0160, 0.0199, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:18:13,167 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.602e+02 2.393e+02 2.671e+02 3.015e+02 4.313e+02, threshold=5.342e+02, percent-clipped=0.0 2023-05-02 06:18:14,934 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:18:24,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1337, 4.2792, 4.0367, 3.7863, 3.6397, 4.1892, 3.8656, 3.8529], device='cuda:0'), covar=tensor([0.0685, 0.0655, 0.0432, 0.0407, 0.1063, 0.0546, 0.0954, 0.0822], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0452, 0.0351, 0.0352, 0.0349, 0.0406, 0.0241, 0.0423], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:18:47,912 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262180.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:02,243 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262189.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:19:24,268 INFO [train.py:904] (0/8) Epoch 26, batch 8450, loss[loss=0.1612, simple_loss=0.2577, pruned_loss=0.03238, over 16672.00 frames. ], tot_loss[loss=0.181, simple_loss=0.2736, pruned_loss=0.04418, over 3069043.63 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:06,394 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9471, 4.2257, 4.0394, 4.1099, 3.7680, 3.7779, 3.8573, 4.2180], device='cuda:0'), covar=tensor([0.1295, 0.1025, 0.1087, 0.0838, 0.0866, 0.1888, 0.1178, 0.1043], device='cuda:0'), in_proj_covar=tensor([0.0691, 0.0829, 0.0683, 0.0637, 0.0529, 0.0529, 0.0698, 0.0647], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:20:08,362 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 06:20:48,811 INFO [train.py:904] (0/8) Epoch 26, batch 8500, loss[loss=0.1522, simple_loss=0.2467, pruned_loss=0.02886, over 15329.00 frames. ], tot_loss[loss=0.1772, simple_loss=0.2701, pruned_loss=0.04216, over 3074480.78 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:20:57,923 INFO [optim.py:368] (0/8) 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 06:22:09,830 INFO [train.py:904] (0/8) Epoch 26, batch 8550, loss[loss=0.1915, simple_loss=0.2896, pruned_loss=0.04667, over 15351.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2678, pruned_loss=0.04118, over 3064631.40 frames. ], batch size: 191, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:07,432 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=262333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:23:40,597 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8388, 2.1383, 2.5188, 1.7089, 2.5386, 2.8260, 2.4505, 2.3010], device='cuda:0'), covar=tensor([0.0968, 0.0282, 0.0304, 0.1290, 0.0171, 0.0265, 0.0414, 0.0572], device='cuda:0'), in_proj_covar=tensor([0.0145, 0.0108, 0.0098, 0.0136, 0.0084, 0.0126, 0.0127, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 06:23:47,972 INFO [train.py:904] (0/8) Epoch 26, batch 8600, loss[loss=0.1647, simple_loss=0.2575, pruned_loss=0.03596, over 16308.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2678, pruned_loss=0.04017, over 3057503.61 frames. ], batch size: 35, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:23:50,641 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7426, 5.0516, 4.8589, 4.8587, 4.5676, 4.5930, 4.5120, 5.1479], device='cuda:0'), covar=tensor([0.1305, 0.1001, 0.0945, 0.0886, 0.0881, 0.1148, 0.1220, 0.0914], device='cuda:0'), in_proj_covar=tensor([0.0686, 0.0825, 0.0678, 0.0633, 0.0525, 0.0526, 0.0692, 0.0645], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:23:59,713 INFO [optim.py:368] (0/8) 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 06:24:29,944 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8909, 5.2340, 5.3635, 5.1480, 5.2344, 5.7501, 5.2567, 4.9881], device='cuda:0'), covar=tensor([0.1034, 0.1715, 0.2219, 0.1904, 0.2392, 0.0964, 0.1482, 0.2279], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0608, 0.0672, 0.0497, 0.0660, 0.0698, 0.0527, 0.0668], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 06:25:27,327 INFO [train.py:904] (0/8) Epoch 26, batch 8650, loss[loss=0.152, simple_loss=0.2608, pruned_loss=0.02162, over 16902.00 frames. ], tot_loss[loss=0.1722, simple_loss=0.2659, pruned_loss=0.03922, over 3040755.23 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:27:13,553 INFO [train.py:904] (0/8) Epoch 26, batch 8700, loss[loss=0.1533, simple_loss=0.2537, pruned_loss=0.0264, over 16900.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.263, pruned_loss=0.03784, over 3041081.92 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:27:25,317 INFO [optim.py:368] (0/8) 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 06:28:02,757 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:20,388 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:28:48,928 INFO [train.py:904] (0/8) Epoch 26, batch 8750, loss[loss=0.1828, simple_loss=0.2923, pruned_loss=0.03669, over 16912.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2629, pruned_loss=0.0375, over 3017961.57 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:28:56,973 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2232, 5.2725, 5.0257, 4.6519, 4.7201, 5.1213, 5.0038, 4.7699], device='cuda:0'), covar=tensor([0.0553, 0.0485, 0.0337, 0.0325, 0.1033, 0.0525, 0.0327, 0.0723], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0446, 0.0347, 0.0349, 0.0346, 0.0403, 0.0240, 0.0419], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:29:48,638 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:09,756 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262537.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:30:41,542 INFO [train.py:904] (0/8) Epoch 26, batch 8800, loss[loss=0.1713, simple_loss=0.2577, pruned_loss=0.0424, over 12355.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2617, pruned_loss=0.03651, over 3028447.72 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:30:52,415 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.510e+02 2.093e+02 2.466e+02 2.812e+02 4.660e+02, threshold=4.932e+02, percent-clipped=0.0 2023-05-02 06:31:48,621 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.53 vs. limit=5.0 2023-05-02 06:32:25,988 INFO [train.py:904] (0/8) Epoch 26, batch 8850, loss[loss=0.1719, simple_loss=0.2771, pruned_loss=0.03333, over 16665.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2641, pruned_loss=0.03611, over 3023092.44 frames. ], batch size: 134, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:33:32,321 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=262633.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:34:13,867 INFO [train.py:904] (0/8) Epoch 26, batch 8900, loss[loss=0.1555, simple_loss=0.2598, pruned_loss=0.02561, over 16870.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2652, pruned_loss=0.0356, over 3042627.85 frames. ], batch size: 96, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:34:26,823 INFO [optim.py:368] (0/8) 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 06:34:35,537 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 06:34:41,469 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0573, 2.0942, 2.5630, 2.9782, 2.7843, 3.4253, 2.3199, 3.3697], device='cuda:0'), covar=tensor([0.0206, 0.0570, 0.0369, 0.0300, 0.0360, 0.0177, 0.0504, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0192, 0.0179, 0.0183, 0.0199, 0.0157, 0.0196, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:35:19,861 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=262681.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:36:18,593 INFO [train.py:904] (0/8) Epoch 26, batch 8950, loss[loss=0.1636, simple_loss=0.2527, pruned_loss=0.03722, over 12822.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2642, pruned_loss=0.03579, over 3041285.87 frames. ], batch size: 250, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:38:00,286 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=262749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:38:08,102 INFO [train.py:904] (0/8) Epoch 26, batch 9000, loss[loss=0.1448, simple_loss=0.2355, pruned_loss=0.02704, over 12152.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2615, pruned_loss=0.03499, over 3037083.55 frames. ], batch size: 248, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:38:08,104 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 06:38:18,513 INFO [train.py:938] (0/8) Epoch 26, validation: loss=0.1435, simple_loss=0.2475, pruned_loss=0.01976, over 944034.00 frames. 2023-05-02 06:38:18,514 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 06:38:30,750 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.305e+02 2.045e+02 2.519e+02 3.077e+02 5.140e+02, threshold=5.037e+02, percent-clipped=0.0 2023-05-02 06:39:15,306 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5336, 1.7960, 2.1228, 2.4596, 2.4482, 2.8497, 1.8744, 2.7790], device='cuda:0'), covar=tensor([0.0251, 0.0582, 0.0405, 0.0378, 0.0366, 0.0201, 0.0597, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0189, 0.0192, 0.0179, 0.0184, 0.0199, 0.0157, 0.0196, 0.0157], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:40:03,322 INFO [train.py:904] (0/8) Epoch 26, batch 9050, loss[loss=0.1578, simple_loss=0.2535, pruned_loss=0.03105, over 16868.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2628, pruned_loss=0.03536, over 3055645.98 frames. ], batch size: 102, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:40:19,869 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=262810.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:41:48,718 INFO [train.py:904] (0/8) Epoch 26, batch 9100, loss[loss=0.1707, simple_loss=0.2684, pruned_loss=0.03654, over 16539.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2621, pruned_loss=0.03565, over 3069152.89 frames. ], batch size: 75, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:42:01,226 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.532e+02 2.262e+02 2.654e+02 3.227e+02 6.240e+02, threshold=5.309e+02, percent-clipped=4.0 2023-05-02 06:43:45,862 INFO [train.py:904] (0/8) Epoch 26, batch 9150, loss[loss=0.163, simple_loss=0.2597, pruned_loss=0.03311, over 16350.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2632, pruned_loss=0.03548, over 3075990.32 frames. ], batch size: 146, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:45:27,082 INFO [train.py:904] (0/8) Epoch 26, batch 9200, loss[loss=0.1844, simple_loss=0.2686, pruned_loss=0.05012, over 16786.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2592, pruned_loss=0.03477, over 3073803.26 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:45:36,601 INFO [optim.py:368] (0/8) 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 06:47:01,960 INFO [train.py:904] (0/8) Epoch 26, batch 9250, loss[loss=0.1635, simple_loss=0.2502, pruned_loss=0.03841, over 16569.00 frames. ], tot_loss[loss=0.164, simple_loss=0.259, pruned_loss=0.03452, over 3090660.18 frames. ], batch size: 62, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:47:06,619 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 06:48:12,674 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-05-02 06:48:19,142 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2996, 4.2607, 4.6459, 4.6089, 4.6138, 4.3595, 4.3327, 4.3569], device='cuda:0'), covar=tensor([0.0433, 0.1356, 0.0539, 0.0537, 0.0658, 0.0599, 0.0988, 0.0538], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0465, 0.0452, 0.0414, 0.0500, 0.0476, 0.0546, 0.0380], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 06:48:49,723 INFO [train.py:904] (0/8) Epoch 26, batch 9300, loss[loss=0.1579, simple_loss=0.2475, pruned_loss=0.0342, over 15358.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2575, pruned_loss=0.03399, over 3086699.76 frames. ], batch size: 190, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:49:02,022 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.028e+02 2.420e+02 3.075e+02 6.199e+02, threshold=4.840e+02, percent-clipped=1.0 2023-05-02 06:50:20,281 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:33,221 INFO [train.py:904] (0/8) Epoch 26, batch 9350, loss[loss=0.1655, simple_loss=0.2584, pruned_loss=0.03632, over 16601.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2581, pruned_loss=0.03417, over 3106389.79 frames. ], batch size: 62, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:50:38,190 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263105.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:50:49,213 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6384, 3.7059, 3.5145, 3.2293, 3.3167, 3.5877, 3.3966, 3.4858], device='cuda:0'), covar=tensor([0.0581, 0.0660, 0.0372, 0.0275, 0.0567, 0.0520, 0.1303, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0441, 0.0345, 0.0347, 0.0342, 0.0399, 0.0238, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:50:51,079 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:51:36,810 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 06:51:48,028 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:51:54,053 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-05-02 06:52:11,661 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 06:52:14,171 INFO [train.py:904] (0/8) Epoch 26, batch 9400, loss[loss=0.1531, simple_loss=0.235, pruned_loss=0.03558, over 12215.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2577, pruned_loss=0.03419, over 3073434.40 frames. ], batch size: 249, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:52:19,885 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263156.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:25,038 INFO [optim.py:368] (0/8) 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 06:52:41,867 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4558, 2.0406, 1.7458, 1.7239, 2.3082, 1.9441, 1.7945, 2.3526], device='cuda:0'), covar=tensor([0.0188, 0.0444, 0.0615, 0.0527, 0.0327, 0.0405, 0.0183, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0233, 0.0224, 0.0224, 0.0234, 0.0232, 0.0230, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:52:52,019 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:52:52,447 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 06:52:54,524 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 06:53:50,966 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:53:51,415 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.56 vs. limit=2.0 2023-05-02 06:53:53,494 INFO [train.py:904] (0/8) Epoch 26, batch 9450, loss[loss=0.155, simple_loss=0.2528, pruned_loss=0.02859, over 16714.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2597, pruned_loss=0.03424, over 3084685.63 frames. ], batch size: 83, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:54:13,816 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:54:55,564 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1852, 4.0447, 4.2805, 4.3754, 4.5026, 4.0907, 4.4698, 4.5153], device='cuda:0'), covar=tensor([0.1847, 0.1210, 0.1370, 0.0692, 0.0537, 0.1231, 0.0679, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0627, 0.0776, 0.0887, 0.0782, 0.0598, 0.0620, 0.0650, 0.0758], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 06:55:19,923 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263246.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:55:33,116 INFO [train.py:904] (0/8) Epoch 26, batch 9500, loss[loss=0.1534, simple_loss=0.254, pruned_loss=0.02638, over 16120.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2591, pruned_loss=0.03402, over 3081461.53 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:55:47,502 INFO [optim.py:368] (0/8) 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 06:56:19,106 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:56:30,888 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 06:56:33,982 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 06:57:07,848 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 06:57:17,625 INFO [train.py:904] (0/8) Epoch 26, batch 9550, loss[loss=0.1907, simple_loss=0.2869, pruned_loss=0.04723, over 16141.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2584, pruned_loss=0.03417, over 3095137.67 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:57:27,261 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 06:58:51,586 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6385, 4.8436, 4.9811, 4.7506, 4.9249, 5.3822, 4.8590, 4.5660], device='cuda:0'), covar=tensor([0.1197, 0.1921, 0.2419, 0.2106, 0.2237, 0.0895, 0.1581, 0.2531], device='cuda:0'), in_proj_covar=tensor([0.0401, 0.0594, 0.0657, 0.0484, 0.0643, 0.0681, 0.0514, 0.0649], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 06:58:58,578 INFO [train.py:904] (0/8) Epoch 26, batch 9600, loss[loss=0.1717, simple_loss=0.2744, pruned_loss=0.03447, over 16166.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2592, pruned_loss=0.03466, over 3067593.04 frames. ], batch size: 165, lr: 2.57e-03, grad_scale: 8.0 2023-05-02 06:59:09,911 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.689e+02 2.249e+02 2.699e+02 3.019e+02 5.664e+02, threshold=5.399e+02, percent-clipped=4.0 2023-05-02 06:59:43,003 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 07:00:01,932 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 07:00:45,662 INFO [train.py:904] (0/8) Epoch 26, batch 9650, loss[loss=0.1638, simple_loss=0.2579, pruned_loss=0.03487, over 12323.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2616, pruned_loss=0.03512, over 3072737.42 frames. ], batch size: 248, lr: 2.56e-03, grad_scale: 8.0 2023-05-02 07:00:51,844 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:28,453 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263451.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:31,842 INFO [train.py:904] (0/8) Epoch 26, batch 9700, loss[loss=0.1601, simple_loss=0.2658, pruned_loss=0.02719, over 16934.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2603, pruned_loss=0.03482, over 3075298.57 frames. ], batch size: 102, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:02:33,072 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:02:42,011 INFO [optim.py:368] (0/8) 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 07:02:46,877 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7963, 5.0527, 5.1021, 4.9421, 5.0294, 5.5189, 5.0039, 4.7096], device='cuda:0'), covar=tensor([0.1003, 0.1906, 0.2457, 0.1941, 0.2288, 0.0838, 0.1488, 0.2314], device='cuda:0'), in_proj_covar=tensor([0.0400, 0.0594, 0.0659, 0.0484, 0.0643, 0.0681, 0.0514, 0.0648], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:02:58,813 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263467.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:03:33,025 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 07:04:01,229 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:04:13,693 INFO [train.py:904] (0/8) Epoch 26, batch 9750, loss[loss=0.166, simple_loss=0.2455, pruned_loss=0.04325, over 11893.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2598, pruned_loss=0.03536, over 3080361.87 frames. ], batch size: 246, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:05:51,433 INFO [train.py:904] (0/8) Epoch 26, batch 9800, loss[loss=0.1563, simple_loss=0.2708, pruned_loss=0.0209, over 16571.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2596, pruned_loss=0.03453, over 3082182.13 frames. ], batch size: 75, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:06:03,263 INFO [optim.py:368] (0/8) 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 07:06:23,919 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:06:33,040 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 07:07:32,699 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=263602.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:07:34,240 INFO [train.py:904] (0/8) Epoch 26, batch 9850, loss[loss=0.1592, simple_loss=0.2596, pruned_loss=0.02938, over 16622.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2607, pruned_loss=0.03416, over 3085089.93 frames. ], batch size: 89, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:24,089 INFO [train.py:904] (0/8) Epoch 26, batch 9900, loss[loss=0.1668, simple_loss=0.2649, pruned_loss=0.03429, over 15236.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2609, pruned_loss=0.03418, over 3061525.31 frames. ], batch size: 191, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:09:35,600 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5214, 4.4918, 4.3078, 3.6960, 4.3893, 1.7465, 4.1601, 4.0258], device='cuda:0'), covar=tensor([0.0090, 0.0097, 0.0217, 0.0289, 0.0108, 0.2822, 0.0151, 0.0304], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0163, 0.0199, 0.0173, 0.0177, 0.0208, 0.0188, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:09:36,803 INFO [optim.py:368] (0/8) 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 07:11:20,749 INFO [train.py:904] (0/8) Epoch 26, batch 9950, loss[loss=0.1798, simple_loss=0.2672, pruned_loss=0.04615, over 12245.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2627, pruned_loss=0.03435, over 3052610.01 frames. ], batch size: 248, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:18,990 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263751.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:13:21,149 INFO [train.py:904] (0/8) Epoch 26, batch 10000, loss[loss=0.1633, simple_loss=0.2669, pruned_loss=0.02982, over 16194.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.262, pruned_loss=0.03422, over 3070294.70 frames. ], batch size: 165, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:13:34,913 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.519e+02 2.135e+02 2.379e+02 2.761e+02 5.378e+02, threshold=4.758e+02, percent-clipped=3.0 2023-05-02 07:13:51,315 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:14:15,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4741, 3.5750, 2.8020, 2.1987, 2.2307, 2.3838, 3.7720, 3.1220], device='cuda:0'), covar=tensor([0.3124, 0.0569, 0.1736, 0.3033, 0.2943, 0.2227, 0.0397, 0.1301], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0266, 0.0303, 0.0315, 0.0290, 0.0265, 0.0294, 0.0337], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:14:48,326 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:00,044 INFO [train.py:904] (0/8) Epoch 26, batch 10050, loss[loss=0.1579, simple_loss=0.2546, pruned_loss=0.03067, over 12143.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.262, pruned_loss=0.03408, over 3069342.12 frames. ], batch size: 247, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:15:23,652 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263815.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:15:57,778 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263833.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:16:17,283 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:16:30,289 INFO [train.py:904] (0/8) Epoch 26, batch 10100, loss[loss=0.149, simple_loss=0.2464, pruned_loss=0.02586, over 16698.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2627, pruned_loss=0.03449, over 3063451.90 frames. ], batch size: 89, lr: 2.56e-03, grad_scale: 16.0 2023-05-02 07:16:39,587 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.545e+02 2.128e+02 2.539e+02 3.188e+02 8.403e+02, threshold=5.079e+02, percent-clipped=3.0 2023-05-02 07:17:02,798 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:17:36,324 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=263894.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 07:17:46,585 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-26.pt 2023-05-02 07:18:08,998 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=263902.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:18:09,866 INFO [train.py:904] (0/8) Epoch 27, batch 0, loss[loss=0.1658, simple_loss=0.2525, pruned_loss=0.03959, over 17163.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2525, pruned_loss=0.03959, over 17163.00 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 16.0 2023-05-02 07:18:09,866 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 07:18:17,112 INFO [train.py:938] (0/8) Epoch 27, validation: loss=0.1434, simple_loss=0.2467, pruned_loss=0.02006, over 944034.00 frames. 2023-05-02 07:18:17,113 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 07:18:30,557 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9439, 4.6958, 4.9944, 5.1489, 5.3136, 4.6292, 5.2713, 5.2996], device='cuda:0'), covar=tensor([0.2010, 0.1428, 0.1645, 0.0847, 0.0659, 0.0995, 0.0629, 0.0734], device='cuda:0'), in_proj_covar=tensor([0.0631, 0.0778, 0.0891, 0.0788, 0.0600, 0.0623, 0.0654, 0.0760], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:18:37,958 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:18:49,599 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6696, 4.6405, 4.9513, 4.9476, 5.0180, 4.7221, 4.6917, 4.5150], device='cuda:0'), covar=tensor([0.0526, 0.0930, 0.0545, 0.0775, 0.0717, 0.0650, 0.1305, 0.0746], device='cuda:0'), in_proj_covar=tensor([0.0404, 0.0457, 0.0444, 0.0408, 0.0492, 0.0467, 0.0537, 0.0374], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 07:18:58,370 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1448, 3.2125, 3.4606, 1.8481, 3.6430, 3.7258, 2.8337, 2.5501], device='cuda:0'), covar=tensor([0.1221, 0.0267, 0.0246, 0.1401, 0.0141, 0.0201, 0.0561, 0.0662], device='cuda:0'), in_proj_covar=tensor([0.0143, 0.0106, 0.0094, 0.0134, 0.0082, 0.0123, 0.0125, 0.0125], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 07:19:22,227 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=263950.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:26,181 INFO [train.py:904] (0/8) Epoch 27, batch 50, loss[loss=0.1436, simple_loss=0.2346, pruned_loss=0.02631, over 17208.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2644, pruned_loss=0.04587, over 747449.87 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:19:36,271 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=263960.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:19:39,892 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.543e+02 2.464e+02 3.043e+02 3.721e+02 7.050e+02, threshold=6.086e+02, percent-clipped=6.0 2023-05-02 07:19:56,152 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7641, 4.7114, 4.6017, 4.0750, 4.6606, 1.8827, 4.3879, 4.3337], device='cuda:0'), covar=tensor([0.0193, 0.0147, 0.0283, 0.0390, 0.0169, 0.2971, 0.0225, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0164, 0.0200, 0.0174, 0.0178, 0.0209, 0.0190, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:20:29,200 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-264000.pt 2023-05-02 07:20:36,110 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 07:20:36,415 INFO [train.py:904] (0/8) Epoch 27, batch 100, loss[loss=0.1846, simple_loss=0.2745, pruned_loss=0.04737, over 16773.00 frames. ], tot_loss[loss=0.1755, simple_loss=0.262, pruned_loss=0.04449, over 1315184.33 frames. ], batch size: 57, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:00,907 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:21:02,348 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 07:21:12,873 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1026, 3.0501, 3.0054, 5.2624, 4.4597, 4.5837, 1.7354, 3.4456], device='cuda:0'), covar=tensor([0.1320, 0.0775, 0.1108, 0.0177, 0.0230, 0.0374, 0.1654, 0.0704], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0176, 0.0195, 0.0192, 0.0199, 0.0213, 0.0205, 0.0193], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:21:44,757 INFO [train.py:904] (0/8) Epoch 27, batch 150, loss[loss=0.1812, simple_loss=0.2621, pruned_loss=0.05012, over 16886.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2606, pruned_loss=0.04301, over 1762902.91 frames. ], batch size: 116, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:21:51,767 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 07:21:57,551 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.626e+02 2.228e+02 2.541e+02 3.123e+02 6.802e+02, threshold=5.083e+02, percent-clipped=1.0 2023-05-02 07:22:53,640 INFO [train.py:904] (0/8) Epoch 27, batch 200, loss[loss=0.1852, simple_loss=0.2695, pruned_loss=0.05047, over 16462.00 frames. ], tot_loss[loss=0.173, simple_loss=0.2603, pruned_loss=0.04286, over 2109360.08 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:23:30,212 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8455, 3.9546, 2.7636, 4.5464, 3.3327, 4.5303, 2.7492, 3.3802], device='cuda:0'), covar=tensor([0.0356, 0.0491, 0.1529, 0.0473, 0.0803, 0.0656, 0.1538, 0.0801], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0176, 0.0192, 0.0167, 0.0176, 0.0213, 0.0202, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:23:30,319 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4970, 2.4330, 2.3854, 4.3739, 2.4606, 2.7438, 2.4781, 2.5754], device='cuda:0'), covar=tensor([0.1327, 0.3738, 0.3384, 0.0526, 0.4258, 0.2682, 0.3688, 0.3669], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0464, 0.0381, 0.0329, 0.0440, 0.0528, 0.0435, 0.0540], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:23:35,030 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7318, 4.2978, 2.9529, 2.3617, 2.6437, 2.5204, 4.7110, 3.5286], device='cuda:0'), covar=tensor([0.3372, 0.0719, 0.2080, 0.3093, 0.3099, 0.2358, 0.0361, 0.1630], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0271, 0.0309, 0.0320, 0.0297, 0.0271, 0.0300, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:23:45,598 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3216, 2.9168, 2.6558, 2.2115, 2.2095, 2.3189, 3.0022, 2.7768], device='cuda:0'), covar=tensor([0.2729, 0.0835, 0.1742, 0.2700, 0.2518, 0.2133, 0.0594, 0.1345], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0271, 0.0309, 0.0320, 0.0297, 0.0271, 0.0300, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:24:00,843 INFO [train.py:904] (0/8) Epoch 27, batch 250, loss[loss=0.1599, simple_loss=0.262, pruned_loss=0.02891, over 17061.00 frames. ], tot_loss[loss=0.1717, simple_loss=0.2579, pruned_loss=0.04269, over 2380800.98 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:24:14,013 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.539e+02 2.137e+02 2.566e+02 3.283e+02 9.180e+02, threshold=5.132e+02, percent-clipped=3.0 2023-05-02 07:24:50,579 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264189.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:25:10,475 INFO [train.py:904] (0/8) Epoch 27, batch 300, loss[loss=0.1706, simple_loss=0.2636, pruned_loss=0.03875, over 17044.00 frames. ], tot_loss[loss=0.1701, simple_loss=0.2559, pruned_loss=0.04211, over 2583172.85 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:25:48,475 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 07:25:52,970 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:26:19,426 INFO [train.py:904] (0/8) Epoch 27, batch 350, loss[loss=0.1687, simple_loss=0.2387, pruned_loss=0.0494, over 16910.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2534, pruned_loss=0.0408, over 2742347.53 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 1.0 2023-05-02 07:26:34,330 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.322e+02 2.070e+02 2.475e+02 2.951e+02 5.112e+02, threshold=4.951e+02, percent-clipped=0.0 2023-05-02 07:26:41,147 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8850, 4.4243, 3.1580, 2.3681, 2.6755, 2.6969, 4.7503, 3.5017], device='cuda:0'), covar=tensor([0.2931, 0.0570, 0.1892, 0.3204, 0.3097, 0.2193, 0.0337, 0.1632], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0272, 0.0311, 0.0322, 0.0299, 0.0273, 0.0301, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:27:12,101 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8848, 4.0543, 2.8383, 4.7386, 3.3696, 4.6075, 2.8053, 3.4474], device='cuda:0'), covar=tensor([0.0326, 0.0436, 0.1384, 0.0258, 0.0778, 0.0524, 0.1390, 0.0702], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0170, 0.0179, 0.0217, 0.0204, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:27:17,352 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:27:23,538 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.42 vs. limit=2.0 2023-05-02 07:27:27,696 INFO [train.py:904] (0/8) Epoch 27, batch 400, loss[loss=0.2038, simple_loss=0.2735, pruned_loss=0.06709, over 16854.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.253, pruned_loss=0.04112, over 2864241.17 frames. ], batch size: 116, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:27:44,539 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264316.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:28:28,155 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6242, 3.6616, 2.3059, 3.8805, 3.0432, 3.7740, 2.3975, 3.0018], device='cuda:0'), covar=tensor([0.0276, 0.0448, 0.1560, 0.0402, 0.0726, 0.0889, 0.1452, 0.0709], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0170, 0.0179, 0.0217, 0.0205, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:28:33,921 INFO [train.py:904] (0/8) Epoch 27, batch 450, loss[loss=0.1484, simple_loss=0.2371, pruned_loss=0.02986, over 16498.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2514, pruned_loss=0.04011, over 2962929.09 frames. ], batch size: 75, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:28:34,522 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7803, 3.8219, 2.9049, 2.2908, 2.4104, 2.4323, 3.9060, 3.2578], device='cuda:0'), covar=tensor([0.2690, 0.0582, 0.1771, 0.3197, 0.2923, 0.2289, 0.0471, 0.1599], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0273, 0.0312, 0.0323, 0.0300, 0.0273, 0.0302, 0.0348], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:28:48,388 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.223e+02 2.597e+02 3.088e+02 6.420e+02, threshold=5.195e+02, percent-clipped=1.0 2023-05-02 07:29:44,916 INFO [train.py:904] (0/8) Epoch 27, batch 500, loss[loss=0.1633, simple_loss=0.2473, pruned_loss=0.03965, over 16462.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2502, pruned_loss=0.03954, over 3029028.96 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:30:24,743 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8342, 4.0431, 2.7386, 4.5871, 3.3397, 4.4672, 2.6709, 3.4191], device='cuda:0'), covar=tensor([0.0344, 0.0372, 0.1463, 0.0310, 0.0737, 0.0710, 0.1470, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0179, 0.0196, 0.0171, 0.0179, 0.0218, 0.0205, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:30:51,108 INFO [train.py:904] (0/8) Epoch 27, batch 550, loss[loss=0.173, simple_loss=0.2496, pruned_loss=0.04815, over 12380.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2492, pruned_loss=0.03866, over 3100757.74 frames. ], batch size: 246, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:31:04,229 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.059e+02 2.347e+02 2.787e+02 4.742e+02, threshold=4.693e+02, percent-clipped=0.0 2023-05-02 07:31:18,787 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5411, 3.5111, 3.4842, 2.8522, 3.3476, 2.1169, 3.2090, 2.7883], device='cuda:0'), covar=tensor([0.0201, 0.0191, 0.0193, 0.0253, 0.0119, 0.2400, 0.0163, 0.0292], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0180, 0.0184, 0.0216, 0.0197, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:31:26,704 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264479.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:31:40,816 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264489.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:31:55,557 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264500.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:31:59,354 INFO [train.py:904] (0/8) Epoch 27, batch 600, loss[loss=0.1401, simple_loss=0.2372, pruned_loss=0.02154, over 17192.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2491, pruned_loss=0.03884, over 3150869.95 frames. ], batch size: 46, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:32:04,370 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 07:32:46,463 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264537.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:32:50,474 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:32:58,160 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8882, 3.4353, 3.9115, 2.1746, 4.0227, 4.0324, 3.3199, 2.9898], device='cuda:0'), covar=tensor([0.0730, 0.0306, 0.0194, 0.1197, 0.0117, 0.0215, 0.0387, 0.0479], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0111, 0.0099, 0.0139, 0.0085, 0.0129, 0.0130, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:33:08,807 INFO [train.py:904] (0/8) Epoch 27, batch 650, loss[loss=0.16, simple_loss=0.2414, pruned_loss=0.03933, over 15440.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2481, pruned_loss=0.0381, over 3191814.48 frames. ], batch size: 190, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:33:17,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6378, 3.6742, 3.9147, 2.7365, 3.6182, 4.0308, 3.6966, 2.4100], device='cuda:0'), covar=tensor([0.0518, 0.0335, 0.0072, 0.0398, 0.0128, 0.0114, 0.0118, 0.0506], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0088, 0.0089, 0.0134, 0.0100, 0.0112, 0.0096, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 07:33:18,877 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264561.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:33:20,708 INFO [optim.py:368] (0/8) 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 07:33:58,326 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:34:15,999 INFO [train.py:904] (0/8) Epoch 27, batch 700, loss[loss=0.1824, simple_loss=0.2658, pruned_loss=0.04956, over 15629.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2481, pruned_loss=0.03786, over 3224037.62 frames. ], batch size: 191, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:34:34,338 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264616.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:35:22,118 INFO [train.py:904] (0/8) Epoch 27, batch 750, loss[loss=0.1604, simple_loss=0.2541, pruned_loss=0.03331, over 17110.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2481, pruned_loss=0.03767, over 3244030.72 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 2.0 2023-05-02 07:35:35,424 INFO [optim.py:368] (0/8) 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 07:35:37,455 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264664.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:36:28,323 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9557, 4.9121, 4.7720, 4.2587, 4.8728, 2.0946, 4.6214, 4.4592], device='cuda:0'), covar=tensor([0.0141, 0.0117, 0.0217, 0.0372, 0.0115, 0.2655, 0.0151, 0.0268], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0169, 0.0205, 0.0179, 0.0183, 0.0214, 0.0195, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:36:29,180 INFO [train.py:904] (0/8) Epoch 27, batch 800, loss[loss=0.1578, simple_loss=0.2441, pruned_loss=0.03572, over 16615.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2481, pruned_loss=0.0377, over 3261014.47 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:36:51,823 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0696, 4.0970, 3.9707, 3.7014, 3.7468, 4.0451, 3.7420, 3.8682], device='cuda:0'), covar=tensor([0.0630, 0.0831, 0.0361, 0.0323, 0.0725, 0.0538, 0.1040, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0463, 0.0361, 0.0364, 0.0361, 0.0420, 0.0249, 0.0435], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:37:36,930 INFO [train.py:904] (0/8) Epoch 27, batch 850, loss[loss=0.1571, simple_loss=0.2557, pruned_loss=0.02925, over 17284.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2474, pruned_loss=0.03788, over 3253453.37 frames. ], batch size: 52, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:37:42,670 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4844, 5.9062, 5.6412, 5.7148, 5.3248, 5.3382, 5.2755, 6.0509], device='cuda:0'), covar=tensor([0.1560, 0.1133, 0.1171, 0.0951, 0.1054, 0.0795, 0.1540, 0.1028], device='cuda:0'), in_proj_covar=tensor([0.0711, 0.0856, 0.0701, 0.0660, 0.0547, 0.0544, 0.0725, 0.0671], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:37:51,806 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.434e+02 2.032e+02 2.277e+02 2.694e+02 3.998e+02, threshold=4.553e+02, percent-clipped=0.0 2023-05-02 07:38:44,875 INFO [train.py:904] (0/8) Epoch 27, batch 900, loss[loss=0.1362, simple_loss=0.226, pruned_loss=0.02325, over 16758.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2467, pruned_loss=0.03746, over 3271237.30 frames. ], batch size: 39, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:38:47,201 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=264804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:28,082 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:39:51,356 INFO [train.py:904] (0/8) Epoch 27, batch 950, loss[loss=0.1667, simple_loss=0.2467, pruned_loss=0.04332, over 16574.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2474, pruned_loss=0.03769, over 3292133.88 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:39:55,580 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=264856.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 07:40:04,152 INFO [optim.py:368] (0/8) 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 07:40:07,705 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=264865.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:15,613 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 07:40:41,680 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=264890.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:40:45,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7287, 3.7464, 2.4071, 4.0263, 3.0694, 3.9741, 2.4481, 3.1077], device='cuda:0'), covar=tensor([0.0283, 0.0502, 0.1626, 0.0417, 0.0739, 0.0846, 0.1615, 0.0732], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0182, 0.0198, 0.0174, 0.0182, 0.0222, 0.0208, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:40:57,200 INFO [train.py:904] (0/8) Epoch 27, batch 1000, loss[loss=0.1359, simple_loss=0.2256, pruned_loss=0.02311, over 16822.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2463, pruned_loss=0.03716, over 3293791.04 frames. ], batch size: 42, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:41:43,549 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4808, 3.5614, 3.9148, 2.0071, 4.0830, 4.1859, 3.2040, 2.8518], device='cuda:0'), covar=tensor([0.1193, 0.0281, 0.0235, 0.1432, 0.0117, 0.0234, 0.0493, 0.0686], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0140, 0.0086, 0.0131, 0.0131, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 07:41:46,842 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=264938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:42:07,410 INFO [train.py:904] (0/8) Epoch 27, batch 1050, loss[loss=0.1579, simple_loss=0.2372, pruned_loss=0.03929, over 16726.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2457, pruned_loss=0.03749, over 3295625.12 frames. ], batch size: 134, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:42:19,712 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.472e+02 2.065e+02 2.389e+02 2.901e+02 6.221e+02, threshold=4.777e+02, percent-clipped=3.0 2023-05-02 07:42:38,211 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 07:43:16,445 INFO [train.py:904] (0/8) Epoch 27, batch 1100, loss[loss=0.1626, simple_loss=0.2578, pruned_loss=0.03371, over 17052.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2454, pruned_loss=0.03699, over 3309345.30 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:43:43,141 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.87 vs. limit=2.0 2023-05-02 07:44:10,285 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6079, 2.7992, 2.4065, 2.5765, 3.0233, 2.7334, 3.1516, 3.2300], device='cuda:0'), covar=tensor([0.0235, 0.0480, 0.0611, 0.0532, 0.0359, 0.0479, 0.0313, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0230, 0.0247, 0.0236, 0.0235, 0.0247, 0.0246, 0.0245, 0.0244], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:44:25,483 INFO [train.py:904] (0/8) Epoch 27, batch 1150, loss[loss=0.1628, simple_loss=0.2387, pruned_loss=0.04345, over 16930.00 frames. ], tot_loss[loss=0.1589, simple_loss=0.2448, pruned_loss=0.03653, over 3312415.37 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 4.0 2023-05-02 07:44:39,253 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.259e+02 1.949e+02 2.374e+02 2.990e+02 6.257e+02, threshold=4.748e+02, percent-clipped=3.0 2023-05-02 07:45:34,345 INFO [train.py:904] (0/8) Epoch 27, batch 1200, loss[loss=0.1605, simple_loss=0.2521, pruned_loss=0.03446, over 17044.00 frames. ], tot_loss[loss=0.158, simple_loss=0.244, pruned_loss=0.03603, over 3307384.03 frames. ], batch size: 55, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:45:51,907 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7512, 2.7866, 3.0791, 2.0524, 2.7322, 2.0930, 3.2826, 3.2257], device='cuda:0'), covar=tensor([0.0271, 0.1163, 0.0688, 0.2092, 0.0946, 0.1161, 0.0641, 0.1074], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0169, 0.0170, 0.0157, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 07:46:19,341 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265135.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:46:34,112 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8443, 4.4179, 3.1167, 2.3944, 2.6458, 2.6591, 4.8259, 3.5419], device='cuda:0'), covar=tensor([0.3051, 0.0566, 0.1877, 0.3144, 0.3230, 0.2252, 0.0312, 0.1520], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0274, 0.0311, 0.0325, 0.0302, 0.0274, 0.0303, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:46:42,970 INFO [train.py:904] (0/8) Epoch 27, batch 1250, loss[loss=0.1634, simple_loss=0.2367, pruned_loss=0.04506, over 16809.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2439, pruned_loss=0.03605, over 3304206.86 frames. ], batch size: 102, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:46:48,423 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265156.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 07:46:53,484 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:46:57,312 INFO [optim.py:368] (0/8) 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 07:47:11,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2776, 5.2300, 5.0205, 4.4603, 5.0527, 2.0424, 4.7894, 4.8808], device='cuda:0'), covar=tensor([0.0091, 0.0080, 0.0237, 0.0418, 0.0110, 0.2720, 0.0163, 0.0242], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0173, 0.0210, 0.0183, 0.0186, 0.0218, 0.0200, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:47:25,263 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265183.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:47:53,013 INFO [train.py:904] (0/8) Epoch 27, batch 1300, loss[loss=0.1363, simple_loss=0.2267, pruned_loss=0.02298, over 17210.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2437, pruned_loss=0.0355, over 3312103.64 frames. ], batch size: 44, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:47:54,314 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265204.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 07:48:15,502 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0962, 5.4343, 5.2001, 5.2270, 4.9530, 4.9344, 4.8233, 5.5583], device='cuda:0'), covar=tensor([0.1401, 0.0973, 0.1089, 0.0940, 0.0868, 0.0973, 0.1509, 0.0897], device='cuda:0'), in_proj_covar=tensor([0.0721, 0.0867, 0.0710, 0.0669, 0.0553, 0.0549, 0.0736, 0.0681], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:49:00,611 INFO [train.py:904] (0/8) Epoch 27, batch 1350, loss[loss=0.1667, simple_loss=0.2616, pruned_loss=0.03591, over 17037.00 frames. ], tot_loss[loss=0.1575, simple_loss=0.244, pruned_loss=0.0355, over 3292734.58 frames. ], batch size: 53, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:49:14,445 INFO [optim.py:368] (0/8) 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 07:49:35,251 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:50:07,774 INFO [train.py:904] (0/8) Epoch 27, batch 1400, loss[loss=0.1868, simple_loss=0.261, pruned_loss=0.05631, over 16890.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2438, pruned_loss=0.03573, over 3301632.04 frames. ], batch size: 109, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:50:26,290 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5434, 4.8593, 5.1865, 5.1375, 5.1754, 4.8562, 4.5474, 4.6670], device='cuda:0'), covar=tensor([0.0710, 0.0829, 0.0649, 0.0661, 0.0740, 0.0657, 0.1502, 0.0646], device='cuda:0'), in_proj_covar=tensor([0.0432, 0.0490, 0.0471, 0.0433, 0.0523, 0.0500, 0.0574, 0.0398], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 07:50:56,882 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:51:10,079 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1876, 5.1830, 4.8979, 4.3614, 5.0657, 1.8461, 4.7543, 4.5553], device='cuda:0'), covar=tensor([0.0107, 0.0085, 0.0236, 0.0408, 0.0101, 0.3184, 0.0149, 0.0317], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0173, 0.0209, 0.0184, 0.0187, 0.0218, 0.0200, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:51:15,554 INFO [train.py:904] (0/8) Epoch 27, batch 1450, loss[loss=0.1383, simple_loss=0.2255, pruned_loss=0.02556, over 17235.00 frames. ], tot_loss[loss=0.1569, simple_loss=0.2424, pruned_loss=0.03567, over 3297807.89 frames. ], batch size: 43, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:51:22,651 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.15 vs. limit=5.0 2023-05-02 07:51:29,672 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.640e+02 2.111e+02 2.470e+02 3.011e+02 5.399e+02, threshold=4.940e+02, percent-clipped=1.0 2023-05-02 07:52:24,303 INFO [train.py:904] (0/8) Epoch 27, batch 1500, loss[loss=0.1579, simple_loss=0.2575, pruned_loss=0.02919, over 17138.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.2423, pruned_loss=0.03594, over 3303889.02 frames. ], batch size: 49, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:52:32,771 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8941, 3.8417, 3.9046, 3.6782, 3.8511, 4.3105, 3.8920, 3.5337], device='cuda:0'), covar=tensor([0.1986, 0.2472, 0.2644, 0.2505, 0.2877, 0.1997, 0.1841, 0.2879], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0638, 0.0707, 0.0525, 0.0693, 0.0733, 0.0550, 0.0698], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:53:10,901 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2503, 2.7451, 2.2284, 2.4883, 3.0612, 2.7550, 3.0992, 3.1214], device='cuda:0'), covar=tensor([0.0242, 0.0442, 0.0586, 0.0480, 0.0277, 0.0415, 0.0268, 0.0344], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0247, 0.0236, 0.0236, 0.0248, 0.0247, 0.0246, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:53:36,985 INFO [train.py:904] (0/8) Epoch 27, batch 1550, loss[loss=0.1933, simple_loss=0.2677, pruned_loss=0.05946, over 16429.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2434, pruned_loss=0.03655, over 3301362.25 frames. ], batch size: 68, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:53:45,916 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265460.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:53:49,835 INFO [optim.py:368] (0/8) 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 07:54:23,888 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1856, 5.9244, 6.0315, 5.7377, 5.9335, 6.3877, 5.9089, 5.5879], device='cuda:0'), covar=tensor([0.0967, 0.2076, 0.2631, 0.2176, 0.2491, 0.0981, 0.1769, 0.2471], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0637, 0.0707, 0.0524, 0.0694, 0.0732, 0.0550, 0.0697], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:54:43,804 INFO [train.py:904] (0/8) Epoch 27, batch 1600, loss[loss=0.1625, simple_loss=0.2602, pruned_loss=0.03239, over 17026.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2459, pruned_loss=0.0373, over 3306891.77 frames. ], batch size: 50, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:54:51,422 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:21,311 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:55:51,086 INFO [train.py:904] (0/8) Epoch 27, batch 1650, loss[loss=0.1674, simple_loss=0.2621, pruned_loss=0.03638, over 17119.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2477, pruned_loss=0.03798, over 3316444.05 frames. ], batch size: 47, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:56:04,417 INFO [optim.py:368] (0/8) 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 07:56:43,921 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:57:00,236 INFO [train.py:904] (0/8) Epoch 27, batch 1700, loss[loss=0.1799, simple_loss=0.2784, pruned_loss=0.04076, over 17218.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2491, pruned_loss=0.03818, over 3317981.44 frames. ], batch size: 45, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:57:07,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0568, 5.5680, 5.7644, 5.4824, 5.5831, 6.1410, 5.6037, 5.3328], device='cuda:0'), covar=tensor([0.0986, 0.2137, 0.2414, 0.1928, 0.2432, 0.0948, 0.1643, 0.2356], device='cuda:0'), in_proj_covar=tensor([0.0429, 0.0637, 0.0705, 0.0523, 0.0692, 0.0730, 0.0549, 0.0695], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 07:57:15,984 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:57:42,668 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265634.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 07:58:07,540 INFO [train.py:904] (0/8) Epoch 27, batch 1750, loss[loss=0.2234, simple_loss=0.3027, pruned_loss=0.07208, over 11946.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2513, pruned_loss=0.03899, over 3302598.68 frames. ], batch size: 247, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:58:22,126 INFO [optim.py:368] (0/8) 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 07:58:38,313 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3395, 4.2024, 4.4099, 4.5345, 4.5941, 4.2616, 4.4479, 4.6240], device='cuda:0'), covar=tensor([0.1644, 0.1265, 0.1344, 0.0708, 0.0602, 0.1195, 0.2783, 0.0866], device='cuda:0'), in_proj_covar=tensor([0.0690, 0.0849, 0.0980, 0.0861, 0.0652, 0.0679, 0.0714, 0.0830], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:58:38,367 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265675.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 07:58:59,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7810, 4.5680, 4.8336, 4.9995, 5.1694, 4.5995, 5.1940, 5.2015], device='cuda:0'), covar=tensor([0.2027, 0.1412, 0.1836, 0.0851, 0.0608, 0.1054, 0.0695, 0.0690], device='cuda:0'), in_proj_covar=tensor([0.0691, 0.0850, 0.0982, 0.0862, 0.0653, 0.0680, 0.0714, 0.0832], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 07:59:15,662 INFO [train.py:904] (0/8) Epoch 27, batch 1800, loss[loss=0.1761, simple_loss=0.2617, pruned_loss=0.04528, over 16531.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2533, pruned_loss=0.03937, over 3308652.30 frames. ], batch size: 146, lr: 2.51e-03, grad_scale: 8.0 2023-05-02 07:59:32,471 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:00:23,855 INFO [train.py:904] (0/8) Epoch 27, batch 1850, loss[loss=0.1743, simple_loss=0.2658, pruned_loss=0.04136, over 16767.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2542, pruned_loss=0.03977, over 3305926.24 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:00:37,847 INFO [optim.py:368] (0/8) 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 08:00:55,789 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:01:24,260 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9332, 5.2076, 5.4368, 5.1579, 5.2207, 5.8726, 5.3842, 5.0622], device='cuda:0'), covar=tensor([0.1213, 0.2113, 0.2548, 0.2124, 0.2918, 0.1039, 0.1769, 0.2584], device='cuda:0'), in_proj_covar=tensor([0.0433, 0.0640, 0.0709, 0.0525, 0.0697, 0.0735, 0.0552, 0.0699], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:01:30,060 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 08:01:33,286 INFO [train.py:904] (0/8) Epoch 27, batch 1900, loss[loss=0.1573, simple_loss=0.241, pruned_loss=0.03679, over 16170.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2532, pruned_loss=0.03906, over 3311580.29 frames. ], batch size: 165, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:12,968 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4464, 4.5337, 4.8025, 4.7993, 4.8519, 4.5455, 4.5489, 4.4529], device='cuda:0'), covar=tensor([0.0399, 0.0625, 0.0445, 0.0425, 0.0550, 0.0477, 0.0878, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0493, 0.0475, 0.0438, 0.0527, 0.0505, 0.0579, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 08:02:15,540 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-05-02 08:02:22,900 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3206, 5.3301, 5.0361, 4.5397, 5.1635, 2.0151, 4.8664, 4.8850], device='cuda:0'), covar=tensor([0.0096, 0.0080, 0.0230, 0.0408, 0.0104, 0.2980, 0.0152, 0.0259], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0174, 0.0211, 0.0185, 0.0188, 0.0219, 0.0202, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:02:42,107 INFO [train.py:904] (0/8) Epoch 27, batch 1950, loss[loss=0.1722, simple_loss=0.262, pruned_loss=0.04115, over 17042.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2528, pruned_loss=0.03842, over 3315449.04 frames. ], batch size: 55, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:02:42,553 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2247, 1.6987, 2.0662, 2.1295, 2.2985, 2.3791, 1.7954, 2.3859], device='cuda:0'), covar=tensor([0.0259, 0.0504, 0.0299, 0.0355, 0.0339, 0.0319, 0.0563, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0198, 0.0187, 0.0192, 0.0208, 0.0166, 0.0204, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:02:54,914 INFO [optim.py:368] (0/8) 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 08:03:07,583 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265872.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:26,224 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:03:36,943 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 08:03:50,718 INFO [train.py:904] (0/8) Epoch 27, batch 2000, loss[loss=0.1412, simple_loss=0.2417, pruned_loss=0.02034, over 17123.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2525, pruned_loss=0.03833, over 3312125.13 frames. ], batch size: 48, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:04:10,568 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=265917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:33,670 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265933.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:04:34,829 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=265934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:01,221 INFO [train.py:904] (0/8) Epoch 27, batch 2050, loss[loss=0.1893, simple_loss=0.2674, pruned_loss=0.05557, over 16726.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2531, pruned_loss=0.0387, over 3304144.32 frames. ], batch size: 89, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:05:14,205 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.410e+02 2.179e+02 2.495e+02 3.022e+02 6.334e+02, threshold=4.990e+02, percent-clipped=2.0 2023-05-02 08:05:24,888 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=265970.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 08:05:36,800 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=265978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:05:41,864 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=265982.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:06:07,013 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-266000.pt 2023-05-02 08:06:14,069 INFO [train.py:904] (0/8) Epoch 27, batch 2100, loss[loss=0.1701, simple_loss=0.2676, pruned_loss=0.03632, over 16996.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2538, pruned_loss=0.03898, over 3312399.39 frames. ], batch size: 50, lr: 2.50e-03, grad_scale: 16.0 2023-05-02 08:07:05,042 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8354, 3.9306, 2.8726, 2.3061, 2.4995, 2.4366, 4.1810, 3.3381], device='cuda:0'), covar=tensor([0.2858, 0.0672, 0.1932, 0.3280, 0.3139, 0.2396, 0.0513, 0.1701], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0276, 0.0313, 0.0326, 0.0305, 0.0275, 0.0305, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:07:22,100 INFO [train.py:904] (0/8) Epoch 27, batch 2150, loss[loss=0.1692, simple_loss=0.2616, pruned_loss=0.03844, over 17247.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2547, pruned_loss=0.03952, over 3319777.52 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:07:37,035 INFO [optim.py:368] (0/8) 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 08:07:43,223 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.76 vs. limit=5.0 2023-05-02 08:07:46,310 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266071.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:08:32,978 INFO [train.py:904] (0/8) Epoch 27, batch 2200, loss[loss=0.1553, simple_loss=0.2537, pruned_loss=0.02851, over 17128.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2551, pruned_loss=0.03932, over 3323625.54 frames. ], batch size: 47, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:41,318 INFO [train.py:904] (0/8) Epoch 27, batch 2250, loss[loss=0.171, simple_loss=0.2708, pruned_loss=0.03561, over 16751.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2551, pruned_loss=0.03864, over 3332544.82 frames. ], batch size: 57, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:09:56,595 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.643e+02 2.173e+02 2.503e+02 2.978e+02 4.988e+02, threshold=5.005e+02, percent-clipped=0.0 2023-05-02 08:10:27,504 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:10:51,425 INFO [train.py:904] (0/8) Epoch 27, batch 2300, loss[loss=0.1707, simple_loss=0.2501, pruned_loss=0.04563, over 16928.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2549, pruned_loss=0.03907, over 3325231.64 frames. ], batch size: 109, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:11:06,836 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266214.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:23,589 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266226.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:11:25,882 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266228.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:34,303 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:11:46,720 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9879, 4.8889, 4.8751, 4.4178, 4.5472, 4.8974, 4.7944, 4.5855], device='cuda:0'), covar=tensor([0.0702, 0.0990, 0.0361, 0.0448, 0.1027, 0.0606, 0.0432, 0.0863], device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0480, 0.0373, 0.0378, 0.0374, 0.0433, 0.0257, 0.0449], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:11:59,657 INFO [train.py:904] (0/8) Epoch 27, batch 2350, loss[loss=0.1875, simple_loss=0.2622, pruned_loss=0.05641, over 16873.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2546, pruned_loss=0.03978, over 3319318.97 frames. ], batch size: 109, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:12:14,161 INFO [optim.py:368] (0/8) 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 08:12:22,365 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266270.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 08:12:26,015 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2441, 3.4152, 3.7382, 2.1442, 3.0694, 2.3680, 3.6197, 3.6355], device='cuda:0'), covar=tensor([0.0296, 0.0941, 0.0527, 0.2131, 0.0853, 0.1034, 0.0615, 0.1050], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0147, 0.0132, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 08:12:27,937 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:30,401 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:12:46,325 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266287.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 08:13:07,763 INFO [train.py:904] (0/8) Epoch 27, batch 2400, loss[loss=0.1858, simple_loss=0.2606, pruned_loss=0.05554, over 16489.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2558, pruned_loss=0.03977, over 3326024.08 frames. ], batch size: 146, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:13:28,958 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:13:30,200 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0354, 5.0736, 5.4626, 5.4283, 5.4780, 5.1186, 5.0141, 4.8435], device='cuda:0'), covar=tensor([0.0409, 0.0565, 0.0391, 0.0396, 0.0539, 0.0460, 0.1117, 0.0522], device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0493, 0.0475, 0.0439, 0.0528, 0.0505, 0.0579, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 08:14:07,154 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266346.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:14:07,445 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 08:14:16,215 INFO [train.py:904] (0/8) Epoch 27, batch 2450, loss[loss=0.1776, simple_loss=0.2589, pruned_loss=0.04818, over 15516.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2564, pruned_loss=0.03964, over 3318326.10 frames. ], batch size: 190, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:14:16,488 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8387, 5.0299, 5.2459, 4.9502, 5.0398, 5.6488, 5.1185, 4.8385], device='cuda:0'), covar=tensor([0.1370, 0.2142, 0.2520, 0.2278, 0.2559, 0.0994, 0.1791, 0.2536], device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0644, 0.0711, 0.0528, 0.0702, 0.0739, 0.0553, 0.0701], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:14:29,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8253, 4.6263, 4.8879, 5.0242, 5.2011, 4.6238, 5.2018, 5.2235], device='cuda:0'), covar=tensor([0.1909, 0.1259, 0.1576, 0.0771, 0.0528, 0.1147, 0.0698, 0.0656], device='cuda:0'), in_proj_covar=tensor([0.0688, 0.0844, 0.0979, 0.0856, 0.0649, 0.0678, 0.0713, 0.0826], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:14:31,086 INFO [optim.py:368] (0/8) 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 08:14:35,943 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 08:14:42,012 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266371.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:15:26,451 INFO [train.py:904] (0/8) Epoch 27, batch 2500, loss[loss=0.2087, simple_loss=0.2905, pruned_loss=0.06348, over 12166.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2556, pruned_loss=0.03924, over 3316914.43 frames. ], batch size: 247, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:15:31,544 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266419.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:16:24,397 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5039, 2.4564, 2.4535, 4.3116, 2.3781, 2.8934, 2.5500, 2.6371], device='cuda:0'), covar=tensor([0.1377, 0.4018, 0.3340, 0.0596, 0.4380, 0.2680, 0.3769, 0.3488], device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0475, 0.0388, 0.0339, 0.0447, 0.0544, 0.0446, 0.0556], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:16:35,711 INFO [train.py:904] (0/8) Epoch 27, batch 2550, loss[loss=0.1413, simple_loss=0.2311, pruned_loss=0.02569, over 16777.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2555, pruned_loss=0.03925, over 3321565.14 frames. ], batch size: 39, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:16:51,125 INFO [optim.py:368] (0/8) 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 08:16:51,496 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266464.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:17:08,575 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7688, 5.0058, 5.1993, 4.9375, 5.0092, 5.6231, 5.0974, 4.7923], device='cuda:0'), covar=tensor([0.1483, 0.2203, 0.2520, 0.2301, 0.2882, 0.1162, 0.1832, 0.2589], device='cuda:0'), in_proj_covar=tensor([0.0438, 0.0648, 0.0714, 0.0530, 0.0706, 0.0741, 0.0556, 0.0705], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:17:45,252 INFO [train.py:904] (0/8) Epoch 27, batch 2600, loss[loss=0.1626, simple_loss=0.2494, pruned_loss=0.03795, over 16848.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2553, pruned_loss=0.03923, over 3330914.40 frames. ], batch size: 102, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:18:16,597 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266525.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:21,082 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266528.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:18:33,598 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3435, 4.2877, 4.2949, 4.0136, 4.0747, 4.3506, 4.0567, 4.1620], device='cuda:0'), covar=tensor([0.0699, 0.0892, 0.0302, 0.0305, 0.0717, 0.0501, 0.0736, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0323, 0.0479, 0.0373, 0.0378, 0.0374, 0.0433, 0.0255, 0.0449], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:18:55,444 INFO [train.py:904] (0/8) Epoch 27, batch 2650, loss[loss=0.1741, simple_loss=0.2716, pruned_loss=0.03833, over 17045.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2557, pruned_loss=0.03899, over 3330875.46 frames. ], batch size: 53, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:19:11,569 INFO [optim.py:368] (0/8) 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 08:19:20,876 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:24,584 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:19:37,435 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266582.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:20:05,810 INFO [train.py:904] (0/8) Epoch 27, batch 2700, loss[loss=0.1526, simple_loss=0.2409, pruned_loss=0.03215, over 17044.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2562, pruned_loss=0.03909, over 3333323.19 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:20:22,276 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5796, 2.4855, 2.4149, 3.7842, 2.8813, 3.8453, 1.5732, 2.6686], device='cuda:0'), covar=tensor([0.1824, 0.0888, 0.1306, 0.0239, 0.0183, 0.0425, 0.1963, 0.1019], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0180, 0.0199, 0.0201, 0.0206, 0.0219, 0.0208, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 08:20:22,463 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 08:20:31,908 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 27, batch 2750, loss[loss=0.1786, simple_loss=0.275, pruned_loss=0.04109, over 17081.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2565, pruned_loss=0.03883, over 3340169.55 frames. ], batch size: 53, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:21:29,197 INFO [optim.py:368] (0/8) 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 08:21:51,192 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:22,673 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:22:23,527 INFO [train.py:904] (0/8) Epoch 27, batch 2800, loss[loss=0.1669, simple_loss=0.2485, pruned_loss=0.04269, over 16692.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2555, pruned_loss=0.03834, over 3338727.17 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:22:25,099 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266704.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:23:14,845 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 27, batch 2850, loss[loss=0.2252, simple_loss=0.2945, pruned_loss=0.07796, over 11762.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2542, pruned_loss=0.03779, over 3333430.84 frames. ], batch size: 247, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:23:48,202 INFO [optim.py:368] (0/8) 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 08:23:49,835 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:24:04,188 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7526, 3.8486, 2.6219, 4.5044, 3.0406, 4.3811, 2.5897, 3.2111], device='cuda:0'), covar=tensor([0.0377, 0.0454, 0.1590, 0.0288, 0.0901, 0.0596, 0.1580, 0.0828], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0184, 0.0199, 0.0177, 0.0183, 0.0225, 0.0207, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 08:24:41,246 INFO [train.py:904] (0/8) Epoch 27, batch 2900, loss[loss=0.1413, simple_loss=0.2299, pruned_loss=0.02635, over 17011.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2534, pruned_loss=0.03828, over 3318493.70 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:24:56,308 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 08:25:04,254 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=266820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:25:48,572 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7459, 6.0836, 5.8336, 5.9183, 5.4777, 5.3628, 5.5015, 6.2434], device='cuda:0'), covar=tensor([0.1366, 0.0907, 0.0975, 0.0934, 0.0963, 0.0724, 0.1326, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0727, 0.0880, 0.0717, 0.0678, 0.0559, 0.0553, 0.0744, 0.0689], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:25:49,407 INFO [train.py:904] (0/8) Epoch 27, batch 2950, loss[loss=0.1408, simple_loss=0.2255, pruned_loss=0.02805, over 16748.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2524, pruned_loss=0.03833, over 3314767.50 frames. ], batch size: 39, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:04,416 INFO [optim.py:368] (0/8) 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 08:26:04,822 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3720, 4.3732, 4.5708, 4.2474, 4.4532, 4.9806, 4.4836, 4.1584], device='cuda:0'), covar=tensor([0.1879, 0.2305, 0.2863, 0.2570, 0.2720, 0.1301, 0.1841, 0.2782], device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0644, 0.0714, 0.0529, 0.0704, 0.0739, 0.0555, 0.0705], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:26:14,737 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:26:29,689 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=266882.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:26:58,276 INFO [train.py:904] (0/8) Epoch 27, batch 3000, loss[loss=0.1431, simple_loss=0.2365, pruned_loss=0.0248, over 17237.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2532, pruned_loss=0.03881, over 3306724.25 frames. ], batch size: 45, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:26:58,277 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 08:27:05,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0751, 2.3511, 2.7715, 3.2405, 3.0694, 3.6809, 2.6320, 3.7156], device='cuda:0'), covar=tensor([0.0285, 0.0494, 0.0398, 0.0290, 0.0306, 0.0195, 0.0459, 0.0136], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0201, 0.0189, 0.0197, 0.0211, 0.0169, 0.0207, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 08:27:07,045 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 08:27:27,844 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:30,453 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=266920.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:27:33,556 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0810, 4.5993, 3.3511, 2.5859, 2.9703, 2.8234, 5.0103, 3.8406], device='cuda:0'), covar=tensor([0.2582, 0.0509, 0.1719, 0.2843, 0.2862, 0.2056, 0.0270, 0.1467], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0276, 0.0313, 0.0326, 0.0304, 0.0275, 0.0304, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:27:45,283 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=266930.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:28:16,687 INFO [train.py:904] (0/8) Epoch 27, batch 3050, loss[loss=0.1738, simple_loss=0.2666, pruned_loss=0.04048, over 17064.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2532, pruned_loss=0.03865, over 3305269.61 frames. ], batch size: 53, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:28:28,881 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7088, 2.7134, 2.2649, 2.5808, 2.9964, 2.8007, 3.2517, 3.2849], device='cuda:0'), covar=tensor([0.0206, 0.0522, 0.0696, 0.0522, 0.0376, 0.0509, 0.0325, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0249, 0.0237, 0.0237, 0.0251, 0.0249, 0.0249, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:28:30,548 INFO [optim.py:368] (0/8) 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 08:28:42,940 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7693, 3.9631, 2.7007, 4.6308, 3.1456, 4.5368, 2.7442, 3.2866], device='cuda:0'), covar=tensor([0.0350, 0.0439, 0.1573, 0.0408, 0.0833, 0.0545, 0.1481, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0184, 0.0199, 0.0177, 0.0183, 0.0225, 0.0207, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 08:28:55,917 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=266981.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:29:00,021 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9872, 5.0267, 5.4427, 5.4214, 5.4392, 5.1032, 5.0456, 4.8856], device='cuda:0'), covar=tensor([0.0351, 0.0634, 0.0405, 0.0416, 0.0397, 0.0405, 0.0903, 0.0468], device='cuda:0'), in_proj_covar=tensor([0.0442, 0.0498, 0.0480, 0.0443, 0.0530, 0.0509, 0.0587, 0.0405], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 08:29:24,860 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 27, batch 3100, loss[loss=0.1734, simple_loss=0.2483, pruned_loss=0.04931, over 16866.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2533, pruned_loss=0.03878, over 3306608.10 frames. ], batch size: 116, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:29:26,155 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7892, 1.9726, 2.5068, 2.8160, 2.6444, 3.2965, 2.1871, 3.3139], device='cuda:0'), covar=tensor([0.0325, 0.0613, 0.0372, 0.0387, 0.0419, 0.0254, 0.0610, 0.0205], device='cuda:0'), in_proj_covar=tensor([0.0202, 0.0201, 0.0190, 0.0197, 0.0211, 0.0169, 0.0207, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 08:29:51,838 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2327, 5.2086, 5.0931, 4.6005, 4.7424, 5.1413, 5.1214, 4.7500], device='cuda:0'), covar=tensor([0.0667, 0.0579, 0.0343, 0.0385, 0.1119, 0.0510, 0.0354, 0.0830], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0479, 0.0373, 0.0378, 0.0372, 0.0434, 0.0256, 0.0449], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:30:11,820 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267035.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:15,847 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2534, 5.6162, 5.3804, 5.4048, 5.0895, 5.0360, 5.0341, 5.7276], device='cuda:0'), covar=tensor([0.1307, 0.0913, 0.1018, 0.0967, 0.0854, 0.0817, 0.1364, 0.0876], device='cuda:0'), in_proj_covar=tensor([0.0732, 0.0883, 0.0722, 0.0683, 0.0561, 0.0556, 0.0749, 0.0693], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:30:33,233 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267050.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:30:36,919 INFO [train.py:904] (0/8) Epoch 27, batch 3150, loss[loss=0.1501, simple_loss=0.2375, pruned_loss=0.03134, over 16804.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2534, pruned_loss=0.03929, over 3307148.27 frames. ], batch size: 39, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:30:45,780 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.616e+02 2.137e+02 2.489e+02 2.954e+02 4.893e+02, threshold=4.979e+02, percent-clipped=1.0 2023-05-02 08:31:15,371 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:37,634 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267098.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:31:43,527 INFO [train.py:904] (0/8) Epoch 27, batch 3200, loss[loss=0.1751, simple_loss=0.2671, pruned_loss=0.04153, over 16688.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.252, pruned_loss=0.03845, over 3323097.02 frames. ], batch size: 62, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:32:08,005 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267120.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:39,058 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:32:50,806 INFO [train.py:904] (0/8) Epoch 27, batch 3250, loss[loss=0.1632, simple_loss=0.2525, pruned_loss=0.037, over 17227.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.252, pruned_loss=0.03827, over 3325158.70 frames. ], batch size: 44, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:33:01,421 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:33:07,330 INFO [optim.py:368] (0/8) 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 08:33:12,831 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267168.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:34:00,066 INFO [train.py:904] (0/8) Epoch 27, batch 3300, loss[loss=0.1357, simple_loss=0.2219, pruned_loss=0.02481, over 16971.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2537, pruned_loss=0.03909, over 3324723.86 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:34:43,915 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-05-02 08:35:07,438 INFO [train.py:904] (0/8) Epoch 27, batch 3350, loss[loss=0.1951, simple_loss=0.2836, pruned_loss=0.0533, over 12244.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2541, pruned_loss=0.03917, over 3316004.66 frames. ], batch size: 246, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:35:22,680 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.323e+02 1.912e+02 2.198e+02 2.588e+02 3.700e+02, threshold=4.395e+02, percent-clipped=0.0 2023-05-02 08:35:31,683 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8687, 5.0724, 5.2577, 5.0279, 5.1292, 5.6925, 5.1719, 4.8593], device='cuda:0'), covar=tensor([0.1324, 0.2071, 0.2445, 0.2092, 0.2492, 0.1058, 0.1730, 0.2446], device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0647, 0.0714, 0.0528, 0.0706, 0.0742, 0.0554, 0.0704], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:35:38,150 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267276.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:36:03,843 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7127, 3.4861, 3.8686, 2.1803, 3.9149, 3.9588, 3.2546, 3.1311], device='cuda:0'), covar=tensor([0.0820, 0.0267, 0.0192, 0.1175, 0.0117, 0.0228, 0.0393, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0112, 0.0102, 0.0140, 0.0087, 0.0133, 0.0131, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 08:36:13,843 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.73 vs. limit=5.0 2023-05-02 08:36:14,882 INFO [train.py:904] (0/8) Epoch 27, batch 3400, loss[loss=0.16, simple_loss=0.2534, pruned_loss=0.03331, over 16883.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2532, pruned_loss=0.0388, over 3317573.84 frames. ], batch size: 96, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:36:44,204 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 08:36:58,135 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267335.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:02,330 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6749, 4.6126, 4.5446, 3.9817, 4.6005, 1.7720, 4.3328, 4.1623], device='cuda:0'), covar=tensor([0.0159, 0.0124, 0.0198, 0.0339, 0.0120, 0.3003, 0.0167, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0177, 0.0215, 0.0188, 0.0192, 0.0221, 0.0205, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:37:14,531 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.39 vs. limit=2.0 2023-05-02 08:37:21,872 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1205, 5.5051, 5.2887, 5.3294, 5.0147, 4.9707, 4.9883, 5.6110], device='cuda:0'), covar=tensor([0.1461, 0.0948, 0.1011, 0.0894, 0.0898, 0.0963, 0.1276, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0736, 0.0890, 0.0726, 0.0685, 0.0564, 0.0558, 0.0751, 0.0696], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:37:22,638 INFO [train.py:904] (0/8) Epoch 27, batch 3450, loss[loss=0.1457, simple_loss=0.2271, pruned_loss=0.03218, over 16478.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2517, pruned_loss=0.03864, over 3313817.45 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:37:31,752 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267360.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:37:36,678 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.354e+02 2.041e+02 2.356e+02 2.660e+02 5.077e+02, threshold=4.713e+02, percent-clipped=1.0 2023-05-02 08:38:02,851 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267383.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:38:20,959 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.63 vs. limit=5.0 2023-05-02 08:38:30,913 INFO [train.py:904] (0/8) Epoch 27, batch 3500, loss[loss=0.177, simple_loss=0.2541, pruned_loss=0.04998, over 16685.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2503, pruned_loss=0.03847, over 3302925.88 frames. ], batch size: 134, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:38:39,647 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267408.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:19,227 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267438.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:40,544 INFO [train.py:904] (0/8) Epoch 27, batch 3550, loss[loss=0.1385, simple_loss=0.2239, pruned_loss=0.0266, over 16984.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2495, pruned_loss=0.03852, over 3280086.74 frames. ], batch size: 41, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:39:41,932 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267454.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:39:53,966 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.045e+02 2.366e+02 2.928e+02 5.405e+02, threshold=4.732e+02, percent-clipped=1.0 2023-05-02 08:40:36,663 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=267494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:40:48,174 INFO [train.py:904] (0/8) Epoch 27, batch 3600, loss[loss=0.1614, simple_loss=0.258, pruned_loss=0.03243, over 17255.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2494, pruned_loss=0.0381, over 3274907.86 frames. ], batch size: 52, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:00,666 INFO [train.py:904] (0/8) Epoch 27, batch 3650, loss[loss=0.1784, simple_loss=0.2601, pruned_loss=0.04834, over 11404.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2483, pruned_loss=0.038, over 3276668.94 frames. ], batch size: 247, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:42:03,442 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=267555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:42:04,776 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3719, 3.4021, 2.2727, 3.6126, 2.7659, 3.5786, 2.3360, 2.8374], device='cuda:0'), covar=tensor([0.0297, 0.0450, 0.1488, 0.0331, 0.0765, 0.0750, 0.1378, 0.0745], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0184, 0.0199, 0.0177, 0.0182, 0.0224, 0.0206, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 08:42:07,894 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6780, 3.4660, 3.8271, 2.2332, 3.9126, 3.9455, 3.3309, 3.0526], device='cuda:0'), covar=tensor([0.0802, 0.0287, 0.0230, 0.1077, 0.0116, 0.0200, 0.0380, 0.0441], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0102, 0.0139, 0.0087, 0.0132, 0.0131, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 08:42:16,689 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.290e+02 2.009e+02 2.392e+02 2.845e+02 4.698e+02, threshold=4.785e+02, percent-clipped=1.0 2023-05-02 08:42:23,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8512, 2.6836, 2.6333, 4.0234, 3.4105, 4.0693, 1.5843, 2.8723], device='cuda:0'), covar=tensor([0.1411, 0.0718, 0.1128, 0.0183, 0.0133, 0.0358, 0.1666, 0.0893], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0180, 0.0200, 0.0202, 0.0208, 0.0220, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 08:42:36,607 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267576.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:43:14,274 INFO [train.py:904] (0/8) Epoch 27, batch 3700, loss[loss=0.1722, simple_loss=0.2448, pruned_loss=0.04982, over 16819.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2469, pruned_loss=0.03939, over 3282974.97 frames. ], batch size: 116, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:43:45,317 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267624.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:44:27,015 INFO [train.py:904] (0/8) Epoch 27, batch 3750, loss[loss=0.1723, simple_loss=0.25, pruned_loss=0.04735, over 16442.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2477, pruned_loss=0.04087, over 3278185.37 frames. ], batch size: 68, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:44:41,897 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8023, 3.7851, 3.9229, 3.5982, 3.8405, 4.2765, 3.9348, 3.5747], device='cuda:0'), covar=tensor([0.2123, 0.2637, 0.2333, 0.2542, 0.2522, 0.1695, 0.1502, 0.2539], device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0645, 0.0712, 0.0527, 0.0704, 0.0739, 0.0554, 0.0703], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:44:42,770 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.122e+02 2.445e+02 2.969e+02 5.138e+02, threshold=4.890e+02, percent-clipped=1.0 2023-05-02 08:44:49,716 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2023, 3.2771, 3.4363, 2.2311, 2.9770, 2.4654, 3.7319, 3.7070], device='cuda:0'), covar=tensor([0.0210, 0.0922, 0.0640, 0.2012, 0.0882, 0.1003, 0.0453, 0.0730], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0158, 0.0149, 0.0133, 0.0148, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 08:45:20,742 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8193, 3.7832, 3.9330, 3.6834, 3.8570, 4.2937, 3.9202, 3.5890], device='cuda:0'), covar=tensor([0.2116, 0.2258, 0.2302, 0.2337, 0.2648, 0.1691, 0.1537, 0.2582], device='cuda:0'), in_proj_covar=tensor([0.0438, 0.0647, 0.0713, 0.0528, 0.0706, 0.0740, 0.0555, 0.0704], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:45:40,198 INFO [train.py:904] (0/8) Epoch 27, batch 3800, loss[loss=0.1631, simple_loss=0.2538, pruned_loss=0.03617, over 16617.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.249, pruned_loss=0.04201, over 3267092.58 frames. ], batch size: 57, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:31,891 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:46:52,913 INFO [train.py:904] (0/8) Epoch 27, batch 3850, loss[loss=0.1565, simple_loss=0.2367, pruned_loss=0.03813, over 16443.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2489, pruned_loss=0.04239, over 3268522.86 frames. ], batch size: 75, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:46:54,308 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=267754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:47:08,877 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.727e+02 2.090e+02 2.451e+02 2.947e+02 6.738e+02, threshold=4.901e+02, percent-clipped=4.0 2023-05-02 08:47:39,703 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:48:02,756 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=267802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:48:03,549 INFO [train.py:904] (0/8) Epoch 27, batch 3900, loss[loss=0.1615, simple_loss=0.2417, pruned_loss=0.04068, over 16203.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2487, pruned_loss=0.04304, over 3272498.14 frames. ], batch size: 165, lr: 2.50e-03, grad_scale: 8.0 2023-05-02 08:49:11,656 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=267850.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:49:14,858 INFO [train.py:904] (0/8) Epoch 27, batch 3950, loss[loss=0.1818, simple_loss=0.2612, pruned_loss=0.05122, over 15503.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.2486, pruned_loss=0.04356, over 3276901.50 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:49:32,330 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.702e+02 2.233e+02 2.530e+02 3.152e+02 5.510e+02, threshold=5.059e+02, percent-clipped=1.0 2023-05-02 08:49:44,346 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.19 vs. limit=5.0 2023-05-02 08:49:55,577 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5904, 4.4443, 4.6471, 4.8136, 4.9061, 4.4590, 4.7630, 4.9111], device='cuda:0'), covar=tensor([0.1913, 0.1262, 0.1503, 0.0710, 0.0734, 0.1151, 0.2492, 0.0860], device='cuda:0'), in_proj_covar=tensor([0.0699, 0.0861, 0.1000, 0.0870, 0.0662, 0.0692, 0.0722, 0.0841], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:50:28,773 INFO [train.py:904] (0/8) Epoch 27, batch 4000, loss[loss=0.1782, simple_loss=0.2528, pruned_loss=0.05176, over 16940.00 frames. ], tot_loss[loss=0.1686, simple_loss=0.2487, pruned_loss=0.0442, over 3284420.30 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:04,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4847, 3.6660, 2.7874, 2.2361, 2.3695, 2.3806, 3.8303, 3.2155], device='cuda:0'), covar=tensor([0.3134, 0.0604, 0.1849, 0.3113, 0.2773, 0.2135, 0.0515, 0.1408], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0279, 0.0315, 0.0329, 0.0309, 0.0277, 0.0307, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 08:51:24,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3320, 5.3778, 5.6618, 5.6735, 5.7536, 5.4032, 5.3597, 5.1199], device='cuda:0'), covar=tensor([0.0270, 0.0435, 0.0371, 0.0370, 0.0385, 0.0325, 0.0783, 0.0480], device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0502, 0.0482, 0.0445, 0.0532, 0.0510, 0.0585, 0.0405], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 08:51:42,550 INFO [train.py:904] (0/8) Epoch 27, batch 4050, loss[loss=0.1775, simple_loss=0.262, pruned_loss=0.04649, over 16971.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2498, pruned_loss=0.04391, over 3271097.23 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:51:58,302 INFO [optim.py:368] (0/8) 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 08:52:51,935 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-268000.pt 2023-05-02 08:52:59,883 INFO [train.py:904] (0/8) Epoch 27, batch 4100, loss[loss=0.1919, simple_loss=0.272, pruned_loss=0.05588, over 16372.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2512, pruned_loss=0.04342, over 3274269.70 frames. ], batch size: 35, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 08:53:20,080 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9145, 2.2489, 1.6918, 1.8930, 2.5863, 2.2151, 2.6288, 2.8068], device='cuda:0'), covar=tensor([0.0221, 0.0608, 0.0814, 0.0747, 0.0381, 0.0535, 0.0301, 0.0368], device='cuda:0'), in_proj_covar=tensor([0.0236, 0.0247, 0.0237, 0.0237, 0.0250, 0.0248, 0.0249, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 08:54:16,177 INFO [train.py:904] (0/8) Epoch 27, batch 4150, loss[loss=0.1925, simple_loss=0.2827, pruned_loss=0.0511, over 15523.00 frames. ], tot_loss[loss=0.1747, simple_loss=0.2581, pruned_loss=0.04563, over 3223923.56 frames. ], batch size: 191, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:54:33,172 INFO [optim.py:368] (0/8) 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 08:54:45,952 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 08:54:58,759 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5049, 4.5391, 4.8032, 4.7706, 4.8256, 4.5464, 4.4987, 4.3862], device='cuda:0'), covar=tensor([0.0322, 0.0523, 0.0388, 0.0419, 0.0420, 0.0400, 0.0892, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0438, 0.0495, 0.0476, 0.0440, 0.0525, 0.0504, 0.0578, 0.0400], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 08:55:00,519 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0107, 2.1134, 2.6535, 3.0668, 2.9530, 3.5465, 2.2089, 3.4649], device='cuda:0'), covar=tensor([0.0300, 0.0574, 0.0373, 0.0342, 0.0341, 0.0181, 0.0627, 0.0155], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0199, 0.0188, 0.0195, 0.0210, 0.0168, 0.0206, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 08:55:32,122 INFO [train.py:904] (0/8) Epoch 27, batch 4200, loss[loss=0.2046, simple_loss=0.2844, pruned_loss=0.06237, over 11762.00 frames. ], tot_loss[loss=0.1789, simple_loss=0.2646, pruned_loss=0.04656, over 3209700.18 frames. ], batch size: 248, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:56:41,013 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268150.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:56:44,663 INFO [train.py:904] (0/8) Epoch 27, batch 4250, loss[loss=0.1732, simple_loss=0.2742, pruned_loss=0.03611, over 16449.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2684, pruned_loss=0.04701, over 3173999.07 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:57:00,875 INFO [optim.py:368] (0/8) 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 08:57:50,366 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:57:57,373 INFO [train.py:904] (0/8) Epoch 27, batch 4300, loss[loss=0.2005, simple_loss=0.2844, pruned_loss=0.05829, over 11700.00 frames. ], tot_loss[loss=0.1811, simple_loss=0.2694, pruned_loss=0.04642, over 3154494.50 frames. ], batch size: 248, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:58:10,894 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268212.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 08:59:12,219 INFO [train.py:904] (0/8) Epoch 27, batch 4350, loss[loss=0.2098, simple_loss=0.3023, pruned_loss=0.05868, over 16465.00 frames. ], tot_loss[loss=0.1837, simple_loss=0.2729, pruned_loss=0.04727, over 3160357.35 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 16.0 2023-05-02 08:59:27,931 INFO [optim.py:368] (0/8) 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 08:59:42,656 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268273.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 08:59:45,651 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:00:26,996 INFO [train.py:904] (0/8) Epoch 27, batch 4400, loss[loss=0.1802, simple_loss=0.2784, pruned_loss=0.04098, over 16784.00 frames. ], tot_loss[loss=0.1857, simple_loss=0.2748, pruned_loss=0.04824, over 3162886.48 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:14,975 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268336.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:01:37,981 INFO [train.py:904] (0/8) Epoch 27, batch 4450, loss[loss=0.2099, simple_loss=0.2988, pruned_loss=0.06053, over 16736.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2781, pruned_loss=0.0496, over 3170521.74 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:01:55,123 INFO [optim.py:368] (0/8) 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 09:02:26,304 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 09:02:42,293 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 09:02:50,812 INFO [train.py:904] (0/8) Epoch 27, batch 4500, loss[loss=0.1799, simple_loss=0.2738, pruned_loss=0.04298, over 16530.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.2788, pruned_loss=0.05034, over 3165429.00 frames. ], batch size: 75, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:03:42,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0389, 3.2944, 3.4411, 1.9758, 2.8887, 2.0319, 3.4015, 3.5251], device='cuda:0'), covar=tensor([0.0251, 0.0829, 0.0559, 0.2383, 0.0935, 0.1209, 0.0668, 0.0864], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 09:04:03,673 INFO [train.py:904] (0/8) Epoch 27, batch 4550, loss[loss=0.1892, simple_loss=0.2741, pruned_loss=0.05211, over 17131.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2798, pruned_loss=0.05129, over 3186233.69 frames. ], batch size: 47, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:04:20,728 INFO [optim.py:368] (0/8) 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 09:04:41,186 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3450, 4.1894, 4.0921, 2.6416, 3.7401, 4.1343, 3.6141, 2.6326], device='cuda:0'), covar=tensor([0.0604, 0.0036, 0.0048, 0.0449, 0.0098, 0.0086, 0.0103, 0.0426], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0089, 0.0091, 0.0136, 0.0102, 0.0115, 0.0099, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 09:05:04,950 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2153, 3.4642, 3.6227, 2.0695, 3.0057, 2.2510, 3.5881, 3.7145], device='cuda:0'), covar=tensor([0.0209, 0.0761, 0.0581, 0.2171, 0.0878, 0.1079, 0.0567, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0171, 0.0169, 0.0155, 0.0147, 0.0131, 0.0145, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 09:05:15,503 INFO [train.py:904] (0/8) Epoch 27, batch 4600, loss[loss=0.1852, simple_loss=0.274, pruned_loss=0.04822, over 16515.00 frames. ], tot_loss[loss=0.1928, simple_loss=0.2812, pruned_loss=0.05221, over 3186821.75 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:05:20,790 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5758, 3.7229, 3.8443, 2.1418, 3.0742, 2.4295, 3.9032, 3.8822], device='cuda:0'), covar=tensor([0.0219, 0.0749, 0.0548, 0.2148, 0.0902, 0.1057, 0.0548, 0.0991], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0171, 0.0169, 0.0155, 0.0147, 0.0132, 0.0145, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 09:05:58,283 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 09:06:23,491 INFO [train.py:904] (0/8) Epoch 27, batch 4650, loss[loss=0.195, simple_loss=0.2731, pruned_loss=0.05845, over 16588.00 frames. ], tot_loss[loss=0.1925, simple_loss=0.2803, pruned_loss=0.05235, over 3212030.93 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:06:40,834 INFO [optim.py:368] (0/8) 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 09:06:45,264 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268568.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:06:51,958 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1368, 2.4345, 2.4520, 3.7969, 2.2823, 2.7183, 2.4560, 2.4972], device='cuda:0'), covar=tensor([0.1445, 0.3125, 0.2852, 0.0636, 0.4175, 0.2322, 0.3222, 0.3376], device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0472, 0.0384, 0.0336, 0.0446, 0.0540, 0.0443, 0.0554], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:07:23,764 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9163, 4.0068, 2.6776, 4.9495, 3.2694, 4.7258, 2.7949, 3.2815], device='cuda:0'), covar=tensor([0.0318, 0.0375, 0.1595, 0.0147, 0.0793, 0.0519, 0.1420, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0181, 0.0197, 0.0172, 0.0179, 0.0221, 0.0203, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:07:32,861 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:07:34,897 INFO [train.py:904] (0/8) Epoch 27, batch 4700, loss[loss=0.1946, simple_loss=0.283, pruned_loss=0.05314, over 16828.00 frames. ], tot_loss[loss=0.1899, simple_loss=0.2776, pruned_loss=0.05109, over 3199674.81 frames. ], batch size: 39, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:07:59,607 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0129, 5.3056, 5.1224, 5.1440, 4.8676, 4.7071, 4.7230, 5.4033], device='cuda:0'), covar=tensor([0.1194, 0.0790, 0.0840, 0.0771, 0.0749, 0.1038, 0.1136, 0.0869], device='cuda:0'), in_proj_covar=tensor([0.0709, 0.0860, 0.0699, 0.0660, 0.0546, 0.0541, 0.0722, 0.0675], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:08:15,189 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268631.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:08:29,508 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9076, 3.1403, 3.3659, 1.9832, 2.8729, 2.1594, 3.3089, 3.3208], device='cuda:0'), covar=tensor([0.0271, 0.0865, 0.0631, 0.2159, 0.0903, 0.1061, 0.0679, 0.1108], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0169, 0.0155, 0.0147, 0.0132, 0.0145, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:08:43,011 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2273, 3.4098, 3.5766, 2.0274, 3.0299, 2.3181, 3.5877, 3.6496], device='cuda:0'), covar=tensor([0.0237, 0.0887, 0.0632, 0.2172, 0.0896, 0.1031, 0.0608, 0.1076], device='cuda:0'), in_proj_covar=tensor([0.0159, 0.0170, 0.0168, 0.0154, 0.0147, 0.0131, 0.0145, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:08:45,865 INFO [train.py:904] (0/8) Epoch 27, batch 4750, loss[loss=0.1674, simple_loss=0.2573, pruned_loss=0.03872, over 16702.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2734, pruned_loss=0.04849, over 3204101.36 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:08:58,283 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2914, 4.1345, 4.3308, 4.4826, 4.6143, 4.2209, 4.5540, 4.6570], device='cuda:0'), covar=tensor([0.1690, 0.1213, 0.1541, 0.0698, 0.0569, 0.1146, 0.0768, 0.0584], device='cuda:0'), in_proj_covar=tensor([0.0669, 0.0822, 0.0957, 0.0836, 0.0633, 0.0664, 0.0693, 0.0806], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:08:59,625 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268662.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:09:04,567 INFO [optim.py:368] (0/8) 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 09:09:59,308 INFO [train.py:904] (0/8) Epoch 27, batch 4800, loss[loss=0.1838, simple_loss=0.2836, pruned_loss=0.04206, over 16318.00 frames. ], tot_loss[loss=0.1816, simple_loss=0.27, pruned_loss=0.04664, over 3195896.09 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:10:28,562 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4418, 3.3672, 3.4448, 3.5315, 3.5896, 3.3283, 3.5586, 3.6395], device='cuda:0'), covar=tensor([0.1164, 0.0966, 0.1077, 0.0623, 0.0590, 0.1943, 0.0957, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0667, 0.0819, 0.0953, 0.0834, 0.0631, 0.0662, 0.0691, 0.0803], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:11:14,179 INFO [train.py:904] (0/8) Epoch 27, batch 4850, loss[loss=0.1922, simple_loss=0.2853, pruned_loss=0.04954, over 16704.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2709, pruned_loss=0.04592, over 3181930.38 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:11:26,633 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.68 vs. limit=5.0 2023-05-02 09:11:31,500 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.342e+02 1.812e+02 2.075e+02 2.444e+02 6.308e+02, threshold=4.151e+02, percent-clipped=1.0 2023-05-02 09:12:27,710 INFO [train.py:904] (0/8) Epoch 27, batch 4900, loss[loss=0.1937, simple_loss=0.2833, pruned_loss=0.05205, over 16882.00 frames. ], tot_loss[loss=0.1803, simple_loss=0.2705, pruned_loss=0.04499, over 3168953.92 frames. ], batch size: 109, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:20,802 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4015, 1.7395, 2.1301, 2.3926, 2.4641, 2.6939, 1.9055, 2.5715], device='cuda:0'), covar=tensor([0.0250, 0.0572, 0.0339, 0.0403, 0.0371, 0.0216, 0.0616, 0.0169], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0199, 0.0187, 0.0194, 0.0208, 0.0166, 0.0205, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:13:37,516 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 09:13:37,967 INFO [train.py:904] (0/8) Epoch 27, batch 4950, loss[loss=0.1788, simple_loss=0.2735, pruned_loss=0.04208, over 16636.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2701, pruned_loss=0.04431, over 3175865.01 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:13:50,518 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5695, 4.5268, 4.4003, 2.7553, 3.8646, 4.4304, 3.8000, 2.4050], device='cuda:0'), covar=tensor([0.0589, 0.0032, 0.0040, 0.0439, 0.0091, 0.0071, 0.0100, 0.0491], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0088, 0.0090, 0.0135, 0.0101, 0.0114, 0.0097, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 09:13:54,420 INFO [optim.py:368] (0/8) 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 09:13:58,125 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8687, 2.2015, 2.4391, 3.1345, 2.1992, 2.3782, 2.3147, 2.3166], device='cuda:0'), covar=tensor([0.1556, 0.3619, 0.2654, 0.0794, 0.4173, 0.2506, 0.3742, 0.3229], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0472, 0.0382, 0.0335, 0.0445, 0.0539, 0.0441, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:13:59,069 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268868.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:14:36,321 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:14:50,666 INFO [train.py:904] (0/8) Epoch 27, batch 5000, loss[loss=0.1799, simple_loss=0.2681, pruned_loss=0.04592, over 17128.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2714, pruned_loss=0.0444, over 3177577.68 frames. ], batch size: 49, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:15:09,313 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268916.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:15:29,460 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=268931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:00,161 INFO [train.py:904] (0/8) Epoch 27, batch 5050, loss[loss=0.1738, simple_loss=0.2691, pruned_loss=0.03923, over 16331.00 frames. ], tot_loss[loss=0.1796, simple_loss=0.2714, pruned_loss=0.04386, over 3201935.20 frames. ], batch size: 165, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:16:03,728 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=268955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:05,790 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=268957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:17,238 INFO [optim.py:368] (0/8) 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 09:16:34,319 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0151, 2.3687, 1.9527, 2.2348, 2.7222, 2.3600, 2.5417, 2.9224], device='cuda:0'), covar=tensor([0.0187, 0.0508, 0.0685, 0.0520, 0.0306, 0.0483, 0.0245, 0.0295], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0244, 0.0233, 0.0233, 0.0246, 0.0244, 0.0242, 0.0243], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:16:37,072 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=268979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:16:54,224 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=268991.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:17:11,193 INFO [train.py:904] (0/8) Epoch 27, batch 5100, loss[loss=0.1469, simple_loss=0.2308, pruned_loss=0.03153, over 16575.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2699, pruned_loss=0.04312, over 3206220.43 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:17:22,935 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.55 vs. limit=2.0 2023-05-02 09:18:00,480 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.20 vs. limit=5.0 2023-05-02 09:18:22,299 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:18:23,604 INFO [train.py:904] (0/8) Epoch 27, batch 5150, loss[loss=0.1669, simple_loss=0.2714, pruned_loss=0.03124, over 16797.00 frames. ], tot_loss[loss=0.1773, simple_loss=0.2697, pruned_loss=0.04245, over 3209561.70 frames. ], batch size: 102, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:18:33,570 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 09:18:41,484 INFO [optim.py:368] (0/8) 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 09:19:15,452 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5970, 3.6651, 2.7773, 2.2755, 2.4306, 2.4170, 3.8271, 3.2408], device='cuda:0'), covar=tensor([0.2984, 0.0661, 0.1960, 0.3157, 0.2548, 0.2092, 0.0552, 0.1400], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0275, 0.0312, 0.0325, 0.0305, 0.0274, 0.0304, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 09:19:36,085 INFO [train.py:904] (0/8) Epoch 27, batch 5200, loss[loss=0.1685, simple_loss=0.2576, pruned_loss=0.03965, over 16664.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2688, pruned_loss=0.04213, over 3196993.04 frames. ], batch size: 134, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:20:47,202 INFO [train.py:904] (0/8) Epoch 27, batch 5250, loss[loss=0.1709, simple_loss=0.2662, pruned_loss=0.03781, over 16461.00 frames. ], tot_loss[loss=0.1752, simple_loss=0.2664, pruned_loss=0.04202, over 3196283.95 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:21:04,388 INFO [optim.py:368] (0/8) 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 09:21:55,562 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4475, 2.5552, 2.5489, 4.1227, 2.3969, 2.8893, 2.5904, 2.7384], device='cuda:0'), covar=tensor([0.1377, 0.3512, 0.2914, 0.0554, 0.3928, 0.2353, 0.3502, 0.3237], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0472, 0.0383, 0.0336, 0.0445, 0.0540, 0.0443, 0.0553], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:21:57,046 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7357, 3.9895, 3.0003, 2.4295, 2.6816, 2.6095, 4.3503, 3.4419], device='cuda:0'), covar=tensor([0.3067, 0.0684, 0.2066, 0.2965, 0.2684, 0.2144, 0.0470, 0.1472], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0276, 0.0313, 0.0326, 0.0305, 0.0274, 0.0304, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 09:22:00,570 INFO [train.py:904] (0/8) Epoch 27, batch 5300, loss[loss=0.1514, simple_loss=0.243, pruned_loss=0.02994, over 16470.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2631, pruned_loss=0.04093, over 3192449.92 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:22:55,407 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269241.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:08,331 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269250.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:12,970 INFO [train.py:904] (0/8) Epoch 27, batch 5350, loss[loss=0.1698, simple_loss=0.2639, pruned_loss=0.03787, over 16486.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2617, pruned_loss=0.04067, over 3178621.60 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:23:18,832 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269257.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:23:29,782 INFO [optim.py:368] (0/8) 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 09:23:31,885 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0995, 2.3254, 1.7557, 2.0762, 2.6763, 2.3923, 2.6374, 2.9621], device='cuda:0'), covar=tensor([0.0214, 0.0589, 0.0828, 0.0622, 0.0365, 0.0532, 0.0299, 0.0345], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0242, 0.0231, 0.0232, 0.0244, 0.0242, 0.0240, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:23:47,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6131, 3.3983, 3.9301, 1.9975, 4.0657, 4.0600, 2.9495, 2.9895], device='cuda:0'), covar=tensor([0.0850, 0.0326, 0.0181, 0.1253, 0.0077, 0.0151, 0.0501, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0111, 0.0101, 0.0140, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:24:24,511 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269302.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:24:25,135 INFO [train.py:904] (0/8) Epoch 27, batch 5400, loss[loss=0.176, simple_loss=0.2696, pruned_loss=0.04119, over 16764.00 frames. ], tot_loss[loss=0.1734, simple_loss=0.2642, pruned_loss=0.04125, over 3180476.33 frames. ], batch size: 76, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:24:28,352 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:31,288 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269347.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:40,787 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:25:42,383 INFO [train.py:904] (0/8) Epoch 27, batch 5450, loss[loss=0.2044, simple_loss=0.292, pruned_loss=0.05844, over 16487.00 frames. ], tot_loss[loss=0.1758, simple_loss=0.2669, pruned_loss=0.04235, over 3179958.55 frames. ], batch size: 146, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:26:01,153 INFO [optim.py:368] (0/8) 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 09:26:19,963 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 09:26:32,035 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6806, 2.6959, 2.5024, 3.9793, 2.8629, 3.9497, 1.5210, 2.7904], device='cuda:0'), covar=tensor([0.1387, 0.0808, 0.1297, 0.0207, 0.0296, 0.0398, 0.1756, 0.0919], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0180, 0.0200, 0.0200, 0.0207, 0.0218, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:26:46,528 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 09:27:00,970 INFO [train.py:904] (0/8) Epoch 27, batch 5500, loss[loss=0.1916, simple_loss=0.2896, pruned_loss=0.04682, over 16736.00 frames. ], tot_loss[loss=0.1825, simple_loss=0.2733, pruned_loss=0.04587, over 3166815.01 frames. ], batch size: 83, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:27:16,662 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269413.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:27:19,674 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0549, 2.8434, 3.1130, 1.8093, 3.2171, 3.2362, 2.6642, 2.5392], device='cuda:0'), covar=tensor([0.0894, 0.0304, 0.0197, 0.1174, 0.0109, 0.0213, 0.0486, 0.0514], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0111, 0.0102, 0.0141, 0.0087, 0.0132, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:28:00,111 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269441.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:28:19,428 INFO [train.py:904] (0/8) Epoch 27, batch 5550, loss[loss=0.2159, simple_loss=0.3011, pruned_loss=0.06541, over 16742.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.2801, pruned_loss=0.0507, over 3137490.74 frames. ], batch size: 124, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:28:38,502 INFO [optim.py:368] (0/8) 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 09:29:37,431 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269502.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:29:38,103 INFO [train.py:904] (0/8) Epoch 27, batch 5600, loss[loss=0.2459, simple_loss=0.3241, pruned_loss=0.08383, over 15211.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2849, pruned_loss=0.0544, over 3125547.89 frames. ], batch size: 190, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:30:56,852 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:31:01,866 INFO [train.py:904] (0/8) Epoch 27, batch 5650, loss[loss=0.1781, simple_loss=0.266, pruned_loss=0.04509, over 16413.00 frames. ], tot_loss[loss=0.203, simple_loss=0.2895, pruned_loss=0.05824, over 3086830.46 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:31:12,488 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.75 vs. limit=5.0 2023-05-02 09:31:20,025 INFO [optim.py:368] (0/8) 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 09:32:00,475 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7399, 2.5179, 2.3354, 3.1910, 2.2821, 3.5859, 1.6503, 2.7049], device='cuda:0'), covar=tensor([0.1402, 0.0766, 0.1308, 0.0218, 0.0214, 0.0451, 0.1717, 0.0875], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0179, 0.0199, 0.0200, 0.0207, 0.0218, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:32:11,867 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:13,056 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:32:20,982 INFO [train.py:904] (0/8) Epoch 27, batch 5700, loss[loss=0.2126, simple_loss=0.314, pruned_loss=0.05554, over 16708.00 frames. ], tot_loss[loss=0.2067, simple_loss=0.2922, pruned_loss=0.06063, over 3062628.59 frames. ], batch size: 89, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:32:54,712 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3516, 3.8044, 3.6936, 2.3641, 3.4690, 3.7814, 3.5131, 1.7532], device='cuda:0'), covar=tensor([0.0681, 0.0072, 0.0101, 0.0593, 0.0133, 0.0169, 0.0128, 0.0751], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0089, 0.0090, 0.0136, 0.0101, 0.0115, 0.0098, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 09:33:31,924 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269647.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:33:41,291 INFO [train.py:904] (0/8) Epoch 27, batch 5750, loss[loss=0.1947, simple_loss=0.2923, pruned_loss=0.04855, over 16879.00 frames. ], tot_loss[loss=0.2098, simple_loss=0.295, pruned_loss=0.06226, over 3038938.65 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:33:59,185 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.119e+02 2.911e+02 3.444e+02 4.109e+02 8.393e+02, threshold=6.889e+02, percent-clipped=0.0 2023-05-02 09:34:49,727 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:35:02,100 INFO [train.py:904] (0/8) Epoch 27, batch 5800, loss[loss=0.1633, simple_loss=0.2592, pruned_loss=0.03374, over 16834.00 frames. ], tot_loss[loss=0.2077, simple_loss=0.2939, pruned_loss=0.06071, over 3045273.62 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:35:07,951 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3183, 2.4092, 2.4554, 4.0330, 2.2940, 2.7690, 2.4406, 2.5796], device='cuda:0'), covar=tensor([0.1355, 0.3339, 0.2867, 0.0553, 0.3921, 0.2303, 0.3611, 0.3061], device='cuda:0'), in_proj_covar=tensor([0.0415, 0.0467, 0.0379, 0.0332, 0.0440, 0.0535, 0.0437, 0.0547], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:35:11,063 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269708.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:36:21,276 INFO [train.py:904] (0/8) Epoch 27, batch 5850, loss[loss=0.1786, simple_loss=0.271, pruned_loss=0.04314, over 16452.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2915, pruned_loss=0.05913, over 3047124.47 frames. ], batch size: 68, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:36:40,944 INFO [optim.py:368] (0/8) 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 09:36:41,494 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0914, 4.9006, 5.1731, 5.3416, 5.5216, 4.8700, 5.4846, 5.5237], device='cuda:0'), covar=tensor([0.2266, 0.1330, 0.1668, 0.0733, 0.0589, 0.0934, 0.0612, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0664, 0.0814, 0.0944, 0.0824, 0.0627, 0.0658, 0.0685, 0.0800], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:36:52,942 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:37:34,888 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=269797.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 09:37:45,800 INFO [train.py:904] (0/8) Epoch 27, batch 5900, loss[loss=0.1844, simple_loss=0.2815, pruned_loss=0.04364, over 16610.00 frames. ], tot_loss[loss=0.2037, simple_loss=0.2903, pruned_loss=0.05853, over 3058495.40 frames. ], batch size: 62, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:37:50,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6959, 3.5538, 4.0236, 2.1163, 4.1775, 4.1815, 3.2740, 3.0681], device='cuda:0'), covar=tensor([0.0807, 0.0278, 0.0203, 0.1244, 0.0080, 0.0181, 0.0370, 0.0521], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0111, 0.0101, 0.0139, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:38:37,610 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269834.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:38:40,771 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=269836.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:38:51,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2598, 3.5530, 3.6812, 2.1515, 3.1390, 2.4567, 3.6893, 3.8712], device='cuda:0'), covar=tensor([0.0269, 0.0832, 0.0611, 0.2150, 0.0839, 0.1000, 0.0585, 0.0859], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0157, 0.0148, 0.0133, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 09:39:06,835 INFO [train.py:904] (0/8) Epoch 27, batch 5950, loss[loss=0.1983, simple_loss=0.2882, pruned_loss=0.05426, over 16704.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2907, pruned_loss=0.05733, over 3076441.95 frames. ], batch size: 89, lr: 2.49e-03, grad_scale: 4.0 2023-05-02 09:39:27,605 INFO [optim.py:368] (0/8) 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 09:39:29,448 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 09:40:16,964 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:17,091 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=269897.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:40:24,914 INFO [train.py:904] (0/8) Epoch 27, batch 6000, loss[loss=0.2167, simple_loss=0.2934, pruned_loss=0.07003, over 11677.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2904, pruned_loss=0.05745, over 3074946.35 frames. ], batch size: 247, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:40:24,915 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 09:40:35,103 INFO [train.py:938] (0/8) Epoch 27, validation: loss=0.148, simple_loss=0.2603, pruned_loss=0.01783, over 944034.00 frames. 2023-05-02 09:40:35,104 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 09:41:37,535 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=269945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:41:52,132 INFO [train.py:904] (0/8) Epoch 27, batch 6050, loss[loss=0.2081, simple_loss=0.314, pruned_loss=0.05114, over 16594.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2892, pruned_loss=0.05681, over 3091736.41 frames. ], batch size: 57, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:42:12,243 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.929e+02 2.707e+02 3.068e+02 3.739e+02 6.756e+02, threshold=6.136e+02, percent-clipped=2.0 2023-05-02 09:43:06,429 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-270000.pt 2023-05-02 09:43:13,685 INFO [train.py:904] (0/8) Epoch 27, batch 6100, loss[loss=0.1721, simple_loss=0.2728, pruned_loss=0.03568, over 16861.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2883, pruned_loss=0.05516, over 3118081.10 frames. ], batch size: 96, lr: 2.49e-03, grad_scale: 8.0 2023-05-02 09:43:22,359 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270008.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:17,804 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 09:44:33,193 INFO [train.py:904] (0/8) Epoch 27, batch 6150, loss[loss=0.2047, simple_loss=0.29, pruned_loss=0.05975, over 16416.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2869, pruned_loss=0.05541, over 3095237.34 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:44:37,807 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270056.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:44:39,373 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9015, 2.3636, 2.0353, 2.1622, 2.7172, 2.3694, 2.5742, 2.8847], device='cuda:0'), covar=tensor([0.0229, 0.0497, 0.0617, 0.0564, 0.0331, 0.0474, 0.0281, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0240, 0.0231, 0.0232, 0.0242, 0.0241, 0.0239, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:44:53,897 INFO [optim.py:368] (0/8) 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 09:45:20,410 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.00 vs. limit=2.0 2023-05-02 09:45:42,308 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270097.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 09:45:51,133 INFO [train.py:904] (0/8) Epoch 27, batch 6200, loss[loss=0.1795, simple_loss=0.2682, pruned_loss=0.04542, over 16254.00 frames. ], tot_loss[loss=0.1968, simple_loss=0.2846, pruned_loss=0.05447, over 3124563.22 frames. ], batch size: 35, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:46:33,644 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270129.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:46:57,438 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:47:09,700 INFO [train.py:904] (0/8) Epoch 27, batch 6250, loss[loss=0.1844, simple_loss=0.2783, pruned_loss=0.04526, over 17073.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2844, pruned_loss=0.05428, over 3136970.58 frames. ], batch size: 55, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:47:29,203 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.849e+02 2.710e+02 3.440e+02 4.218e+02 8.566e+02, threshold=6.880e+02, percent-clipped=4.0 2023-05-02 09:48:08,662 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:48:25,160 INFO [train.py:904] (0/8) Epoch 27, batch 6300, loss[loss=0.1993, simple_loss=0.2857, pruned_loss=0.05641, over 16857.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.2841, pruned_loss=0.0538, over 3126604.76 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:49:29,736 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8321, 2.6050, 2.3699, 3.1237, 2.1354, 3.6108, 1.5475, 2.7460], device='cuda:0'), covar=tensor([0.1368, 0.0720, 0.1268, 0.0227, 0.0169, 0.0425, 0.1830, 0.0852], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0179, 0.0200, 0.0201, 0.0208, 0.0219, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:49:45,982 INFO [train.py:904] (0/8) Epoch 27, batch 6350, loss[loss=0.1772, simple_loss=0.2718, pruned_loss=0.04133, over 16385.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2846, pruned_loss=0.05464, over 3116019.08 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:49:48,930 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0301, 3.0766, 1.9806, 3.2480, 2.3854, 3.3119, 2.1328, 2.5777], device='cuda:0'), covar=tensor([0.0322, 0.0415, 0.1621, 0.0308, 0.0779, 0.0665, 0.1462, 0.0781], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0180, 0.0195, 0.0171, 0.0179, 0.0219, 0.0202, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 09:50:05,611 INFO [optim.py:368] (0/8) 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 09:50:28,131 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270280.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:51:02,490 INFO [train.py:904] (0/8) Epoch 27, batch 6400, loss[loss=0.2074, simple_loss=0.2904, pruned_loss=0.06217, over 15449.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.2851, pruned_loss=0.05623, over 3088439.37 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:01,045 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270341.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 09:52:19,099 INFO [train.py:904] (0/8) Epoch 27, batch 6450, loss[loss=0.2007, simple_loss=0.2756, pruned_loss=0.0629, over 11795.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2862, pruned_loss=0.05674, over 3050190.13 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:52:30,529 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1730, 3.3441, 3.5841, 2.0611, 3.0226, 2.3560, 3.6478, 3.7118], device='cuda:0'), covar=tensor([0.0255, 0.0877, 0.0580, 0.2190, 0.0879, 0.0998, 0.0568, 0.0954], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0148, 0.0132, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 09:52:39,052 INFO [optim.py:368] (0/8) 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 09:53:26,955 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:53:37,255 INFO [train.py:904] (0/8) Epoch 27, batch 6500, loss[loss=0.1898, simple_loss=0.2797, pruned_loss=0.04991, over 16400.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2845, pruned_loss=0.056, over 3059195.06 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:53:50,373 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8209, 3.9033, 4.1238, 4.0959, 4.1235, 3.9160, 3.9211, 3.9485], device='cuda:0'), covar=tensor([0.0366, 0.0713, 0.0430, 0.0484, 0.0485, 0.0516, 0.0880, 0.0524], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0479, 0.0465, 0.0426, 0.0513, 0.0490, 0.0564, 0.0392], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 09:54:17,008 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270429.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 09:54:20,042 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5607, 4.8577, 4.6552, 4.6265, 4.4022, 4.3269, 4.3072, 4.9349], device='cuda:0'), covar=tensor([0.1360, 0.0849, 0.1045, 0.0956, 0.0880, 0.1453, 0.1270, 0.0881], device='cuda:0'), in_proj_covar=tensor([0.0706, 0.0854, 0.0697, 0.0658, 0.0539, 0.0538, 0.0717, 0.0666], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:54:25,730 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 09:54:33,582 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270439.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:54:58,115 INFO [train.py:904] (0/8) Epoch 27, batch 6550, loss[loss=0.1818, simple_loss=0.2843, pruned_loss=0.03966, over 16682.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2871, pruned_loss=0.05653, over 3072420.14 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:55:04,111 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:16,564 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1154, 5.4226, 5.1914, 5.2133, 4.9797, 4.8506, 4.8142, 5.5251], device='cuda:0'), covar=tensor([0.1268, 0.0826, 0.1025, 0.0849, 0.0814, 0.0958, 0.1306, 0.0895], device='cuda:0'), in_proj_covar=tensor([0.0704, 0.0852, 0.0694, 0.0656, 0.0538, 0.0537, 0.0715, 0.0665], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:55:17,873 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:55:55,934 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2103, 5.1718, 4.9988, 4.2463, 5.1264, 2.0191, 4.8049, 4.5630], device='cuda:0'), covar=tensor([0.0105, 0.0100, 0.0216, 0.0399, 0.0091, 0.2859, 0.0145, 0.0301], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0173, 0.0210, 0.0185, 0.0186, 0.0215, 0.0199, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:55:57,712 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:56:09,999 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270500.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:56:11,606 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4295, 5.7170, 5.4284, 5.4922, 5.1590, 5.1232, 5.0943, 5.8331], device='cuda:0'), covar=tensor([0.1318, 0.0866, 0.1118, 0.0866, 0.0897, 0.0746, 0.1276, 0.0881], device='cuda:0'), in_proj_covar=tensor([0.0705, 0.0853, 0.0697, 0.0657, 0.0539, 0.0538, 0.0717, 0.0667], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:56:14,141 INFO [train.py:904] (0/8) Epoch 27, batch 6600, loss[loss=0.2009, simple_loss=0.2849, pruned_loss=0.05848, over 15285.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2885, pruned_loss=0.05662, over 3059857.69 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:09,703 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:57:30,479 INFO [train.py:904] (0/8) Epoch 27, batch 6650, loss[loss=0.1875, simple_loss=0.2761, pruned_loss=0.04947, over 16629.00 frames. ], tot_loss[loss=0.2029, simple_loss=0.2896, pruned_loss=0.05807, over 3050464.86 frames. ], batch size: 76, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:57:50,324 INFO [optim.py:368] (0/8) 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 09:58:46,100 INFO [train.py:904] (0/8) Epoch 27, batch 6700, loss[loss=0.1869, simple_loss=0.278, pruned_loss=0.04794, over 16889.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.288, pruned_loss=0.0577, over 3072074.05 frames. ], batch size: 96, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 09:58:46,580 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5258, 3.5843, 3.3549, 2.9794, 3.2240, 3.5007, 3.3274, 3.3135], device='cuda:0'), covar=tensor([0.0559, 0.0606, 0.0301, 0.0313, 0.0492, 0.0492, 0.1311, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0461, 0.0358, 0.0361, 0.0357, 0.0415, 0.0246, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 09:59:01,631 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 09:59:35,303 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270636.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:00:01,089 INFO [train.py:904] (0/8) Epoch 27, batch 6750, loss[loss=0.1783, simple_loss=0.2667, pruned_loss=0.04498, over 16576.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.287, pruned_loss=0.0578, over 3061736.70 frames. ], batch size: 75, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:00:02,184 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 10:00:09,211 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5701, 4.3957, 4.5862, 4.7458, 4.9293, 4.4672, 4.8697, 4.9239], device='cuda:0'), covar=tensor([0.1905, 0.1307, 0.1622, 0.0724, 0.0511, 0.1157, 0.0660, 0.0689], device='cuda:0'), in_proj_covar=tensor([0.0664, 0.0812, 0.0944, 0.0824, 0.0627, 0.0658, 0.0687, 0.0799], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:00:20,167 INFO [optim.py:368] (0/8) 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 10:00:22,380 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7571, 5.0092, 5.1979, 4.9067, 5.0208, 5.5503, 4.9969, 4.7639], device='cuda:0'), covar=tensor([0.1107, 0.1680, 0.2158, 0.1907, 0.2198, 0.0961, 0.1666, 0.2391], device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0627, 0.0687, 0.0510, 0.0679, 0.0716, 0.0538, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 10:00:32,929 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:01:04,269 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4361, 3.4400, 3.4064, 2.4264, 3.3071, 2.0894, 3.1048, 2.5619], device='cuda:0'), covar=tensor([0.0242, 0.0220, 0.0264, 0.0405, 0.0153, 0.3053, 0.0212, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0173, 0.0211, 0.0185, 0.0186, 0.0216, 0.0199, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:01:15,220 INFO [train.py:904] (0/8) Epoch 27, batch 6800, loss[loss=0.2017, simple_loss=0.2914, pruned_loss=0.05604, over 15277.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2868, pruned_loss=0.05733, over 3080641.30 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:01:53,385 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270726.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:33,316 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:02:34,172 INFO [train.py:904] (0/8) Epoch 27, batch 6850, loss[loss=0.2046, simple_loss=0.3091, pruned_loss=0.05007, over 16747.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2883, pruned_loss=0.05792, over 3069034.90 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:02:53,222 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.651e+02 2.704e+02 3.220e+02 3.695e+02 5.610e+02, threshold=6.440e+02, percent-clipped=0.0 2023-05-02 10:03:25,309 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:31,587 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6758, 3.8040, 2.8921, 2.3056, 2.5069, 2.4914, 4.0422, 3.2870], device='cuda:0'), covar=tensor([0.2883, 0.0651, 0.1900, 0.3007, 0.2528, 0.2109, 0.0505, 0.1387], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0275, 0.0312, 0.0327, 0.0305, 0.0275, 0.0305, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 10:03:37,600 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270795.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:03:49,492 INFO [train.py:904] (0/8) Epoch 27, batch 6900, loss[loss=0.1756, simple_loss=0.2752, pruned_loss=0.03798, over 16744.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2903, pruned_loss=0.05667, over 3104496.09 frames. ], batch size: 83, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:04:42,166 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3717, 4.4355, 4.7360, 4.6947, 4.7208, 4.4615, 4.4215, 4.3885], device='cuda:0'), covar=tensor([0.0331, 0.0576, 0.0383, 0.0418, 0.0450, 0.0375, 0.0931, 0.0480], device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0476, 0.0463, 0.0424, 0.0509, 0.0488, 0.0561, 0.0390], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 10:05:03,193 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7735, 3.7876, 3.8950, 3.6576, 3.8463, 4.2161, 3.8851, 3.6148], device='cuda:0'), covar=tensor([0.2054, 0.2232, 0.2447, 0.2287, 0.2501, 0.1711, 0.1598, 0.2441], device='cuda:0'), in_proj_covar=tensor([0.0424, 0.0629, 0.0692, 0.0513, 0.0682, 0.0720, 0.0540, 0.0687], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 10:05:07,700 INFO [train.py:904] (0/8) Epoch 27, batch 6950, loss[loss=0.1952, simple_loss=0.2835, pruned_loss=0.0534, over 17202.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2913, pruned_loss=0.05775, over 3101224.73 frames. ], batch size: 44, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:05:28,450 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.875e+02 2.915e+02 3.513e+02 4.432e+02 8.204e+02, threshold=7.026e+02, percent-clipped=5.0 2023-05-02 10:05:59,253 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7667, 1.7433, 1.5647, 1.3802, 1.9111, 1.4715, 1.4874, 1.9288], device='cuda:0'), covar=tensor([0.0232, 0.0414, 0.0566, 0.0459, 0.0283, 0.0348, 0.0181, 0.0249], device='cuda:0'), in_proj_covar=tensor([0.0224, 0.0240, 0.0230, 0.0231, 0.0241, 0.0240, 0.0238, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:06:23,850 INFO [train.py:904] (0/8) Epoch 27, batch 7000, loss[loss=0.2054, simple_loss=0.3018, pruned_loss=0.05454, over 16733.00 frames. ], tot_loss[loss=0.2033, simple_loss=0.2914, pruned_loss=0.0576, over 3079428.49 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:06:45,966 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=270917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:07:05,152 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2987, 4.0712, 3.9559, 2.4851, 3.6220, 4.0365, 3.6101, 2.2472], device='cuda:0'), covar=tensor([0.0593, 0.0062, 0.0064, 0.0480, 0.0110, 0.0112, 0.0109, 0.0500], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0090, 0.0091, 0.0137, 0.0102, 0.0116, 0.0098, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 10:07:15,675 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=270936.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:07:27,463 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6598, 4.8070, 4.9860, 4.7413, 4.8634, 5.3640, 4.8774, 4.6241], device='cuda:0'), covar=tensor([0.1197, 0.1883, 0.2352, 0.1982, 0.2309, 0.0998, 0.1655, 0.2461], device='cuda:0'), in_proj_covar=tensor([0.0424, 0.0630, 0.0692, 0.0514, 0.0682, 0.0721, 0.0540, 0.0689], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 10:07:40,704 INFO [train.py:904] (0/8) Epoch 27, batch 7050, loss[loss=0.2025, simple_loss=0.2909, pruned_loss=0.05705, over 16220.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.2923, pruned_loss=0.05722, over 3100296.12 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:08:01,042 INFO [optim.py:368] (0/8) 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 10:08:05,080 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=270969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:18,544 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=270978.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:08:27,520 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=270984.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 10:08:58,239 INFO [train.py:904] (0/8) Epoch 27, batch 7100, loss[loss=0.2114, simple_loss=0.3128, pruned_loss=0.05496, over 16889.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.2909, pruned_loss=0.05684, over 3099854.34 frames. ], batch size: 109, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:10:15,337 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:10:16,498 INFO [train.py:904] (0/8) Epoch 27, batch 7150, loss[loss=0.2436, simple_loss=0.3097, pruned_loss=0.08871, over 11391.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2892, pruned_loss=0.05682, over 3105439.80 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:10:21,404 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 10:10:37,971 INFO [optim.py:368] (0/8) 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 10:10:59,676 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:18,941 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271095.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:27,575 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:11:31,648 INFO [train.py:904] (0/8) Epoch 27, batch 7200, loss[loss=0.1915, simple_loss=0.293, pruned_loss=0.04502, over 16327.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2875, pruned_loss=0.05592, over 3089751.69 frames. ], batch size: 146, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:11:54,700 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6486, 3.6964, 2.4286, 4.3680, 2.8825, 4.2609, 2.5886, 3.0039], device='cuda:0'), covar=tensor([0.0320, 0.0429, 0.1631, 0.0194, 0.0865, 0.0513, 0.1496, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0180, 0.0194, 0.0170, 0.0178, 0.0218, 0.0201, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 10:12:37,366 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271143.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:12:53,499 INFO [train.py:904] (0/8) Epoch 27, batch 7250, loss[loss=0.1936, simple_loss=0.2715, pruned_loss=0.05781, over 16892.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2851, pruned_loss=0.05489, over 3081465.78 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:13:16,041 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.855e+02 2.666e+02 3.085e+02 3.699e+02 8.603e+02, threshold=6.169e+02, percent-clipped=3.0 2023-05-02 10:13:26,804 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9751, 2.1917, 2.2550, 3.3863, 2.1322, 2.4661, 2.2980, 2.3405], device='cuda:0'), covar=tensor([0.1499, 0.3288, 0.2902, 0.0698, 0.4207, 0.2353, 0.3308, 0.3218], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0470, 0.0381, 0.0333, 0.0443, 0.0537, 0.0441, 0.0548], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:14:10,773 INFO [train.py:904] (0/8) Epoch 27, batch 7300, loss[loss=0.1815, simple_loss=0.2765, pruned_loss=0.04322, over 16710.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2845, pruned_loss=0.05478, over 3085429.93 frames. ], batch size: 89, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:29,751 INFO [train.py:904] (0/8) Epoch 27, batch 7350, loss[loss=0.1897, simple_loss=0.2804, pruned_loss=0.04952, over 16189.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2855, pruned_loss=0.05581, over 3065890.17 frames. ], batch size: 165, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:15:42,229 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:15:50,538 INFO [optim.py:368] (0/8) 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 10:15:54,206 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:00,591 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:16:48,725 INFO [train.py:904] (0/8) Epoch 27, batch 7400, loss[loss=0.2335, simple_loss=0.3191, pruned_loss=0.07396, over 15487.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2865, pruned_loss=0.05606, over 3073886.66 frames. ], batch size: 190, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:17:10,663 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:17:15,050 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2263, 4.3007, 4.1172, 3.8550, 3.8509, 4.2286, 3.9356, 4.0083], device='cuda:0'), covar=tensor([0.0627, 0.0750, 0.0322, 0.0307, 0.0771, 0.0528, 0.0880, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0301, 0.0455, 0.0354, 0.0356, 0.0352, 0.0409, 0.0243, 0.0423], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:17:18,228 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:06,486 INFO [train.py:904] (0/8) Epoch 27, batch 7450, loss[loss=0.21, simple_loss=0.297, pruned_loss=0.06148, over 16917.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2875, pruned_loss=0.05712, over 3063210.69 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:18:30,891 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.940e+02 2.819e+02 3.560e+02 4.384e+02 9.484e+02, threshold=7.119e+02, percent-clipped=1.0 2023-05-02 10:18:45,328 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:18:55,900 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271382.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:19:15,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8641, 5.1592, 5.3524, 5.0886, 5.1673, 5.6820, 5.1283, 4.8951], device='cuda:0'), covar=tensor([0.1051, 0.1856, 0.2387, 0.1917, 0.2192, 0.0916, 0.1693, 0.2349], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0631, 0.0694, 0.0514, 0.0686, 0.0724, 0.0545, 0.0691], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 10:19:30,003 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 10:19:30,458 INFO [train.py:904] (0/8) Epoch 27, batch 7500, loss[loss=0.1813, simple_loss=0.2642, pruned_loss=0.04925, over 16644.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2877, pruned_loss=0.05605, over 3059393.89 frames. ], batch size: 57, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:20:08,064 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:14,062 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:22,952 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271436.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:20:49,677 INFO [train.py:904] (0/8) Epoch 27, batch 7550, loss[loss=0.1634, simple_loss=0.2585, pruned_loss=0.03417, over 16811.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2872, pruned_loss=0.05653, over 3048381.40 frames. ], batch size: 102, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:21:11,194 INFO [optim.py:368] (0/8) 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 10:21:41,051 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:21:56,017 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.71 vs. limit=5.0 2023-05-02 10:22:05,503 INFO [train.py:904] (0/8) Epoch 27, batch 7600, loss[loss=0.1793, simple_loss=0.2659, pruned_loss=0.04634, over 16694.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2859, pruned_loss=0.05672, over 3058425.87 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:22,945 INFO [train.py:904] (0/8) Epoch 27, batch 7650, loss[loss=0.2111, simple_loss=0.3025, pruned_loss=0.05979, over 15240.00 frames. ], tot_loss[loss=0.2, simple_loss=0.286, pruned_loss=0.05702, over 3055682.07 frames. ], batch size: 191, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:23:45,490 INFO [optim.py:368] (0/8) 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 10:23:55,624 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271573.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:23:58,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2696, 3.0134, 3.2905, 1.7602, 3.3994, 3.4632, 2.7327, 2.6421], device='cuda:0'), covar=tensor([0.0833, 0.0318, 0.0236, 0.1283, 0.0112, 0.0249, 0.0525, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0110, 0.0101, 0.0138, 0.0086, 0.0130, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 10:24:43,693 INFO [train.py:904] (0/8) Epoch 27, batch 7700, loss[loss=0.1812, simple_loss=0.275, pruned_loss=0.04366, over 16527.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2853, pruned_loss=0.05679, over 3064216.13 frames. ], batch size: 75, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:24:58,649 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 10:25:06,084 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:25:12,499 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271621.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:25:12,546 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9262, 5.2157, 5.4209, 5.1722, 5.2464, 5.7999, 5.2118, 4.9286], device='cuda:0'), covar=tensor([0.1013, 0.1857, 0.2363, 0.1957, 0.2268, 0.0829, 0.1618, 0.2388], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0630, 0.0694, 0.0515, 0.0685, 0.0722, 0.0544, 0.0690], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 10:25:25,327 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2992, 5.2543, 5.1063, 4.2506, 5.1762, 1.7429, 4.9097, 4.7458], device='cuda:0'), covar=tensor([0.0132, 0.0116, 0.0245, 0.0477, 0.0115, 0.3080, 0.0170, 0.0274], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0171, 0.0209, 0.0182, 0.0184, 0.0213, 0.0196, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:26:02,428 INFO [train.py:904] (0/8) Epoch 27, batch 7750, loss[loss=0.1907, simple_loss=0.2833, pruned_loss=0.04902, over 16855.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2853, pruned_loss=0.05623, over 3083342.98 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:26:24,500 INFO [optim.py:368] (0/8) 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 10:27:20,091 INFO [train.py:904] (0/8) Epoch 27, batch 7800, loss[loss=0.2421, simple_loss=0.3099, pruned_loss=0.08712, over 11381.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2873, pruned_loss=0.0577, over 3062955.30 frames. ], batch size: 248, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:27:56,718 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6604, 3.5327, 3.9624, 1.7915, 4.0598, 4.1433, 3.1726, 3.0427], device='cuda:0'), covar=tensor([0.0880, 0.0302, 0.0243, 0.1491, 0.0117, 0.0198, 0.0445, 0.0535], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0138, 0.0086, 0.0131, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 10:28:04,238 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271731.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:28:29,902 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7021, 4.5542, 4.7513, 4.8983, 5.0762, 4.5776, 5.0518, 5.0682], device='cuda:0'), covar=tensor([0.2102, 0.1309, 0.1678, 0.0785, 0.0630, 0.1071, 0.0714, 0.0703], device='cuda:0'), in_proj_covar=tensor([0.0658, 0.0801, 0.0933, 0.0812, 0.0622, 0.0651, 0.0682, 0.0794], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:28:36,668 INFO [train.py:904] (0/8) Epoch 27, batch 7850, loss[loss=0.2395, simple_loss=0.3097, pruned_loss=0.08465, over 11714.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2882, pruned_loss=0.05742, over 3070132.94 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:28:57,994 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.815e+02 2.838e+02 3.379e+02 4.179e+02 9.334e+02, threshold=6.757e+02, percent-clipped=5.0 2023-05-02 10:29:20,353 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=271782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:29:52,548 INFO [train.py:904] (0/8) Epoch 27, batch 7900, loss[loss=0.1982, simple_loss=0.2854, pruned_loss=0.05545, over 16841.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.287, pruned_loss=0.05663, over 3083663.63 frames. ], batch size: 116, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:31:12,164 INFO [train.py:904] (0/8) Epoch 27, batch 7950, loss[loss=0.1962, simple_loss=0.2795, pruned_loss=0.0564, over 16777.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.2874, pruned_loss=0.05723, over 3071582.61 frames. ], batch size: 39, lr: 2.48e-03, grad_scale: 4.0 2023-05-02 10:31:15,290 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271854.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:31:36,733 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.948e+02 2.764e+02 3.233e+02 3.778e+02 5.723e+02, threshold=6.466e+02, percent-clipped=0.0 2023-05-02 10:31:50,005 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.26 vs. limit=5.0 2023-05-02 10:32:14,118 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9834, 3.3599, 3.3879, 2.1719, 2.9470, 2.1926, 3.4251, 3.6990], device='cuda:0'), covar=tensor([0.0362, 0.0827, 0.0690, 0.2248, 0.0969, 0.1183, 0.0742, 0.0945], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0170, 0.0170, 0.0156, 0.0148, 0.0133, 0.0146, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 10:32:17,849 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7689, 3.9462, 2.2308, 4.6230, 3.0243, 4.4507, 2.1414, 2.9851], device='cuda:0'), covar=tensor([0.0328, 0.0408, 0.2053, 0.0225, 0.0856, 0.0566, 0.2138, 0.1039], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0172, 0.0180, 0.0220, 0.0204, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 10:32:29,930 INFO [train.py:904] (0/8) Epoch 27, batch 8000, loss[loss=0.1868, simple_loss=0.2785, pruned_loss=0.0475, over 16767.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.288, pruned_loss=0.05786, over 3068961.35 frames. ], batch size: 124, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:32:42,423 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-05-02 10:32:49,727 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271915.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:32:52,971 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=271917.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:07,795 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1650, 5.4646, 5.2320, 5.2600, 4.9695, 4.9121, 4.8344, 5.5829], device='cuda:0'), covar=tensor([0.1275, 0.0834, 0.1017, 0.0932, 0.0814, 0.0901, 0.1267, 0.0806], device='cuda:0'), in_proj_covar=tensor([0.0703, 0.0850, 0.0699, 0.0656, 0.0536, 0.0539, 0.0714, 0.0667], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:33:19,443 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=271934.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:33:47,193 INFO [train.py:904] (0/8) Epoch 27, batch 8050, loss[loss=0.1814, simple_loss=0.271, pruned_loss=0.04584, over 16661.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2874, pruned_loss=0.05755, over 3052635.20 frames. ], batch size: 62, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:33:50,093 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.40 vs. limit=2.0 2023-05-02 10:34:05,656 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=271965.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:34:09,135 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.787e+02 2.577e+02 3.144e+02 3.701e+02 6.468e+02, threshold=6.287e+02, percent-clipped=1.0 2023-05-02 10:34:49,582 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=271995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:34:57,125 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-272000.pt 2023-05-02 10:35:04,683 INFO [train.py:904] (0/8) Epoch 27, batch 8100, loss[loss=0.1895, simple_loss=0.2793, pruned_loss=0.04981, over 16938.00 frames. ], tot_loss[loss=0.2007, simple_loss=0.2872, pruned_loss=0.05711, over 3053909.86 frames. ], batch size: 109, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:35:45,761 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:36:17,763 INFO [train.py:904] (0/8) Epoch 27, batch 8150, loss[loss=0.1649, simple_loss=0.2518, pruned_loss=0.03899, over 17218.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2836, pruned_loss=0.05531, over 3082829.72 frames. ], batch size: 45, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:36:39,767 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:01,597 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:37:32,138 INFO [train.py:904] (0/8) Epoch 27, batch 8200, loss[loss=0.2076, simple_loss=0.2808, pruned_loss=0.06716, over 11838.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2811, pruned_loss=0.05426, over 3100846.99 frames. ], batch size: 246, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:38:12,575 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:16,440 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:38:33,077 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5878, 3.6491, 3.3914, 3.0342, 3.2256, 3.5299, 3.3552, 3.3350], device='cuda:0'), covar=tensor([0.0545, 0.0608, 0.0306, 0.0277, 0.0534, 0.0455, 0.1383, 0.0480], device='cuda:0'), in_proj_covar=tensor([0.0300, 0.0454, 0.0353, 0.0354, 0.0352, 0.0407, 0.0243, 0.0422], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:38:53,594 INFO [train.py:904] (0/8) Epoch 27, batch 8250, loss[loss=0.1721, simple_loss=0.2641, pruned_loss=0.0401, over 11797.00 frames. ], tot_loss[loss=0.1921, simple_loss=0.2801, pruned_loss=0.05206, over 3074524.35 frames. ], batch size: 247, lr: 2.48e-03, grad_scale: 8.0 2023-05-02 10:39:19,037 INFO [optim.py:368] (0/8) 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 10:39:40,111 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272181.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:39:50,241 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272188.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:03,752 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7204, 2.1994, 1.9305, 2.0458, 2.4946, 2.2352, 2.1277, 2.6470], device='cuda:0'), covar=tensor([0.0248, 0.0503, 0.0604, 0.0583, 0.0342, 0.0462, 0.0270, 0.0328], device='cuda:0'), in_proj_covar=tensor([0.0223, 0.0240, 0.0229, 0.0230, 0.0241, 0.0239, 0.0237, 0.0237], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:40:14,330 INFO [train.py:904] (0/8) Epoch 27, batch 8300, loss[loss=0.1482, simple_loss=0.238, pruned_loss=0.02922, over 12035.00 frames. ], tot_loss[loss=0.188, simple_loss=0.2776, pruned_loss=0.04921, over 3069432.19 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:40:26,351 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:40:45,487 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272223.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:15,132 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272242.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 10:41:22,862 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272247.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:41:31,419 INFO [train.py:904] (0/8) Epoch 27, batch 8350, loss[loss=0.1852, simple_loss=0.2817, pruned_loss=0.04433, over 16134.00 frames. ], tot_loss[loss=0.1862, simple_loss=0.2773, pruned_loss=0.04753, over 3066476.35 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:41:54,866 INFO [optim.py:368] (0/8) 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 10:42:20,359 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272284.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:30,783 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272290.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:42:50,447 INFO [train.py:904] (0/8) Epoch 27, batch 8400, loss[loss=0.1645, simple_loss=0.2597, pruned_loss=0.03467, over 16463.00 frames. ], tot_loss[loss=0.1827, simple_loss=0.2748, pruned_loss=0.04532, over 3081452.35 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:42:59,281 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:43:41,161 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.45 vs. limit=2.0 2023-05-02 10:44:09,905 INFO [train.py:904] (0/8) Epoch 27, batch 8450, loss[loss=0.1585, simple_loss=0.2607, pruned_loss=0.02816, over 16897.00 frames. ], tot_loss[loss=0.1805, simple_loss=0.2732, pruned_loss=0.0439, over 3076561.09 frames. ], batch size: 90, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:44:34,141 INFO [optim.py:368] (0/8) 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 10:44:42,795 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272373.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:45:00,446 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0496, 1.8404, 1.6288, 1.4885, 1.9881, 1.5830, 1.5527, 1.9664], device='cuda:0'), covar=tensor([0.0282, 0.0379, 0.0564, 0.0468, 0.0296, 0.0350, 0.0210, 0.0250], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0238, 0.0227, 0.0229, 0.0239, 0.0237, 0.0236, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:45:32,403 INFO [train.py:904] (0/8) Epoch 27, batch 8500, loss[loss=0.156, simple_loss=0.2576, pruned_loss=0.02719, over 15125.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.269, pruned_loss=0.04188, over 3038516.76 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:46:23,562 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:46:57,064 INFO [train.py:904] (0/8) Epoch 27, batch 8550, loss[loss=0.1769, simple_loss=0.275, pruned_loss=0.03935, over 15365.00 frames. ], tot_loss[loss=0.1746, simple_loss=0.2672, pruned_loss=0.04102, over 3026809.35 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:47:24,044 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4948, 3.2885, 2.7064, 2.1737, 2.1717, 2.3588, 3.5191, 3.0198], device='cuda:0'), covar=tensor([0.2905, 0.0786, 0.1941, 0.3305, 0.3161, 0.2352, 0.0466, 0.1535], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0270, 0.0307, 0.0321, 0.0299, 0.0271, 0.0299, 0.0342], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 10:47:25,009 INFO [optim.py:368] (0/8) 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 10:47:55,127 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272483.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:48:09,970 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5970, 3.5365, 3.5239, 2.5543, 3.4118, 1.9540, 3.1931, 2.8028], device='cuda:0'), covar=tensor([0.0129, 0.0125, 0.0182, 0.0201, 0.0108, 0.2684, 0.0124, 0.0268], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0170, 0.0207, 0.0181, 0.0183, 0.0213, 0.0196, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:48:35,734 INFO [train.py:904] (0/8) Epoch 27, batch 8600, loss[loss=0.1814, simple_loss=0.2807, pruned_loss=0.041, over 16523.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2675, pruned_loss=0.04021, over 3032229.81 frames. ], batch size: 68, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:48:50,716 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272510.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:48:53,611 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7762, 2.2450, 1.9158, 2.1191, 2.5411, 2.2494, 2.1685, 2.7142], device='cuda:0'), covar=tensor([0.0202, 0.0506, 0.0655, 0.0549, 0.0327, 0.0484, 0.0204, 0.0313], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0239, 0.0227, 0.0229, 0.0239, 0.0237, 0.0236, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:49:30,751 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.48 vs. limit=2.0 2023-05-02 10:49:34,437 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9645, 1.8813, 1.6870, 1.5508, 2.0147, 1.6496, 1.5749, 1.9818], device='cuda:0'), covar=tensor([0.0223, 0.0339, 0.0453, 0.0410, 0.0265, 0.0323, 0.0182, 0.0231], device='cuda:0'), in_proj_covar=tensor([0.0220, 0.0239, 0.0228, 0.0229, 0.0240, 0.0237, 0.0236, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 10:49:43,066 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272537.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 10:50:13,939 INFO [train.py:904] (0/8) Epoch 27, batch 8650, loss[loss=0.1473, simple_loss=0.249, pruned_loss=0.02283, over 16171.00 frames. ], tot_loss[loss=0.1718, simple_loss=0.2655, pruned_loss=0.03902, over 3007185.72 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:50:26,983 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:50:29,715 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.36 vs. limit=2.0 2023-05-02 10:50:48,991 INFO [optim.py:368] (0/8) 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 10:51:15,186 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:51:36,583 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:00,252 INFO [train.py:904] (0/8) Epoch 27, batch 8700, loss[loss=0.1683, simple_loss=0.2658, pruned_loss=0.03543, over 15181.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2628, pruned_loss=0.03762, over 3047451.27 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:52:01,219 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:20,937 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=272613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:52:56,713 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8144, 3.0959, 2.7343, 4.8475, 3.4725, 4.3291, 1.7574, 3.2972], device='cuda:0'), covar=tensor([0.1574, 0.0763, 0.1229, 0.0184, 0.0189, 0.0352, 0.1771, 0.0697], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0177, 0.0197, 0.0196, 0.0203, 0.0215, 0.0206, 0.0195], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 10:53:06,584 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272638.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:53:35,411 INFO [train.py:904] (0/8) Epoch 27, batch 8750, loss[loss=0.153, simple_loss=0.2439, pruned_loss=0.03108, over 12015.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2622, pruned_loss=0.0372, over 3043505.48 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 10:53:56,729 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 10:54:15,880 INFO [optim.py:368] (0/8) 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 10:54:28,904 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=272674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:54:57,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 10:55:27,081 INFO [train.py:904] (0/8) Epoch 27, batch 8800, loss[loss=0.1439, simple_loss=0.2459, pruned_loss=0.02094, over 17126.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2602, pruned_loss=0.03578, over 3049096.15 frames. ], batch size: 49, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:56:22,549 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:57:12,476 INFO [train.py:904] (0/8) Epoch 27, batch 8850, loss[loss=0.1505, simple_loss=0.243, pruned_loss=0.02898, over 12419.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2632, pruned_loss=0.03542, over 3036701.06 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 10:57:46,554 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.085e+02 2.509e+02 2.907e+02 5.151e+02, threshold=5.018e+02, percent-clipped=0.0 2023-05-02 10:58:17,295 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272783.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 10:58:57,772 INFO [train.py:904] (0/8) Epoch 27, batch 8900, loss[loss=0.1733, simple_loss=0.2712, pruned_loss=0.0377, over 16881.00 frames. ], tot_loss[loss=0.167, simple_loss=0.264, pruned_loss=0.03496, over 3043248.24 frames. ], batch size: 116, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:00:01,531 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:00:20,784 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272837.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 11:00:48,065 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7137, 5.0421, 4.8316, 4.8600, 4.5815, 4.5551, 4.4439, 5.1405], device='cuda:0'), covar=tensor([0.1330, 0.0984, 0.1139, 0.0910, 0.0872, 0.1193, 0.1342, 0.0856], device='cuda:0'), in_proj_covar=tensor([0.0691, 0.0837, 0.0687, 0.0643, 0.0527, 0.0529, 0.0700, 0.0655], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:01:00,400 INFO [train.py:904] (0/8) Epoch 27, batch 8950, loss[loss=0.1512, simple_loss=0.259, pruned_loss=0.02174, over 16853.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2641, pruned_loss=0.03529, over 3062442.69 frames. ], batch size: 90, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:01:35,340 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.412e+02 1.963e+02 2.469e+02 3.081e+02 5.293e+02, threshold=4.938e+02, percent-clipped=1.0 2023-05-02 11:01:56,793 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272879.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:02:11,912 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272885.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 11:02:33,673 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0844, 4.0451, 3.9538, 3.1854, 3.9919, 1.8419, 3.8001, 3.5686], device='cuda:0'), covar=tensor([0.0119, 0.0113, 0.0180, 0.0288, 0.0108, 0.2852, 0.0139, 0.0291], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0169, 0.0206, 0.0180, 0.0183, 0.0213, 0.0195, 0.0173], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:02:49,725 INFO [train.py:904] (0/8) Epoch 27, batch 9000, loss[loss=0.1429, simple_loss=0.2344, pruned_loss=0.02567, over 16612.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2609, pruned_loss=0.0342, over 3067110.02 frames. ], batch size: 89, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:02:49,726 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 11:02:59,738 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 11:03:00,998 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=272903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:03:51,482 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:42,536 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=272951.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:04:44,610 INFO [train.py:904] (0/8) Epoch 27, batch 9050, loss[loss=0.1713, simple_loss=0.2632, pruned_loss=0.03973, over 16216.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2619, pruned_loss=0.03489, over 3071984.09 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:05:18,777 INFO [optim.py:368] (0/8) 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 11:05:19,346 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=272969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:06:32,241 INFO [train.py:904] (0/8) Epoch 27, batch 9100, loss[loss=0.166, simple_loss=0.2526, pruned_loss=0.03969, over 12192.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2615, pruned_loss=0.0352, over 3072859.60 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:07:30,828 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4270, 4.5561, 4.6891, 4.4291, 4.6069, 5.0532, 4.5648, 4.2234], device='cuda:0'), covar=tensor([0.1409, 0.2015, 0.2334, 0.2077, 0.2156, 0.0927, 0.1763, 0.2700], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0614, 0.0672, 0.0501, 0.0666, 0.0706, 0.0531, 0.0668], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:07:36,046 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273029.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:07:49,100 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5574, 3.7484, 2.7858, 2.2245, 2.3824, 2.4468, 4.0845, 3.1878], device='cuda:0'), covar=tensor([0.3285, 0.0650, 0.2109, 0.3356, 0.3101, 0.2393, 0.0410, 0.1632], device='cuda:0'), in_proj_covar=tensor([0.0328, 0.0268, 0.0306, 0.0319, 0.0297, 0.0270, 0.0297, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:08:18,713 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-02 11:08:32,215 INFO [train.py:904] (0/8) Epoch 27, batch 9150, loss[loss=0.1554, simple_loss=0.2434, pruned_loss=0.03372, over 12100.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2614, pruned_loss=0.03462, over 3068326.71 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:09:08,132 INFO [optim.py:368] (0/8) 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 11:09:27,709 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273077.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:10:03,537 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3867, 3.0040, 2.7065, 2.3005, 2.1762, 2.3463, 3.0481, 2.7526], device='cuda:0'), covar=tensor([0.2830, 0.0720, 0.1836, 0.2792, 0.2765, 0.2334, 0.0474, 0.1652], device='cuda:0'), in_proj_covar=tensor([0.0329, 0.0270, 0.0308, 0.0321, 0.0298, 0.0271, 0.0299, 0.0343], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:10:18,361 INFO [train.py:904] (0/8) Epoch 27, batch 9200, loss[loss=0.1712, simple_loss=0.2615, pruned_loss=0.0404, over 16681.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2575, pruned_loss=0.03402, over 3069313.74 frames. ], batch size: 62, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:10:46,305 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1267, 5.1064, 4.8147, 4.3586, 4.9948, 1.9548, 4.7603, 4.7033], device='cuda:0'), covar=tensor([0.0180, 0.0207, 0.0313, 0.0437, 0.0187, 0.2849, 0.0191, 0.0282], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0169, 0.0205, 0.0178, 0.0182, 0.0211, 0.0194, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:11:24,895 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 11:11:38,908 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.89 vs. limit=5.0 2023-05-02 11:11:53,372 INFO [train.py:904] (0/8) Epoch 27, batch 9250, loss[loss=0.1661, simple_loss=0.2582, pruned_loss=0.03699, over 16977.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2575, pruned_loss=0.03411, over 3078657.69 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:12:25,536 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.747e+02 2.225e+02 2.595e+02 3.229e+02 5.341e+02, threshold=5.191e+02, percent-clipped=1.0 2023-05-02 11:12:33,863 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1150, 2.1967, 2.6024, 3.1426, 2.8926, 3.6535, 2.3228, 3.4670], device='cuda:0'), covar=tensor([0.0230, 0.0564, 0.0406, 0.0317, 0.0353, 0.0150, 0.0554, 0.0189], device='cuda:0'), in_proj_covar=tensor([0.0191, 0.0193, 0.0181, 0.0184, 0.0201, 0.0159, 0.0197, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:12:33,934 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8519, 4.4212, 3.2324, 2.4159, 2.7469, 2.7548, 4.8234, 3.6072], device='cuda:0'), covar=tensor([0.2815, 0.0463, 0.1700, 0.2998, 0.2981, 0.1999, 0.0272, 0.1293], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0268, 0.0306, 0.0319, 0.0296, 0.0269, 0.0297, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:12:41,093 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.43 vs. limit=5.0 2023-05-02 11:13:43,822 INFO [train.py:904] (0/8) Epoch 27, batch 9300, loss[loss=0.1535, simple_loss=0.2349, pruned_loss=0.03602, over 12230.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2559, pruned_loss=0.0336, over 3067683.16 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:14:12,046 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 11:15:28,955 INFO [train.py:904] (0/8) Epoch 27, batch 9350, loss[loss=0.172, simple_loss=0.2679, pruned_loss=0.03805, over 16200.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2555, pruned_loss=0.03348, over 3073185.30 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:15:45,229 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 11:16:02,980 INFO [optim.py:368] (0/8) 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 11:16:03,990 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=273269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:16:15,934 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 11:16:47,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8760, 2.8139, 2.6059, 1.9522, 2.5372, 2.8285, 2.6328, 1.9359], device='cuda:0'), covar=tensor([0.0443, 0.0086, 0.0085, 0.0402, 0.0167, 0.0112, 0.0126, 0.0485], device='cuda:0'), in_proj_covar=tensor([0.0135, 0.0087, 0.0088, 0.0134, 0.0099, 0.0112, 0.0095, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 11:17:10,303 INFO [train.py:904] (0/8) Epoch 27, batch 9400, loss[loss=0.1718, simple_loss=0.2772, pruned_loss=0.03318, over 16514.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2557, pruned_loss=0.03369, over 3054023.36 frames. ], batch size: 68, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:17:39,076 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=273317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:18:45,076 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0631, 4.1433, 3.9834, 3.6894, 3.7210, 4.0716, 3.7393, 3.8377], device='cuda:0'), covar=tensor([0.0570, 0.0552, 0.0321, 0.0304, 0.0666, 0.0550, 0.0986, 0.0629], device='cuda:0'), in_proj_covar=tensor([0.0297, 0.0445, 0.0347, 0.0349, 0.0343, 0.0401, 0.0239, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:18:52,060 INFO [train.py:904] (0/8) Epoch 27, batch 9450, loss[loss=0.1738, simple_loss=0.2713, pruned_loss=0.03819, over 15278.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.257, pruned_loss=0.03372, over 3036529.60 frames. ], batch size: 190, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:19:21,972 INFO [optim.py:368] (0/8) 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 11:20:32,363 INFO [train.py:904] (0/8) Epoch 27, batch 9500, loss[loss=0.168, simple_loss=0.2588, pruned_loss=0.03859, over 16941.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2563, pruned_loss=0.03342, over 3043588.47 frames. ], batch size: 109, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:20:49,267 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273409.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:20:49,790 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 11:21:13,519 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.5331, 5.9074, 5.6318, 5.7131, 5.2824, 5.3555, 5.2967, 5.9855], device='cuda:0'), covar=tensor([0.1357, 0.0866, 0.0948, 0.0828, 0.0838, 0.0614, 0.1278, 0.0836], device='cuda:0'), in_proj_covar=tensor([0.0689, 0.0835, 0.0682, 0.0642, 0.0525, 0.0528, 0.0698, 0.0653], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:21:38,351 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=273435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:21:40,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1044, 5.4923, 5.6414, 5.3566, 5.4645, 5.9930, 5.5434, 5.1678], device='cuda:0'), covar=tensor([0.0893, 0.1800, 0.2462, 0.2063, 0.2438, 0.0864, 0.1558, 0.2527], device='cuda:0'), in_proj_covar=tensor([0.0406, 0.0605, 0.0664, 0.0492, 0.0655, 0.0697, 0.0523, 0.0656], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:22:02,047 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7221, 1.8703, 2.2977, 2.7280, 2.5675, 3.0633, 2.0451, 2.9954], device='cuda:0'), covar=tensor([0.0247, 0.0651, 0.0409, 0.0355, 0.0403, 0.0214, 0.0616, 0.0191], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0194, 0.0182, 0.0185, 0.0202, 0.0160, 0.0198, 0.0160], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:22:18,628 INFO [train.py:904] (0/8) Epoch 27, batch 9550, loss[loss=0.1587, simple_loss=0.2478, pruned_loss=0.03484, over 12149.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2561, pruned_loss=0.03347, over 3047054.36 frames. ], batch size: 246, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:22:51,234 INFO [optim.py:368] (0/8) 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 11:22:53,722 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=273470.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:23:46,456 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 27, batch 9600, loss[loss=0.1685, simple_loss=0.255, pruned_loss=0.04096, over 12686.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2581, pruned_loss=0.03439, over 3045936.72 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:24:57,492 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.74 vs. limit=5.0 2023-05-02 11:25:44,222 INFO [train.py:904] (0/8) Epoch 27, batch 9650, loss[loss=0.1563, simple_loss=0.2595, pruned_loss=0.02662, over 16181.00 frames. ], tot_loss[loss=0.164, simple_loss=0.259, pruned_loss=0.03444, over 3019180.24 frames. ], batch size: 165, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:26:23,247 INFO [optim.py:368] (0/8) 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 11:27:30,806 INFO [train.py:904] (0/8) Epoch 27, batch 9700, loss[loss=0.2029, simple_loss=0.2857, pruned_loss=0.06004, over 16903.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2584, pruned_loss=0.03431, over 3038608.72 frames. ], batch size: 116, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:27:31,403 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9567, 4.9973, 5.3542, 5.3370, 5.3470, 5.0810, 5.0091, 4.9246], device='cuda:0'), covar=tensor([0.0333, 0.0750, 0.0426, 0.0392, 0.0399, 0.0410, 0.0823, 0.0425], device='cuda:0'), in_proj_covar=tensor([0.0411, 0.0460, 0.0448, 0.0412, 0.0494, 0.0473, 0.0540, 0.0379], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 11:28:04,325 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7942, 2.2456, 1.9985, 1.9671, 2.5488, 2.2079, 2.2779, 2.6552], device='cuda:0'), covar=tensor([0.0191, 0.0441, 0.0538, 0.0552, 0.0315, 0.0432, 0.0230, 0.0302], device='cuda:0'), in_proj_covar=tensor([0.0214, 0.0234, 0.0223, 0.0225, 0.0235, 0.0232, 0.0229, 0.0229], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:29:12,479 INFO [train.py:904] (0/8) Epoch 27, batch 9750, loss[loss=0.1717, simple_loss=0.2629, pruned_loss=0.04023, over 12638.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2583, pruned_loss=0.03453, over 3054029.85 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:29:23,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4799, 3.5135, 2.6366, 2.1048, 2.1807, 2.3304, 3.8071, 3.0215], device='cuda:0'), covar=tensor([0.3163, 0.0667, 0.2115, 0.3181, 0.3083, 0.2405, 0.0379, 0.1500], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0268, 0.0306, 0.0319, 0.0295, 0.0269, 0.0297, 0.0341], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:29:42,115 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.247e+02 2.115e+02 2.683e+02 3.309e+02 9.068e+02, threshold=5.365e+02, percent-clipped=4.0 2023-05-02 11:30:08,795 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8365, 1.3140, 1.7951, 1.6925, 1.8345, 1.9027, 1.6833, 1.8389], device='cuda:0'), covar=tensor([0.0318, 0.0526, 0.0262, 0.0381, 0.0362, 0.0255, 0.0490, 0.0181], device='cuda:0'), in_proj_covar=tensor([0.0192, 0.0194, 0.0181, 0.0185, 0.0201, 0.0160, 0.0198, 0.0159], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:30:48,443 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 11:30:51,297 INFO [train.py:904] (0/8) Epoch 27, batch 9800, loss[loss=0.1591, simple_loss=0.262, pruned_loss=0.02812, over 16358.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2584, pruned_loss=0.03384, over 3055172.85 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:31:27,421 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.63 vs. limit=2.0 2023-05-02 11:32:36,072 INFO [train.py:904] (0/8) Epoch 27, batch 9850, loss[loss=0.1724, simple_loss=0.266, pruned_loss=0.03944, over 15377.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2601, pruned_loss=0.03386, over 3066967.84 frames. ], batch size: 192, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:33:00,307 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273765.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:33:11,083 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.328e+02 2.029e+02 2.351e+02 3.028e+02 6.435e+02, threshold=4.703e+02, percent-clipped=1.0 2023-05-02 11:34:02,567 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=273791.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:34:25,790 INFO [train.py:904] (0/8) Epoch 27, batch 9900, loss[loss=0.1672, simple_loss=0.2689, pruned_loss=0.03272, over 15396.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2608, pruned_loss=0.0338, over 3069209.81 frames. ], batch size: 191, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:36:22,587 INFO [train.py:904] (0/8) Epoch 27, batch 9950, loss[loss=0.163, simple_loss=0.2649, pruned_loss=0.03058, over 16334.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2627, pruned_loss=0.03389, over 3080721.49 frames. ], batch size: 146, lr: 2.47e-03, grad_scale: 4.0 2023-05-02 11:37:03,260 INFO [optim.py:368] (0/8) 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 11:38:25,131 INFO [train.py:904] (0/8) Epoch 27, batch 10000, loss[loss=0.1743, simple_loss=0.2677, pruned_loss=0.04042, over 16787.00 frames. ], tot_loss[loss=0.165, simple_loss=0.2619, pruned_loss=0.03406, over 3072937.09 frames. ], batch size: 124, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:38:32,943 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.60 vs. limit=2.0 2023-05-02 11:38:55,443 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.50 vs. limit=2.0 2023-05-02 11:40:06,639 INFO [train.py:904] (0/8) Epoch 27, batch 10050, loss[loss=0.1731, simple_loss=0.2659, pruned_loss=0.04021, over 12345.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.262, pruned_loss=0.0341, over 3070253.80 frames. ], batch size: 247, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:40:39,175 INFO [optim.py:368] (0/8) 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 11:41:34,532 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0260, 5.0226, 4.7516, 4.2717, 4.8820, 1.9684, 4.6788, 4.6181], device='cuda:0'), covar=tensor([0.0149, 0.0185, 0.0243, 0.0342, 0.0128, 0.2641, 0.0154, 0.0225], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0166, 0.0201, 0.0174, 0.0179, 0.0209, 0.0191, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:41:35,619 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-274000.pt 2023-05-02 11:41:40,578 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274001.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:41:43,727 INFO [train.py:904] (0/8) Epoch 27, batch 10100, loss[loss=0.1648, simple_loss=0.2492, pruned_loss=0.0402, over 12624.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.262, pruned_loss=0.03408, over 3069559.27 frames. ], batch size: 248, lr: 2.47e-03, grad_scale: 8.0 2023-05-02 11:43:03,196 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-27.pt 2023-05-02 11:43:27,176 INFO [train.py:904] (0/8) Epoch 28, batch 0, loss[loss=0.1658, simple_loss=0.2639, pruned_loss=0.03387, over 17140.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2639, pruned_loss=0.03387, over 17140.00 frames. ], batch size: 49, lr: 2.42e-03, grad_scale: 8.0 2023-05-02 11:43:27,177 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 11:43:34,595 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 11:43:48,103 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274062.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:43:52,104 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:01,204 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.517e+02 2.305e+02 2.657e+02 3.293e+02 5.505e+02, threshold=5.314e+02, percent-clipped=2.0 2023-05-02 11:44:27,541 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274091.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:44:43,859 INFO [train.py:904] (0/8) Epoch 28, batch 50, loss[loss=0.1614, simple_loss=0.2461, pruned_loss=0.0383, over 16501.00 frames. ], tot_loss[loss=0.1765, simple_loss=0.2633, pruned_loss=0.04485, over 754518.29 frames. ], batch size: 75, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:44:59,067 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274113.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:09,261 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5283, 4.4445, 4.4285, 4.1798, 4.2299, 4.4717, 4.2804, 4.2500], device='cuda:0'), covar=tensor([0.0689, 0.0934, 0.0398, 0.0366, 0.0734, 0.0637, 0.0598, 0.0673], device='cuda:0'), in_proj_covar=tensor([0.0294, 0.0440, 0.0344, 0.0345, 0.0340, 0.0396, 0.0237, 0.0410], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:45:24,160 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:34,059 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274139.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:45:54,006 INFO [train.py:904] (0/8) Epoch 28, batch 100, loss[loss=0.1778, simple_loss=0.2574, pruned_loss=0.04908, over 16719.00 frames. ], tot_loss[loss=0.1727, simple_loss=0.2606, pruned_loss=0.04235, over 1323778.33 frames. ], batch size: 134, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:46:20,300 INFO [optim.py:368] (0/8) 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 11:46:36,772 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.14 vs. limit=5.0 2023-05-02 11:46:47,953 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:47:02,022 INFO [train.py:904] (0/8) Epoch 28, batch 150, loss[loss=0.1906, simple_loss=0.2661, pruned_loss=0.05756, over 16700.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.258, pruned_loss=0.0426, over 1769229.74 frames. ], batch size: 124, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:47:17,171 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-02 11:48:08,600 INFO [train.py:904] (0/8) Epoch 28, batch 200, loss[loss=0.1804, simple_loss=0.261, pruned_loss=0.04987, over 15891.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2593, pruned_loss=0.04363, over 2115462.58 frames. ], batch size: 35, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:48:19,118 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 11:48:21,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 11:48:29,604 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:48:33,861 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 2.228e+02 2.594e+02 3.067e+02 5.865e+02, threshold=5.188e+02, percent-clipped=1.0 2023-05-02 11:48:51,062 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.56 vs. limit=5.0 2023-05-02 11:48:54,304 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8603, 2.8633, 2.7571, 4.9217, 3.9194, 4.3439, 1.6841, 3.2909], device='cuda:0'), covar=tensor([0.1439, 0.0843, 0.1276, 0.0233, 0.0203, 0.0404, 0.1762, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0178, 0.0199, 0.0197, 0.0202, 0.0216, 0.0208, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 11:49:16,131 INFO [train.py:904] (0/8) Epoch 28, batch 250, loss[loss=0.1615, simple_loss=0.2604, pruned_loss=0.03131, over 17138.00 frames. ], tot_loss[loss=0.1714, simple_loss=0.2571, pruned_loss=0.04289, over 2380511.23 frames. ], batch size: 48, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:49:16,555 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274303.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:49:52,018 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274330.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:12,616 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0497, 3.0983, 2.8345, 2.9584, 3.3732, 3.1687, 3.5947, 3.5749], device='cuda:0'), covar=tensor([0.0158, 0.0453, 0.0506, 0.0457, 0.0298, 0.0358, 0.0255, 0.0266], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0244, 0.0233, 0.0234, 0.0246, 0.0243, 0.0239, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:50:23,028 INFO [train.py:904] (0/8) Epoch 28, batch 300, loss[loss=0.1546, simple_loss=0.2429, pruned_loss=0.03311, over 15517.00 frames. ], tot_loss[loss=0.1687, simple_loss=0.2546, pruned_loss=0.04137, over 2583354.13 frames. ], batch size: 191, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:50:29,905 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:35,979 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8378, 4.0309, 2.9452, 2.3846, 2.6619, 2.5760, 4.2527, 3.4397], device='cuda:0'), covar=tensor([0.2947, 0.0712, 0.1987, 0.3185, 0.2882, 0.2238, 0.0568, 0.1601], device='cuda:0'), in_proj_covar=tensor([0.0331, 0.0271, 0.0309, 0.0322, 0.0298, 0.0273, 0.0300, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:50:39,278 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:50:49,960 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.456e+02 2.173e+02 2.525e+02 3.031e+02 5.896e+02, threshold=5.050e+02, percent-clipped=1.0 2023-05-02 11:51:31,015 INFO [train.py:904] (0/8) Epoch 28, batch 350, loss[loss=0.161, simple_loss=0.247, pruned_loss=0.0375, over 15701.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2522, pruned_loss=0.04025, over 2744348.81 frames. ], batch size: 191, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:52:37,198 INFO [train.py:904] (0/8) Epoch 28, batch 400, loss[loss=0.1429, simple_loss=0.2338, pruned_loss=0.026, over 16859.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2511, pruned_loss=0.03959, over 2882220.19 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 11:52:47,528 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9908, 5.3802, 5.4697, 5.2804, 5.2904, 5.8845, 5.3219, 5.0547], device='cuda:0'), covar=tensor([0.1126, 0.2055, 0.2686, 0.2192, 0.2488, 0.1049, 0.1768, 0.2502], device='cuda:0'), in_proj_covar=tensor([0.0416, 0.0623, 0.0682, 0.0505, 0.0674, 0.0714, 0.0536, 0.0670], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 11:53:03,896 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.542e+02 2.265e+02 2.649e+02 3.069e+02 5.270e+02, threshold=5.298e+02, percent-clipped=3.0 2023-05-02 11:53:24,167 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:53:44,151 INFO [train.py:904] (0/8) Epoch 28, batch 450, loss[loss=0.144, simple_loss=0.2354, pruned_loss=0.02627, over 17220.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2507, pruned_loss=0.0394, over 2989607.40 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:54:02,263 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.41 vs. limit=2.0 2023-05-02 11:54:35,054 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274540.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:54:52,739 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7473, 2.7641, 2.3973, 2.6570, 3.0428, 2.8898, 3.3446, 3.3020], device='cuda:0'), covar=tensor([0.0174, 0.0514, 0.0644, 0.0530, 0.0394, 0.0504, 0.0310, 0.0348], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0246, 0.0235, 0.0235, 0.0247, 0.0245, 0.0242, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 11:54:53,422 INFO [train.py:904] (0/8) Epoch 28, batch 500, loss[loss=0.1633, simple_loss=0.2558, pruned_loss=0.03543, over 17235.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2493, pruned_loss=0.03892, over 3060112.31 frames. ], batch size: 45, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:55:01,597 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6253, 3.7692, 2.3920, 4.3639, 2.9048, 4.2912, 2.6412, 3.2420], device='cuda:0'), covar=tensor([0.0398, 0.0461, 0.1661, 0.0366, 0.0889, 0.0583, 0.1468, 0.0706], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0179, 0.0195, 0.0170, 0.0178, 0.0217, 0.0204, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 11:55:21,741 INFO [optim.py:368] (0/8) 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 11:55:58,897 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:56:00,846 INFO [train.py:904] (0/8) Epoch 28, batch 550, loss[loss=0.1769, simple_loss=0.2578, pruned_loss=0.048, over 16503.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2487, pruned_loss=0.03841, over 3111976.46 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:56:20,968 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 11:56:30,888 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274625.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:56:41,942 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7072, 3.7754, 2.4474, 4.1488, 3.0300, 4.1039, 2.5943, 3.2001], device='cuda:0'), covar=tensor([0.0318, 0.0421, 0.1546, 0.0363, 0.0813, 0.0645, 0.1434, 0.0676], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0180, 0.0197, 0.0172, 0.0179, 0.0219, 0.0206, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 11:56:56,819 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 11:57:09,888 INFO [train.py:904] (0/8) Epoch 28, batch 600, loss[loss=0.1458, simple_loss=0.2239, pruned_loss=0.03379, over 16615.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2484, pruned_loss=0.03851, over 3160416.43 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:57:15,312 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274657.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:17,489 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:57:36,637 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 28, batch 650, loss[loss=0.142, simple_loss=0.2345, pruned_loss=0.0247, over 16512.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2472, pruned_loss=0.03773, over 3205500.77 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:58:18,639 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 11:58:57,799 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8420, 2.0123, 2.3744, 2.7100, 2.7200, 2.7533, 2.0604, 2.9333], device='cuda:0'), covar=tensor([0.0222, 0.0567, 0.0397, 0.0347, 0.0377, 0.0359, 0.0618, 0.0210], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0193, 0.0208, 0.0167, 0.0206, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 11:59:25,798 INFO [train.py:904] (0/8) Epoch 28, batch 700, loss[loss=0.1583, simple_loss=0.238, pruned_loss=0.03928, over 16817.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2466, pruned_loss=0.03752, over 3238901.10 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 11:59:54,347 INFO [optim.py:368] (0/8) 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 12:00:13,321 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:00:34,714 INFO [train.py:904] (0/8) Epoch 28, batch 750, loss[loss=0.2034, simple_loss=0.2764, pruned_loss=0.06515, over 16800.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2466, pruned_loss=0.03736, over 3266722.39 frames. ], batch size: 102, lr: 2.42e-03, grad_scale: 1.0 2023-05-02 12:01:19,363 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:01:22,576 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8192, 2.8069, 2.6824, 5.0288, 3.9402, 4.3565, 1.8169, 3.1648], device='cuda:0'), covar=tensor([0.1431, 0.0856, 0.1271, 0.0173, 0.0232, 0.0407, 0.1579, 0.0827], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0179, 0.0200, 0.0200, 0.0204, 0.0218, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 12:01:45,406 INFO [train.py:904] (0/8) Epoch 28, batch 800, loss[loss=0.178, simple_loss=0.2518, pruned_loss=0.05205, over 16516.00 frames. ], tot_loss[loss=0.1605, simple_loss=0.246, pruned_loss=0.03749, over 3272309.19 frames. ], batch size: 68, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:02:15,003 INFO [optim.py:368] (0/8) 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 12:02:43,495 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=274894.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:46,256 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=274896.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:02:57,016 INFO [train.py:904] (0/8) Epoch 28, batch 850, loss[loss=0.1487, simple_loss=0.2434, pruned_loss=0.02698, over 16731.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2458, pruned_loss=0.03717, over 3289823.96 frames. ], batch size: 62, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:03:26,697 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274925.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:03:47,367 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7164, 3.7945, 2.8904, 2.3259, 2.4441, 2.4734, 3.9310, 3.3097], device='cuda:0'), covar=tensor([0.2833, 0.0600, 0.1789, 0.3235, 0.2864, 0.2230, 0.0546, 0.1596], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0275, 0.0312, 0.0325, 0.0303, 0.0276, 0.0303, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:04:05,739 INFO [train.py:904] (0/8) Epoch 28, batch 900, loss[loss=0.1637, simple_loss=0.2568, pruned_loss=0.0353, over 17032.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2448, pruned_loss=0.03628, over 3290874.32 frames. ], batch size: 55, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:04:09,242 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=274955.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:13,706 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=274959.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:19,902 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9611, 5.3876, 5.4205, 5.2456, 5.2616, 5.8951, 5.3384, 5.0874], device='cuda:0'), covar=tensor([0.1242, 0.2082, 0.2670, 0.2130, 0.2656, 0.1004, 0.1892, 0.2493], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0638, 0.0700, 0.0517, 0.0689, 0.0727, 0.0549, 0.0686], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:04:33,594 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=274973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:04:34,542 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.476e+02 2.075e+02 2.378e+02 2.870e+02 6.521e+02, threshold=4.756e+02, percent-clipped=4.0 2023-05-02 12:05:14,553 INFO [train.py:904] (0/8) Epoch 28, batch 950, loss[loss=0.1624, simple_loss=0.2574, pruned_loss=0.03374, over 17264.00 frames. ], tot_loss[loss=0.1599, simple_loss=0.2459, pruned_loss=0.03693, over 3293878.86 frames. ], batch size: 52, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:05:20,459 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275007.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:05:56,265 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2255, 4.3000, 4.5878, 4.5949, 4.6193, 4.3536, 4.3540, 4.2425], device='cuda:0'), covar=tensor([0.0464, 0.0870, 0.0468, 0.0460, 0.0504, 0.0529, 0.0845, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0493, 0.0477, 0.0438, 0.0524, 0.0502, 0.0575, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 12:06:23,169 INFO [train.py:904] (0/8) Epoch 28, batch 1000, loss[loss=0.1619, simple_loss=0.2333, pruned_loss=0.04524, over 16829.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2451, pruned_loss=0.03701, over 3301117.88 frames. ], batch size: 116, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:06:52,348 INFO [optim.py:368] (0/8) 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 12:07:31,789 INFO [train.py:904] (0/8) Epoch 28, batch 1050, loss[loss=0.1722, simple_loss=0.2462, pruned_loss=0.04912, over 16380.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2453, pruned_loss=0.03696, over 3299846.64 frames. ], batch size: 146, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:07:50,724 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5125, 3.5476, 2.2439, 3.7694, 2.8209, 3.7345, 2.4194, 2.9280], device='cuda:0'), covar=tensor([0.0307, 0.0436, 0.1624, 0.0379, 0.0804, 0.0710, 0.1357, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0181, 0.0197, 0.0173, 0.0180, 0.0220, 0.0206, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 12:08:39,636 INFO [train.py:904] (0/8) Epoch 28, batch 1100, loss[loss=0.1459, simple_loss=0.2257, pruned_loss=0.03307, over 16909.00 frames. ], tot_loss[loss=0.1589, simple_loss=0.2447, pruned_loss=0.03656, over 3314393.06 frames. ], batch size: 116, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:08:43,090 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2102, 3.2085, 3.5609, 2.2755, 2.9574, 2.3784, 3.6776, 3.5229], device='cuda:0'), covar=tensor([0.0241, 0.1052, 0.0589, 0.2074, 0.0911, 0.1040, 0.0540, 0.1051], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0170, 0.0170, 0.0157, 0.0148, 0.0133, 0.0146, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 12:08:47,425 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275158.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:07,972 INFO [optim.py:368] (0/8) 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 12:09:38,973 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:09:49,917 INFO [train.py:904] (0/8) Epoch 28, batch 1150, loss[loss=0.1532, simple_loss=0.2414, pruned_loss=0.03251, over 16759.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2446, pruned_loss=0.03623, over 3318436.58 frames. ], batch size: 39, lr: 2.42e-03, grad_scale: 2.0 2023-05-02 12:09:52,241 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 12:10:10,261 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1388, 5.1321, 4.8609, 4.3512, 4.9868, 1.7809, 4.7159, 4.6478], device='cuda:0'), covar=tensor([0.0115, 0.0099, 0.0240, 0.0426, 0.0115, 0.3144, 0.0163, 0.0282], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0174, 0.0210, 0.0182, 0.0187, 0.0217, 0.0200, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:10:11,435 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:30,511 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7014, 2.6928, 2.3752, 2.6253, 3.0299, 2.7674, 3.1665, 3.1759], device='cuda:0'), covar=tensor([0.0236, 0.0591, 0.0671, 0.0568, 0.0388, 0.0514, 0.0451, 0.0441], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0250, 0.0238, 0.0238, 0.0251, 0.0248, 0.0247, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:10:44,956 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:53,461 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275250.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:10:57,370 INFO [train.py:904] (0/8) Epoch 28, batch 1200, loss[loss=0.1517, simple_loss=0.2417, pruned_loss=0.0308, over 16850.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2439, pruned_loss=0.03606, over 3308816.06 frames. ], batch size: 42, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:11:20,281 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7204, 3.9261, 4.0490, 2.9930, 3.6331, 4.1178, 3.7781, 2.3877], device='cuda:0'), covar=tensor([0.0527, 0.0362, 0.0083, 0.0386, 0.0146, 0.0147, 0.0122, 0.0556], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0090, 0.0091, 0.0135, 0.0102, 0.0114, 0.0098, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 12:11:21,750 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.26 vs. limit=2.0 2023-05-02 12:11:26,949 INFO [optim.py:368] (0/8) 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 12:12:06,869 INFO [train.py:904] (0/8) Epoch 28, batch 1250, loss[loss=0.1904, simple_loss=0.262, pruned_loss=0.05938, over 16838.00 frames. ], tot_loss[loss=0.158, simple_loss=0.2439, pruned_loss=0.03603, over 3309909.52 frames. ], batch size: 96, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:12:12,683 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275307.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:08,433 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275347.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:13:17,203 INFO [train.py:904] (0/8) Epoch 28, batch 1300, loss[loss=0.1554, simple_loss=0.2377, pruned_loss=0.03657, over 16727.00 frames. ], tot_loss[loss=0.1577, simple_loss=0.2432, pruned_loss=0.03607, over 3312035.33 frames. ], batch size: 124, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:13:39,186 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275368.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:13:46,873 INFO [optim.py:368] (0/8) 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 12:14:02,333 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275385.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:14:26,887 INFO [train.py:904] (0/8) Epoch 28, batch 1350, loss[loss=0.1403, simple_loss=0.2208, pruned_loss=0.02986, over 16779.00 frames. ], tot_loss[loss=0.1576, simple_loss=0.2439, pruned_loss=0.03564, over 3309207.67 frames. ], batch size: 39, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:14:35,342 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275408.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 12:15:28,628 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:15:37,729 INFO [train.py:904] (0/8) Epoch 28, batch 1400, loss[loss=0.1754, simple_loss=0.2654, pruned_loss=0.04272, over 17046.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2438, pruned_loss=0.03546, over 3309332.91 frames. ], batch size: 53, lr: 2.42e-03, grad_scale: 4.0 2023-05-02 12:16:06,679 INFO [optim.py:368] (0/8) 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 12:16:46,826 INFO [train.py:904] (0/8) Epoch 28, batch 1450, loss[loss=0.1441, simple_loss=0.2262, pruned_loss=0.03094, over 15943.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2432, pruned_loss=0.03574, over 3309808.46 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:17:03,277 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:05,694 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9597, 2.2126, 2.5363, 2.8862, 2.8091, 3.2305, 2.3184, 3.2486], device='cuda:0'), covar=tensor([0.0285, 0.0539, 0.0410, 0.0378, 0.0435, 0.0266, 0.0567, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0200, 0.0189, 0.0194, 0.0209, 0.0168, 0.0205, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 12:17:53,384 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:17:57,227 INFO [train.py:904] (0/8) Epoch 28, batch 1500, loss[loss=0.1836, simple_loss=0.2587, pruned_loss=0.0542, over 16823.00 frames. ], tot_loss[loss=0.1582, simple_loss=0.2437, pruned_loss=0.03638, over 3315782.65 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:18:26,335 INFO [optim.py:368] (0/8) 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 12:19:01,052 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275598.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:19:07,375 INFO [train.py:904] (0/8) Epoch 28, batch 1550, loss[loss=0.1477, simple_loss=0.2467, pruned_loss=0.02438, over 17029.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2447, pruned_loss=0.03721, over 3315126.48 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:19:18,153 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 12:19:31,502 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9455, 3.6710, 4.2069, 2.1671, 4.3499, 4.4352, 3.2744, 3.3497], device='cuda:0'), covar=tensor([0.0723, 0.0306, 0.0246, 0.1197, 0.0090, 0.0208, 0.0469, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0111, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 12:20:18,424 INFO [train.py:904] (0/8) Epoch 28, batch 1600, loss[loss=0.1639, simple_loss=0.2607, pruned_loss=0.03353, over 17136.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2462, pruned_loss=0.03834, over 3309497.52 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:20:33,172 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275663.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:20:48,277 INFO [optim.py:368] (0/8) 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 12:21:10,194 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9552, 4.4900, 4.4770, 3.2258, 3.6471, 4.4634, 4.0264, 2.6941], device='cuda:0'), covar=tensor([0.0525, 0.0071, 0.0052, 0.0384, 0.0169, 0.0108, 0.0094, 0.0520], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0136, 0.0103, 0.0115, 0.0099, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 12:21:28,987 INFO [train.py:904] (0/8) Epoch 28, batch 1650, loss[loss=0.1762, simple_loss=0.2498, pruned_loss=0.05125, over 16707.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2484, pruned_loss=0.03915, over 3288559.85 frames. ], batch size: 124, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:21:29,296 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=275703.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:22:21,196 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 28, batch 1700, loss[loss=0.1779, simple_loss=0.2582, pruned_loss=0.04883, over 16758.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.2505, pruned_loss=0.03992, over 3293946.00 frames. ], batch size: 134, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:23:08,713 INFO [optim.py:368] (0/8) 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] (0/8) attn_weights_entropy = tensor([4.9865, 4.7399, 5.0214, 5.2124, 5.4100, 4.7748, 5.3991, 5.4291], device='cuda:0'), covar=tensor([0.2061, 0.1423, 0.1798, 0.0799, 0.0584, 0.1004, 0.0651, 0.0682], device='cuda:0'), in_proj_covar=tensor([0.0692, 0.0842, 0.0978, 0.0855, 0.0652, 0.0683, 0.0718, 0.0831], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:23:49,164 INFO [train.py:904] (0/8) Epoch 28, batch 1750, loss[loss=0.1733, simple_loss=0.2598, pruned_loss=0.04336, over 16480.00 frames. ], tot_loss[loss=0.165, simple_loss=0.251, pruned_loss=0.03949, over 3307787.46 frames. ], batch size: 68, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:24:03,927 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4108, 5.3271, 5.2669, 4.7222, 4.8778, 5.2633, 5.2687, 4.8419], device='cuda:0'), covar=tensor([0.0603, 0.0609, 0.0334, 0.0391, 0.1156, 0.0542, 0.0315, 0.0963], device='cuda:0'), in_proj_covar=tensor([0.0319, 0.0479, 0.0373, 0.0376, 0.0370, 0.0430, 0.0256, 0.0447], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:24:05,559 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275814.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:24:19,764 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.66 vs. limit=2.0 2023-05-02 12:24:59,139 INFO [train.py:904] (0/8) Epoch 28, batch 1800, loss[loss=0.1491, simple_loss=0.2356, pruned_loss=0.03129, over 16809.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2523, pruned_loss=0.03962, over 3301243.47 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:25:09,009 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275860.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:11,182 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=275862.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:25:13,624 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.85 vs. limit=5.0 2023-05-02 12:25:30,538 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.248e+02 2.316e+02 2.618e+02 3.157e+02 1.143e+03, threshold=5.235e+02, percent-clipped=6.0 2023-05-02 12:26:07,535 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7039, 4.6770, 4.6032, 4.0735, 4.6695, 1.7988, 4.4373, 4.2750], device='cuda:0'), covar=tensor([0.0167, 0.0139, 0.0202, 0.0353, 0.0109, 0.3104, 0.0177, 0.0268], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0177, 0.0214, 0.0186, 0.0190, 0.0220, 0.0203, 0.0181], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:26:08,321 INFO [train.py:904] (0/8) Epoch 28, batch 1850, loss[loss=0.1498, simple_loss=0.2351, pruned_loss=0.03218, over 16830.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2538, pruned_loss=0.04, over 3303888.44 frames. ], batch size: 90, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:26:20,883 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8847, 5.0510, 5.3925, 5.3332, 5.4063, 5.1124, 4.8867, 4.8475], device='cuda:0'), covar=tensor([0.0538, 0.0706, 0.0527, 0.0694, 0.0656, 0.0634, 0.1383, 0.0593], device='cuda:0'), in_proj_covar=tensor([0.0441, 0.0494, 0.0477, 0.0441, 0.0525, 0.0505, 0.0580, 0.0403], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 12:26:33,131 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=275921.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:26:48,397 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4669, 3.1414, 3.5879, 1.9095, 3.6521, 3.6683, 3.0779, 2.7248], device='cuda:0'), covar=tensor([0.0850, 0.0320, 0.0193, 0.1234, 0.0121, 0.0243, 0.0418, 0.0508], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0111, 0.0102, 0.0139, 0.0087, 0.0132, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 12:27:05,287 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4667, 5.8689, 5.6224, 5.7215, 5.2991, 5.2928, 5.2895, 5.9975], device='cuda:0'), covar=tensor([0.1490, 0.0927, 0.1054, 0.0891, 0.0980, 0.0792, 0.1298, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0724, 0.0880, 0.0716, 0.0678, 0.0556, 0.0555, 0.0740, 0.0686], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:27:16,685 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=275952.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:17,452 INFO [train.py:904] (0/8) Epoch 28, batch 1900, loss[loss=0.1347, simple_loss=0.2166, pruned_loss=0.02637, over 17027.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2537, pruned_loss=0.03942, over 3304464.78 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:27:22,918 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-02 12:27:31,503 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=275963.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:27:39,241 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8710, 3.6262, 4.1060, 2.1507, 4.2265, 4.3156, 3.3007, 3.3010], device='cuda:0'), covar=tensor([0.0799, 0.0324, 0.0248, 0.1251, 0.0108, 0.0272, 0.0446, 0.0476], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0139, 0.0086, 0.0131, 0.0129, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 12:27:43,729 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0915, 4.8406, 5.1263, 5.3064, 5.5106, 4.8373, 5.4850, 5.5001], device='cuda:0'), covar=tensor([0.2042, 0.1336, 0.1822, 0.0805, 0.0554, 0.1004, 0.0544, 0.0633], device='cuda:0'), in_proj_covar=tensor([0.0693, 0.0843, 0.0981, 0.0858, 0.0652, 0.0685, 0.0719, 0.0832], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:27:47,691 INFO [optim.py:368] (0/8) 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 12:28:22,514 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-276000.pt 2023-05-02 12:28:30,507 INFO [train.py:904] (0/8) Epoch 28, batch 1950, loss[loss=0.1835, simple_loss=0.2589, pruned_loss=0.054, over 16854.00 frames. ], tot_loss[loss=0.1649, simple_loss=0.2525, pruned_loss=0.03869, over 3301273.63 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:28:30,831 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276003.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 12:28:40,348 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:28:43,422 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276013.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:24,696 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:29:37,517 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276051.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:29:40,570 INFO [train.py:904] (0/8) Epoch 28, batch 2000, loss[loss=0.154, simple_loss=0.2572, pruned_loss=0.02541, over 17131.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2519, pruned_loss=0.0383, over 3311162.79 frames. ], batch size: 49, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:11,367 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.574e+02 2.104e+02 2.511e+02 2.919e+02 7.567e+02, threshold=5.023e+02, percent-clipped=3.0 2023-05-02 12:30:31,278 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276089.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:30:50,235 INFO [train.py:904] (0/8) Epoch 28, batch 2050, loss[loss=0.1657, simple_loss=0.2624, pruned_loss=0.0345, over 17260.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2514, pruned_loss=0.03842, over 3313063.69 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:30:54,889 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 12:31:09,003 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 12:31:21,932 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0991, 5.4172, 5.2124, 5.2684, 4.9856, 4.8796, 4.9250, 5.5416], device='cuda:0'), covar=tensor([0.1359, 0.0963, 0.1101, 0.0900, 0.0842, 0.1044, 0.1182, 0.0888], device='cuda:0'), in_proj_covar=tensor([0.0727, 0.0889, 0.0723, 0.0683, 0.0560, 0.0557, 0.0745, 0.0691], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:31:26,607 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2235, 2.2581, 2.3223, 3.9434, 2.3277, 2.5963, 2.3240, 2.4042], device='cuda:0'), covar=tensor([0.1566, 0.3864, 0.3320, 0.0684, 0.3963, 0.2739, 0.4048, 0.3273], device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0475, 0.0388, 0.0338, 0.0447, 0.0545, 0.0447, 0.0556], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:31:31,830 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8138, 3.9610, 4.0757, 2.9703, 3.6255, 4.1165, 3.7728, 2.6493], device='cuda:0'), covar=tensor([0.0468, 0.0401, 0.0069, 0.0385, 0.0148, 0.0121, 0.0123, 0.0451], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0092, 0.0092, 0.0137, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 12:32:00,358 INFO [train.py:904] (0/8) Epoch 28, batch 2100, loss[loss=0.1735, simple_loss=0.2672, pruned_loss=0.03983, over 17074.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2516, pruned_loss=0.03848, over 3326456.39 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:32:15,341 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3676, 3.4990, 3.6333, 3.6080, 3.6291, 3.4909, 3.4816, 3.5132], device='cuda:0'), covar=tensor([0.0479, 0.0686, 0.0536, 0.0551, 0.0640, 0.0590, 0.0843, 0.0590], device='cuda:0'), in_proj_covar=tensor([0.0443, 0.0498, 0.0481, 0.0444, 0.0529, 0.0508, 0.0583, 0.0406], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 12:32:30,700 INFO [optim.py:368] (0/8) 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 12:33:09,368 INFO [train.py:904] (0/8) Epoch 28, batch 2150, loss[loss=0.1808, simple_loss=0.2698, pruned_loss=0.04593, over 16663.00 frames. ], tot_loss[loss=0.1648, simple_loss=0.2528, pruned_loss=0.03837, over 3318473.78 frames. ], batch size: 62, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:33:27,334 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276216.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:34:18,063 INFO [train.py:904] (0/8) Epoch 28, batch 2200, loss[loss=0.1728, simple_loss=0.2693, pruned_loss=0.03813, over 16724.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2534, pruned_loss=0.0393, over 3318188.61 frames. ], batch size: 57, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:34:23,933 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6952, 6.0234, 5.7745, 5.8412, 5.3795, 5.4132, 5.3867, 6.1716], device='cuda:0'), covar=tensor([0.1358, 0.0925, 0.1143, 0.0980, 0.0990, 0.0724, 0.1421, 0.0890], device='cuda:0'), in_proj_covar=tensor([0.0728, 0.0889, 0.0721, 0.0683, 0.0559, 0.0558, 0.0745, 0.0691], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:34:50,534 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.368e+02 2.258e+02 2.667e+02 3.359e+02 8.724e+02, threshold=5.333e+02, percent-clipped=6.0 2023-05-02 12:35:27,994 INFO [train.py:904] (0/8) Epoch 28, batch 2250, loss[loss=0.1538, simple_loss=0.2531, pruned_loss=0.02729, over 17041.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2532, pruned_loss=0.03891, over 3321222.62 frames. ], batch size: 50, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:35:35,391 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276308.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:35:35,448 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9194, 4.7557, 4.9998, 5.2145, 5.3802, 4.7838, 5.3831, 5.4084], device='cuda:0'), covar=tensor([0.2198, 0.1369, 0.1756, 0.0751, 0.0541, 0.1002, 0.0615, 0.0702], device='cuda:0'), in_proj_covar=tensor([0.0700, 0.0852, 0.0990, 0.0867, 0.0659, 0.0691, 0.0725, 0.0840], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:35:47,431 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:35:58,325 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 12:36:32,852 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8756, 4.8996, 5.2982, 5.2602, 5.2983, 4.9671, 4.9003, 4.7462], device='cuda:0'), covar=tensor([0.0377, 0.0634, 0.0402, 0.0443, 0.0522, 0.0408, 0.1022, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0444, 0.0501, 0.0483, 0.0446, 0.0531, 0.0510, 0.0586, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 12:36:37,084 INFO [train.py:904] (0/8) Epoch 28, batch 2300, loss[loss=0.1468, simple_loss=0.242, pruned_loss=0.02584, over 17208.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2531, pruned_loss=0.03883, over 3324248.63 frames. ], batch size: 44, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:36:52,469 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4387, 5.3800, 5.2796, 4.7306, 4.9349, 5.3383, 5.2450, 4.8936], device='cuda:0'), covar=tensor([0.0597, 0.0519, 0.0318, 0.0387, 0.1075, 0.0472, 0.0314, 0.0892], device='cuda:0'), in_proj_covar=tensor([0.0322, 0.0482, 0.0376, 0.0378, 0.0371, 0.0433, 0.0257, 0.0450], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:37:08,709 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.645e+02 2.043e+02 2.409e+02 2.878e+02 4.638e+02, threshold=4.818e+02, percent-clipped=0.0 2023-05-02 12:37:11,205 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1936, 2.6684, 2.1265, 2.4457, 2.9655, 2.7803, 3.0584, 3.1290], device='cuda:0'), covar=tensor([0.0223, 0.0431, 0.0616, 0.0494, 0.0339, 0.0387, 0.0282, 0.0318], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0253, 0.0240, 0.0241, 0.0254, 0.0252, 0.0251, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:37:12,213 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276378.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:37:16,727 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 12:37:46,507 INFO [train.py:904] (0/8) Epoch 28, batch 2350, loss[loss=0.165, simple_loss=0.269, pruned_loss=0.0305, over 17272.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.254, pruned_loss=0.03978, over 3327544.37 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 4.0 2023-05-02 12:38:10,345 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3703, 3.6049, 3.8973, 2.2934, 3.1639, 2.4912, 3.7739, 3.7858], device='cuda:0'), covar=tensor([0.0319, 0.0943, 0.0528, 0.2047, 0.0868, 0.0981, 0.0640, 0.1059], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0170, 0.0157, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 12:38:54,376 INFO [train.py:904] (0/8) Epoch 28, batch 2400, loss[loss=0.183, simple_loss=0.2794, pruned_loss=0.0433, over 16049.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2549, pruned_loss=0.03962, over 3323420.00 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:39:26,391 INFO [optim.py:368] (0/8) 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 12:39:39,823 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5096, 3.7717, 3.9246, 2.7670, 3.5488, 4.0315, 3.6805, 2.4175], device='cuda:0'), covar=tensor([0.0552, 0.0255, 0.0071, 0.0405, 0.0152, 0.0096, 0.0123, 0.0516], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0116, 0.0100, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 12:40:04,325 INFO [train.py:904] (0/8) Epoch 28, batch 2450, loss[loss=0.1769, simple_loss=0.272, pruned_loss=0.0409, over 17122.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2547, pruned_loss=0.03932, over 3331243.86 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:40:23,638 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276516.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:40:23,812 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1851, 2.6337, 2.1468, 2.4140, 2.9787, 2.7117, 2.9587, 3.1089], device='cuda:0'), covar=tensor([0.0228, 0.0491, 0.0610, 0.0526, 0.0279, 0.0399, 0.0269, 0.0312], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0253, 0.0241, 0.0242, 0.0254, 0.0252, 0.0251, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:40:38,046 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3974, 3.8909, 4.4363, 2.2610, 4.5838, 4.7115, 3.5065, 3.7180], device='cuda:0'), covar=tensor([0.0581, 0.0291, 0.0219, 0.1183, 0.0080, 0.0176, 0.0422, 0.0375], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0138, 0.0086, 0.0131, 0.0129, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 12:41:13,986 INFO [train.py:904] (0/8) Epoch 28, batch 2500, loss[loss=0.1576, simple_loss=0.2532, pruned_loss=0.03102, over 16851.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2551, pruned_loss=0.03909, over 3325370.04 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:41:23,393 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4875, 3.3214, 2.6930, 2.1630, 2.2252, 2.3180, 3.4325, 2.9524], device='cuda:0'), covar=tensor([0.2900, 0.0657, 0.1847, 0.2910, 0.2753, 0.2282, 0.0583, 0.1592], device='cuda:0'), in_proj_covar=tensor([0.0338, 0.0280, 0.0317, 0.0331, 0.0309, 0.0281, 0.0309, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:41:30,323 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:41:45,571 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.411e+02 2.004e+02 2.344e+02 2.885e+02 4.375e+02, threshold=4.688e+02, percent-clipped=0.0 2023-05-02 12:42:23,399 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9991, 2.0737, 2.6497, 3.0055, 2.8059, 3.4012, 2.3328, 3.4818], device='cuda:0'), covar=tensor([0.0352, 0.0640, 0.0392, 0.0408, 0.0425, 0.0261, 0.0624, 0.0202], device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0201, 0.0189, 0.0196, 0.0210, 0.0169, 0.0205, 0.0168], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 12:42:24,144 INFO [train.py:904] (0/8) Epoch 28, batch 2550, loss[loss=0.1485, simple_loss=0.2526, pruned_loss=0.02219, over 17107.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2554, pruned_loss=0.03939, over 3327486.85 frames. ], batch size: 48, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:42:31,835 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:42:35,367 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276611.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:28,819 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1423, 5.6943, 5.8212, 5.4797, 5.5705, 6.1429, 5.5933, 5.3103], device='cuda:0'), covar=tensor([0.1014, 0.1981, 0.2714, 0.2173, 0.2687, 0.0984, 0.1565, 0.2358], device='cuda:0'), in_proj_covar=tensor([0.0437, 0.0654, 0.0718, 0.0532, 0.0707, 0.0745, 0.0559, 0.0708], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:43:33,173 INFO [train.py:904] (0/8) Epoch 28, batch 2600, loss[loss=0.1599, simple_loss=0.2468, pruned_loss=0.03647, over 16816.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2558, pruned_loss=0.03947, over 3324120.51 frames. ], batch size: 96, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:43:37,974 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=276656.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:43:55,761 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276668.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:01,044 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276672.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:01,993 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276673.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:44:05,674 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.592e+02 2.267e+02 2.519e+02 3.133e+02 7.416e+02, threshold=5.037e+02, percent-clipped=2.0 2023-05-02 12:44:41,541 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 12:44:43,697 INFO [train.py:904] (0/8) Epoch 28, batch 2650, loss[loss=0.1771, simple_loss=0.2715, pruned_loss=0.0413, over 17043.00 frames. ], tot_loss[loss=0.168, simple_loss=0.2571, pruned_loss=0.03942, over 3330380.40 frames. ], batch size: 55, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:45:19,138 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:45:50,749 INFO [train.py:904] (0/8) Epoch 28, batch 2700, loss[loss=0.1706, simple_loss=0.2605, pruned_loss=0.04039, over 16844.00 frames. ], tot_loss[loss=0.1682, simple_loss=0.2576, pruned_loss=0.03937, over 3328051.29 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:46:23,526 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.276e+02 2.127e+02 2.476e+02 3.244e+02 1.015e+03, threshold=4.951e+02, percent-clipped=5.0 2023-05-02 12:46:42,603 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:46:56,315 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8396, 2.5638, 2.5132, 3.8175, 3.0734, 3.8803, 1.5925, 2.8642], device='cuda:0'), covar=tensor([0.1401, 0.0761, 0.1209, 0.0200, 0.0155, 0.0373, 0.1668, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0205, 0.0207, 0.0220, 0.0210, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 12:47:00,589 INFO [train.py:904] (0/8) Epoch 28, batch 2750, loss[loss=0.16, simple_loss=0.2516, pruned_loss=0.03416, over 17229.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2572, pruned_loss=0.03838, over 3337858.41 frames. ], batch size: 45, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:06,666 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276851.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:09,126 INFO [train.py:904] (0/8) Epoch 28, batch 2800, loss[loss=0.1577, simple_loss=0.2446, pruned_loss=0.03542, over 16770.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2566, pruned_loss=0.03856, over 3333320.49 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:48:24,989 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276863.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:48:25,226 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 12:48:41,264 INFO [optim.py:368] (0/8) 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 12:49:20,036 INFO [train.py:904] (0/8) Epoch 28, batch 2850, loss[loss=0.1571, simple_loss=0.244, pruned_loss=0.03507, over 16051.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2561, pruned_loss=0.03843, over 3328436.16 frames. ], batch size: 35, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:49:50,126 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276924.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:49:57,703 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=276931.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:03,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6512, 2.6445, 2.1988, 2.4001, 2.9441, 2.6868, 3.1888, 3.1812], device='cuda:0'), covar=tensor([0.0203, 0.0544, 0.0665, 0.0575, 0.0387, 0.0481, 0.0338, 0.0341], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0253, 0.0240, 0.0241, 0.0254, 0.0251, 0.0252, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:50:27,302 INFO [train.py:904] (0/8) Epoch 28, batch 2900, loss[loss=0.1847, simple_loss=0.2589, pruned_loss=0.05526, over 16285.00 frames. ], tot_loss[loss=0.1664, simple_loss=0.2548, pruned_loss=0.03906, over 3319023.77 frames. ], batch size: 165, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:50:46,680 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=276967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:54,997 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=276973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:50:58,050 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.403e+02 2.124e+02 2.391e+02 2.943e+02 6.733e+02, threshold=4.783e+02, percent-clipped=5.0 2023-05-02 12:51:22,032 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=276992.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 12:51:27,115 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7116, 4.7978, 4.9924, 4.8031, 4.8344, 5.4105, 4.9505, 4.6863], device='cuda:0'), covar=tensor([0.1560, 0.2107, 0.2468, 0.2227, 0.2750, 0.1111, 0.1738, 0.2731], device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0655, 0.0718, 0.0532, 0.0706, 0.0742, 0.0558, 0.0708], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:51:36,329 INFO [train.py:904] (0/8) Epoch 28, batch 2950, loss[loss=0.1818, simple_loss=0.2799, pruned_loss=0.04185, over 17145.00 frames. ], tot_loss[loss=0.1669, simple_loss=0.2548, pruned_loss=0.03948, over 3321459.59 frames. ], batch size: 49, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:51:48,544 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277011.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:00,583 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277021.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:05,403 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277024.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:52:45,420 INFO [train.py:904] (0/8) Epoch 28, batch 3000, loss[loss=0.1603, simple_loss=0.2612, pruned_loss=0.02971, over 17258.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2539, pruned_loss=0.03953, over 3324983.88 frames. ], batch size: 52, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:52:45,421 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 12:52:54,790 INFO [train.py:938] (0/8) Epoch 28, validation: loss=0.1335, simple_loss=0.2385, pruned_loss=0.01427, over 944034.00 frames. 2023-05-02 12:52:54,791 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 12:53:21,132 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277072.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:53:25,663 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.470e+02 2.148e+02 2.585e+02 3.132e+02 1.139e+03, threshold=5.170e+02, percent-clipped=2.0 2023-05-02 12:53:27,159 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1969, 5.5065, 5.2735, 5.2916, 5.0239, 4.9393, 4.9045, 5.6073], device='cuda:0'), covar=tensor([0.1265, 0.0933, 0.1046, 0.0934, 0.0793, 0.0897, 0.1258, 0.0889], device='cuda:0'), in_proj_covar=tensor([0.0737, 0.0899, 0.0734, 0.0692, 0.0567, 0.0562, 0.0754, 0.0700], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:53:55,165 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2068, 2.0188, 2.7712, 3.1750, 2.9792, 3.6328, 2.1743, 3.6563], device='cuda:0'), covar=tensor([0.0267, 0.0725, 0.0416, 0.0350, 0.0379, 0.0234, 0.0794, 0.0185], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0197, 0.0212, 0.0170, 0.0207, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 12:54:02,154 INFO [train.py:904] (0/8) Epoch 28, batch 3050, loss[loss=0.1474, simple_loss=0.2328, pruned_loss=0.03099, over 16805.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2534, pruned_loss=0.03923, over 3328811.50 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:54:13,061 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2314, 4.0341, 4.2892, 4.3959, 4.4643, 4.0745, 4.2529, 4.4772], device='cuda:0'), covar=tensor([0.1562, 0.1187, 0.1183, 0.0715, 0.0634, 0.1357, 0.2659, 0.0786], device='cuda:0'), in_proj_covar=tensor([0.0698, 0.0851, 0.0989, 0.0867, 0.0658, 0.0691, 0.0725, 0.0837], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:55:01,332 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277146.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:55:11,790 INFO [train.py:904] (0/8) Epoch 28, batch 3100, loss[loss=0.1593, simple_loss=0.2459, pruned_loss=0.03638, over 15512.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2528, pruned_loss=0.0392, over 3324434.34 frames. ], batch size: 191, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:55:43,285 INFO [optim.py:368] (0/8) 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 12:55:45,338 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277177.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:05,032 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3962, 4.3643, 4.5346, 4.2617, 4.3250, 4.9884, 4.4506, 4.0849], device='cuda:0'), covar=tensor([0.1702, 0.2247, 0.2367, 0.2484, 0.3027, 0.1214, 0.2046, 0.3076], device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0655, 0.0720, 0.0533, 0.0707, 0.0743, 0.0558, 0.0709], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:56:21,093 INFO [train.py:904] (0/8) Epoch 28, batch 3150, loss[loss=0.1466, simple_loss=0.2347, pruned_loss=0.02925, over 16202.00 frames. ], tot_loss[loss=0.1644, simple_loss=0.2514, pruned_loss=0.03872, over 3325179.94 frames. ], batch size: 36, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:56:44,550 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:44,665 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277219.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:56:50,317 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8791, 3.9410, 3.0371, 2.3526, 2.5365, 2.5751, 4.0908, 3.3840], device='cuda:0'), covar=tensor([0.2647, 0.0577, 0.1791, 0.3219, 0.2929, 0.2158, 0.0524, 0.1587], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0279, 0.0315, 0.0330, 0.0308, 0.0280, 0.0308, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:57:10,227 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277238.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:57:15,083 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4235, 5.8402, 5.5831, 5.6584, 5.2840, 5.3349, 5.2024, 5.9640], device='cuda:0'), covar=tensor([0.1609, 0.1016, 0.1105, 0.0939, 0.0951, 0.0703, 0.1352, 0.1014], device='cuda:0'), in_proj_covar=tensor([0.0739, 0.0902, 0.0735, 0.0693, 0.0568, 0.0563, 0.0755, 0.0701], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:57:30,738 INFO [train.py:904] (0/8) Epoch 28, batch 3200, loss[loss=0.1495, simple_loss=0.2359, pruned_loss=0.03151, over 17043.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2512, pruned_loss=0.03855, over 3325873.49 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:57:49,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8295, 3.9868, 2.7531, 2.3852, 2.4664, 2.3530, 4.2322, 3.3189], device='cuda:0'), covar=tensor([0.3117, 0.0877, 0.2382, 0.3169, 0.3376, 0.2720, 0.0569, 0.1697], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0279, 0.0315, 0.0330, 0.0309, 0.0280, 0.0308, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 12:57:50,700 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277267.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:01,818 INFO [optim.py:368] (0/8) 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 12:58:07,062 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277280.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:17,854 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277287.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 12:58:29,280 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4487, 2.6341, 2.1922, 2.3939, 2.9103, 2.6334, 2.9913, 3.1033], device='cuda:0'), covar=tensor([0.0245, 0.0550, 0.0672, 0.0556, 0.0366, 0.0464, 0.0327, 0.0354], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0253, 0.0240, 0.0241, 0.0255, 0.0251, 0.0252, 0.0251], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 12:58:39,071 INFO [train.py:904] (0/8) Epoch 28, batch 3250, loss[loss=0.1821, simple_loss=0.2622, pruned_loss=0.05099, over 16427.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2513, pruned_loss=0.03885, over 3327537.53 frames. ], batch size: 146, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 12:58:42,430 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277305.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:58:55,902 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:09,085 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277324.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 12:59:47,704 INFO [train.py:904] (0/8) Epoch 28, batch 3300, loss[loss=0.1713, simple_loss=0.2616, pruned_loss=0.04048, over 17030.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2527, pruned_loss=0.03894, over 3329012.92 frames. ], batch size: 53, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:00:05,234 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277366.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:06,241 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277367.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:12,573 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277372.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:00:18,090 INFO [optim.py:368] (0/8) 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 13:00:46,829 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4901, 3.6534, 3.9847, 2.2460, 3.2025, 2.4821, 3.9062, 3.8369], device='cuda:0'), covar=tensor([0.0265, 0.0926, 0.0497, 0.2061, 0.0832, 0.1012, 0.0607, 0.1028], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0174, 0.0171, 0.0158, 0.0150, 0.0134, 0.0148, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:00:55,416 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 13:00:55,921 INFO [train.py:904] (0/8) Epoch 28, batch 3350, loss[loss=0.1297, simple_loss=0.2165, pruned_loss=0.02138, over 17041.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2538, pruned_loss=0.03894, over 3320148.59 frames. ], batch size: 41, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:01:25,841 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9903, 4.0696, 4.3150, 4.2791, 4.3275, 4.0783, 4.1182, 4.0343], device='cuda:0'), covar=tensor([0.0391, 0.0644, 0.0435, 0.0455, 0.0557, 0.0463, 0.0758, 0.0594], device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0509, 0.0489, 0.0452, 0.0538, 0.0516, 0.0596, 0.0414], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 13:01:25,999 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6134, 2.9133, 3.0769, 2.0192, 2.7479, 2.1436, 3.2640, 3.2823], device='cuda:0'), covar=tensor([0.0255, 0.0940, 0.0653, 0.2086, 0.0930, 0.1074, 0.0604, 0.0856], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0174, 0.0171, 0.0158, 0.0150, 0.0134, 0.0148, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:01:55,025 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277446.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:02:03,954 INFO [train.py:904] (0/8) Epoch 28, batch 3400, loss[loss=0.1657, simple_loss=0.2434, pruned_loss=0.04405, over 16866.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2533, pruned_loss=0.03869, over 3324563.42 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:02:23,267 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1954, 5.6277, 5.7396, 5.3715, 5.4976, 6.0888, 5.5888, 5.3528], device='cuda:0'), covar=tensor([0.1016, 0.2103, 0.2434, 0.2386, 0.2955, 0.1088, 0.1675, 0.2539], device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0654, 0.0718, 0.0533, 0.0708, 0.0742, 0.0556, 0.0707], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:02:34,599 INFO [optim.py:368] (0/8) 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 13:03:01,408 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277494.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:13,672 INFO [train.py:904] (0/8) Epoch 28, batch 3450, loss[loss=0.1567, simple_loss=0.229, pruned_loss=0.04221, over 16847.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2519, pruned_loss=0.0383, over 3323790.57 frames. ], batch size: 102, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:03:14,051 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3706, 4.2794, 4.2900, 4.0127, 4.1012, 4.3595, 3.9770, 4.1351], device='cuda:0'), covar=tensor([0.0614, 0.0837, 0.0338, 0.0312, 0.0692, 0.0517, 0.0748, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0490, 0.0382, 0.0384, 0.0378, 0.0440, 0.0260, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:03:35,940 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277519.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:54,331 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6857, 6.0716, 5.7679, 5.8862, 5.4936, 5.4647, 5.4133, 6.1665], device='cuda:0'), covar=tensor([0.1422, 0.0953, 0.0983, 0.0911, 0.0935, 0.0665, 0.1333, 0.0887], device='cuda:0'), in_proj_covar=tensor([0.0741, 0.0905, 0.0735, 0.0694, 0.0569, 0.0564, 0.0758, 0.0703], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:03:55,441 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277533.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:03:57,657 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6117, 5.5568, 5.4495, 4.9825, 5.1184, 5.5293, 5.4435, 5.1634], device='cuda:0'), covar=tensor([0.0512, 0.0515, 0.0300, 0.0336, 0.0971, 0.0450, 0.0280, 0.0680], device='cuda:0'), in_proj_covar=tensor([0.0326, 0.0491, 0.0382, 0.0384, 0.0379, 0.0440, 0.0260, 0.0457], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:04:23,507 INFO [train.py:904] (0/8) Epoch 28, batch 3500, loss[loss=0.1422, simple_loss=0.2258, pruned_loss=0.0293, over 16827.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2501, pruned_loss=0.03748, over 3331677.89 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:04:25,808 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277554.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:27,124 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5643, 3.6967, 4.0421, 2.3111, 3.2496, 2.6555, 4.0245, 3.8920], device='cuda:0'), covar=tensor([0.0266, 0.0950, 0.0468, 0.2035, 0.0818, 0.0979, 0.0559, 0.1067], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0174, 0.0172, 0.0158, 0.0150, 0.0134, 0.0148, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:04:42,188 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277567.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:47,072 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277570.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:54,500 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:04:55,304 INFO [optim.py:368] (0/8) 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 13:05:10,720 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277587.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 13:05:32,042 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0940, 3.1510, 2.9978, 5.2052, 4.3768, 4.5742, 1.9650, 3.3884], device='cuda:0'), covar=tensor([0.1297, 0.0759, 0.1119, 0.0234, 0.0205, 0.0397, 0.1560, 0.0738], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0206, 0.0208, 0.0221, 0.0210, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:05:32,650 INFO [train.py:904] (0/8) Epoch 28, batch 3550, loss[loss=0.1369, simple_loss=0.2194, pruned_loss=0.02724, over 16767.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2494, pruned_loss=0.03748, over 3322693.04 frames. ], batch size: 39, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:05:49,394 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277615.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:11,061 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:06:35,592 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.07 vs. limit=5.0 2023-05-02 13:06:42,328 INFO [train.py:904] (0/8) Epoch 28, batch 3600, loss[loss=0.1317, simple_loss=0.2231, pruned_loss=0.02009, over 16885.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2482, pruned_loss=0.03707, over 3319874.79 frames. ], batch size: 42, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:06:47,445 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 13:06:52,887 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:01,955 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277667.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:14,468 INFO [optim.py:368] (0/8) 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 13:07:36,972 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:07:53,658 INFO [train.py:904] (0/8) Epoch 28, batch 3650, loss[loss=0.1708, simple_loss=0.2494, pruned_loss=0.04609, over 11349.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2481, pruned_loss=0.03794, over 3290582.02 frames. ], batch size: 246, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:08:12,453 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:08:56,256 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2799, 5.2449, 5.0136, 4.3727, 5.2157, 1.7160, 4.9464, 4.7106], device='cuda:0'), covar=tensor([0.0122, 0.0105, 0.0233, 0.0408, 0.0080, 0.3202, 0.0131, 0.0265], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0180, 0.0219, 0.0191, 0.0195, 0.0224, 0.0208, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:09:06,655 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=277752.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 13:09:07,238 INFO [train.py:904] (0/8) Epoch 28, batch 3700, loss[loss=0.1819, simple_loss=0.2523, pruned_loss=0.05577, over 16867.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.2473, pruned_loss=0.03969, over 3262046.54 frames. ], batch size: 109, lr: 2.41e-03, grad_scale: 8.0 2023-05-02 13:09:41,446 INFO [optim.py:368] (0/8) 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 13:09:54,826 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6858, 4.4998, 4.5941, 4.8658, 4.9743, 4.5726, 4.9856, 5.0190], device='cuda:0'), covar=tensor([0.1965, 0.1437, 0.2084, 0.0931, 0.0842, 0.1079, 0.1503, 0.1020], device='cuda:0'), in_proj_covar=tensor([0.0708, 0.0860, 0.1003, 0.0878, 0.0667, 0.0699, 0.0734, 0.0847], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:10:22,512 INFO [train.py:904] (0/8) Epoch 28, batch 3750, loss[loss=0.1745, simple_loss=0.2691, pruned_loss=0.03996, over 17101.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2478, pruned_loss=0.04071, over 3258342.64 frames. ], batch size: 49, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:10:50,070 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 13:11:06,895 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:11:35,749 INFO [train.py:904] (0/8) Epoch 28, batch 3800, loss[loss=0.1632, simple_loss=0.2436, pruned_loss=0.0414, over 16778.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2487, pruned_loss=0.04157, over 3262684.37 frames. ], batch size: 102, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:08,763 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277875.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:11,082 INFO [optim.py:368] (0/8) 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 13:12:17,665 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:12:48,155 INFO [train.py:904] (0/8) Epoch 28, batch 3850, loss[loss=0.1604, simple_loss=0.2387, pruned_loss=0.04105, over 16808.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2495, pruned_loss=0.04268, over 3256258.65 frames. ], batch size: 83, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:12:59,678 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277910.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:14,036 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-05-02 13:13:18,282 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=277923.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:23,017 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=277926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:13:29,144 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2279, 5.5357, 5.2972, 5.3261, 5.0489, 4.9374, 4.9594, 5.6534], device='cuda:0'), covar=tensor([0.1250, 0.0863, 0.1011, 0.0876, 0.0782, 0.0905, 0.1213, 0.0813], device='cuda:0'), in_proj_covar=tensor([0.0738, 0.0900, 0.0734, 0.0691, 0.0567, 0.0563, 0.0753, 0.0700], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:13:47,038 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 13:14:00,822 INFO [train.py:904] (0/8) Epoch 28, batch 3900, loss[loss=0.1584, simple_loss=0.2345, pruned_loss=0.04114, over 16746.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2486, pruned_loss=0.04284, over 3259548.04 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:14:13,208 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=277961.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:17,850 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8529, 4.0722, 3.0532, 2.4355, 2.6554, 2.6189, 4.2287, 3.5029], device='cuda:0'), covar=tensor([0.2752, 0.0548, 0.1854, 0.3107, 0.2956, 0.2161, 0.0540, 0.1403], device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0278, 0.0315, 0.0329, 0.0309, 0.0279, 0.0306, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:14:27,572 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=277971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:14:32,416 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0957, 3.8701, 4.4131, 2.2679, 4.6306, 4.6872, 3.2816, 3.5453], device='cuda:0'), covar=tensor([0.0740, 0.0304, 0.0202, 0.1211, 0.0067, 0.0109, 0.0418, 0.0442], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0140, 0.0088, 0.0133, 0.0131, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:14:36,143 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.618e+02 2.147e+02 2.438e+02 2.765e+02 5.958e+02, threshold=4.876e+02, percent-clipped=2.0 2023-05-02 13:14:58,158 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.09 vs. limit=5.0 2023-05-02 13:15:09,540 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-278000.pt 2023-05-02 13:15:16,810 INFO [train.py:904] (0/8) Epoch 28, batch 3950, loss[loss=0.1657, simple_loss=0.2446, pruned_loss=0.04345, over 16696.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2482, pruned_loss=0.04309, over 3271083.38 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 4.0 2023-05-02 13:15:25,691 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278009.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:15:58,263 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278032.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:15:59,496 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5739, 3.2383, 3.6365, 1.9716, 3.7176, 3.7077, 3.0838, 2.8314], device='cuda:0'), covar=tensor([0.0752, 0.0322, 0.0199, 0.1172, 0.0111, 0.0248, 0.0402, 0.0476], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0139, 0.0087, 0.0133, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:16:04,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3997, 3.4037, 2.1791, 3.5635, 2.7232, 3.5920, 2.2953, 2.7925], device='cuda:0'), covar=tensor([0.0283, 0.0425, 0.1537, 0.0342, 0.0744, 0.0768, 0.1428, 0.0769], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0185, 0.0199, 0.0178, 0.0183, 0.0226, 0.0207, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:16:19,770 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278047.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 13:16:19,861 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1581, 4.0566, 4.2212, 4.3308, 4.3994, 4.0058, 4.1828, 4.4203], device='cuda:0'), covar=tensor([0.1659, 0.1127, 0.1242, 0.0709, 0.0666, 0.1376, 0.2787, 0.0742], device='cuda:0'), in_proj_covar=tensor([0.0702, 0.0856, 0.0994, 0.0873, 0.0663, 0.0694, 0.0728, 0.0840], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:16:26,879 INFO [train.py:904] (0/8) Epoch 28, batch 4000, loss[loss=0.179, simple_loss=0.2669, pruned_loss=0.0456, over 16497.00 frames. ], tot_loss[loss=0.1683, simple_loss=0.2488, pruned_loss=0.04391, over 3268143.70 frames. ], batch size: 75, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:17:01,537 INFO [optim.py:368] (0/8) 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 13:17:16,111 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-02 13:17:18,416 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4041, 3.4817, 3.7323, 2.3345, 3.1899, 2.5293, 3.8623, 3.8835], device='cuda:0'), covar=tensor([0.0239, 0.0823, 0.0597, 0.2003, 0.0878, 0.0950, 0.0492, 0.0840], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0173, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:17:32,730 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2581, 4.2538, 4.1262, 3.3073, 4.1394, 1.6754, 3.9077, 3.5072], device='cuda:0'), covar=tensor([0.0112, 0.0093, 0.0205, 0.0337, 0.0090, 0.3398, 0.0134, 0.0341], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0179, 0.0218, 0.0191, 0.0194, 0.0223, 0.0207, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:17:38,531 INFO [train.py:904] (0/8) Epoch 28, batch 4050, loss[loss=0.1743, simple_loss=0.2516, pruned_loss=0.04847, over 16737.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2499, pruned_loss=0.04348, over 3260266.38 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:17:55,661 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9565, 2.7021, 2.6076, 4.6552, 3.4754, 4.0401, 1.7693, 3.0538], device='cuda:0'), covar=tensor([0.1243, 0.0849, 0.1261, 0.0140, 0.0327, 0.0396, 0.1623, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0181, 0.0201, 0.0205, 0.0208, 0.0219, 0.0210, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:18:22,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3417, 2.5756, 2.5051, 4.3727, 2.4300, 2.9094, 2.5873, 2.6397], device='cuda:0'), covar=tensor([0.1393, 0.3321, 0.2760, 0.0426, 0.3671, 0.2269, 0.3352, 0.2981], device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0481, 0.0390, 0.0342, 0.0448, 0.0551, 0.0451, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:18:52,733 INFO [train.py:904] (0/8) Epoch 28, batch 4100, loss[loss=0.1808, simple_loss=0.2575, pruned_loss=0.05211, over 12141.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2514, pruned_loss=0.04272, over 3245723.48 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:19:14,571 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 13:19:29,064 INFO [optim.py:368] (0/8) 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 13:20:00,985 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1452, 5.4713, 5.0044, 5.3746, 5.0109, 4.7519, 5.0145, 5.5244], device='cuda:0'), covar=tensor([0.1902, 0.1311, 0.2104, 0.1358, 0.1370, 0.1468, 0.2253, 0.1603], device='cuda:0'), in_proj_covar=tensor([0.0729, 0.0892, 0.0725, 0.0684, 0.0562, 0.0558, 0.0746, 0.0692], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:20:09,080 INFO [train.py:904] (0/8) Epoch 28, batch 4150, loss[loss=0.2386, simple_loss=0.3084, pruned_loss=0.0844, over 11411.00 frames. ], tot_loss[loss=0.1738, simple_loss=0.2578, pruned_loss=0.04494, over 3203970.34 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:20:19,199 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278210.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:20:40,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3416, 4.1897, 4.1657, 4.5817, 4.6329, 4.2886, 4.6131, 4.7393], device='cuda:0'), covar=tensor([0.1842, 0.1451, 0.2198, 0.0932, 0.0908, 0.1728, 0.1330, 0.0999], device='cuda:0'), in_proj_covar=tensor([0.0697, 0.0849, 0.0985, 0.0864, 0.0658, 0.0688, 0.0721, 0.0833], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:20:42,437 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278226.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:07,624 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4371, 2.1195, 1.7763, 1.8602, 2.4190, 2.0529, 2.1707, 2.5502], device='cuda:0'), covar=tensor([0.0225, 0.0482, 0.0647, 0.0535, 0.0315, 0.0450, 0.0225, 0.0299], device='cuda:0'), in_proj_covar=tensor([0.0238, 0.0250, 0.0237, 0.0238, 0.0251, 0.0248, 0.0249, 0.0249], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:21:22,143 INFO [train.py:904] (0/8) Epoch 28, batch 4200, loss[loss=0.2018, simple_loss=0.2949, pruned_loss=0.0543, over 16864.00 frames. ], tot_loss[loss=0.1785, simple_loss=0.2647, pruned_loss=0.04615, over 3194066.63 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:21:22,888 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 13:21:30,671 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278258.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:54,356 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278274.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:21:58,172 INFO [optim.py:368] (0/8) 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 13:22:00,299 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 13:22:04,583 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.50 vs. limit=5.0 2023-05-02 13:22:22,421 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3410, 3.4321, 2.1635, 3.7394, 2.5535, 3.7677, 2.3169, 2.7621], device='cuda:0'), covar=tensor([0.0303, 0.0400, 0.1733, 0.0297, 0.0865, 0.0581, 0.1581, 0.0839], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0184, 0.0198, 0.0177, 0.0182, 0.0224, 0.0206, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:22:37,404 INFO [train.py:904] (0/8) Epoch 28, batch 4250, loss[loss=0.1909, simple_loss=0.2832, pruned_loss=0.04928, over 17263.00 frames. ], tot_loss[loss=0.1804, simple_loss=0.268, pruned_loss=0.04634, over 3182431.62 frames. ], batch size: 43, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:22:52,635 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 13:23:07,779 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:13,458 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:23:27,785 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0047, 3.2546, 3.6267, 2.1131, 2.9667, 2.3254, 3.3831, 3.4689], device='cuda:0'), covar=tensor([0.0258, 0.0961, 0.0548, 0.2176, 0.0929, 0.0998, 0.0662, 0.1035], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0173, 0.0171, 0.0157, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:23:44,001 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278347.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 13:23:52,560 INFO [train.py:904] (0/8) Epoch 28, batch 4300, loss[loss=0.1879, simple_loss=0.2804, pruned_loss=0.04768, over 17128.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2693, pruned_loss=0.04542, over 3193782.55 frames. ], batch size: 48, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:24:25,547 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278374.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:29,327 INFO [optim.py:368] (0/8) 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 13:24:40,249 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:24:56,296 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278395.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:25:07,429 INFO [train.py:904] (0/8) Epoch 28, batch 4350, loss[loss=0.1866, simple_loss=0.2754, pruned_loss=0.0489, over 16748.00 frames. ], tot_loss[loss=0.1826, simple_loss=0.2724, pruned_loss=0.04644, over 3189444.90 frames. ], batch size: 76, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:25:56,491 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278435.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:26:22,397 INFO [train.py:904] (0/8) Epoch 28, batch 4400, loss[loss=0.1948, simple_loss=0.2777, pruned_loss=0.05598, over 11542.00 frames. ], tot_loss[loss=0.1851, simple_loss=0.2749, pruned_loss=0.04765, over 3180219.34 frames. ], batch size: 246, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:26:58,181 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.489e+02 2.247e+02 2.621e+02 3.049e+02 4.749e+02, threshold=5.243e+02, percent-clipped=0.0 2023-05-02 13:27:35,977 INFO [train.py:904] (0/8) Epoch 28, batch 4450, loss[loss=0.2058, simple_loss=0.3034, pruned_loss=0.05413, over 16805.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2784, pruned_loss=0.04912, over 3197086.87 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:27:47,357 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=278511.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:28:49,082 INFO [train.py:904] (0/8) Epoch 28, batch 4500, loss[loss=0.1845, simple_loss=0.277, pruned_loss=0.04601, over 16480.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2788, pruned_loss=0.04988, over 3196539.30 frames. ], batch size: 75, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:29:17,584 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=278572.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:29:23,508 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.361e+02 1.834e+02 2.148e+02 2.381e+02 3.720e+02, threshold=4.296e+02, percent-clipped=0.0 2023-05-02 13:30:01,305 INFO [train.py:904] (0/8) Epoch 28, batch 4550, loss[loss=0.1967, simple_loss=0.2865, pruned_loss=0.05348, over 16443.00 frames. ], tot_loss[loss=0.1903, simple_loss=0.2791, pruned_loss=0.05071, over 3206636.14 frames. ], batch size: 68, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:30:36,967 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278627.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:12,171 INFO [train.py:904] (0/8) Epoch 28, batch 4600, loss[loss=0.1993, simple_loss=0.2921, pruned_loss=0.05323, over 16691.00 frames. ], tot_loss[loss=0.1908, simple_loss=0.28, pruned_loss=0.05078, over 3216056.82 frames. ], batch size: 57, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:31:43,310 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=278675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:31:46,046 INFO [optim.py:368] (0/8) 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 13:31:49,278 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:32:07,752 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 13:32:09,292 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.70 vs. limit=5.0 2023-05-02 13:32:09,885 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7304, 5.9656, 5.7393, 5.8266, 5.4255, 5.2273, 5.4573, 6.1309], device='cuda:0'), covar=tensor([0.1107, 0.0834, 0.0953, 0.0839, 0.0755, 0.0686, 0.1119, 0.0756], device='cuda:0'), in_proj_covar=tensor([0.0723, 0.0882, 0.0718, 0.0678, 0.0555, 0.0552, 0.0735, 0.0686], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:32:22,032 INFO [train.py:904] (0/8) Epoch 28, batch 4650, loss[loss=0.184, simple_loss=0.2654, pruned_loss=0.0513, over 16996.00 frames. ], tot_loss[loss=0.1911, simple_loss=0.2795, pruned_loss=0.05134, over 3211298.46 frames. ], batch size: 55, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:32:55,215 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.8358, 6.1445, 5.8516, 6.0394, 5.6066, 5.4032, 5.6688, 6.3086], device='cuda:0'), covar=tensor([0.1228, 0.0837, 0.1035, 0.0760, 0.0825, 0.0615, 0.1047, 0.0787], device='cuda:0'), in_proj_covar=tensor([0.0723, 0.0881, 0.0718, 0.0677, 0.0555, 0.0552, 0.0735, 0.0686], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:33:00,846 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278730.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:33:33,728 INFO [train.py:904] (0/8) Epoch 28, batch 4700, loss[loss=0.1584, simple_loss=0.245, pruned_loss=0.03595, over 16241.00 frames. ], tot_loss[loss=0.1887, simple_loss=0.2768, pruned_loss=0.05026, over 3217384.77 frames. ], batch size: 35, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:34:07,043 INFO [optim.py:368] (0/8) 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 13:34:45,690 INFO [train.py:904] (0/8) Epoch 28, batch 4750, loss[loss=0.1768, simple_loss=0.2666, pruned_loss=0.04351, over 16351.00 frames. ], tot_loss[loss=0.1855, simple_loss=0.2736, pruned_loss=0.04869, over 3213247.26 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:34:48,486 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2591, 3.4565, 3.6644, 2.0867, 3.0851, 2.3568, 3.6475, 3.7440], device='cuda:0'), covar=tensor([0.0258, 0.0843, 0.0638, 0.2140, 0.0871, 0.0972, 0.0580, 0.0998], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0172, 0.0170, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:35:12,087 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-05-02 13:35:57,547 INFO [train.py:904] (0/8) Epoch 28, batch 4800, loss[loss=0.1864, simple_loss=0.2804, pruned_loss=0.04626, over 16725.00 frames. ], tot_loss[loss=0.1807, simple_loss=0.2691, pruned_loss=0.0461, over 3216097.93 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:36:18,308 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=278867.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:36:27,732 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2232, 5.4874, 5.2832, 5.3049, 5.0273, 4.9524, 4.9097, 5.5965], device='cuda:0'), covar=tensor([0.1133, 0.0797, 0.0881, 0.0797, 0.0768, 0.0715, 0.1044, 0.0793], device='cuda:0'), in_proj_covar=tensor([0.0719, 0.0879, 0.0715, 0.0675, 0.0553, 0.0549, 0.0733, 0.0683], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:36:32,585 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.390e+02 1.887e+02 2.183e+02 2.468e+02 4.688e+02, threshold=4.367e+02, percent-clipped=1.0 2023-05-02 13:37:13,031 INFO [train.py:904] (0/8) Epoch 28, batch 4850, loss[loss=0.1817, simple_loss=0.2741, pruned_loss=0.04464, over 16747.00 frames. ], tot_loss[loss=0.1806, simple_loss=0.2701, pruned_loss=0.04554, over 3198981.10 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:38:20,018 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6146, 3.8082, 2.8262, 2.2537, 2.4464, 2.5048, 4.1096, 3.2332], device='cuda:0'), covar=tensor([0.2980, 0.0612, 0.1929, 0.2930, 0.2589, 0.2016, 0.0463, 0.1325], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0276, 0.0313, 0.0327, 0.0307, 0.0277, 0.0304, 0.0354], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:38:26,577 INFO [train.py:904] (0/8) Epoch 28, batch 4900, loss[loss=0.1814, simple_loss=0.2671, pruned_loss=0.04781, over 11940.00 frames. ], tot_loss[loss=0.1794, simple_loss=0.2693, pruned_loss=0.04469, over 3181743.89 frames. ], batch size: 247, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:00,731 INFO [optim.py:368] (0/8) 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 13:39:04,610 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=278979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:39:39,802 INFO [train.py:904] (0/8) Epoch 28, batch 4950, loss[loss=0.1817, simple_loss=0.274, pruned_loss=0.04476, over 16524.00 frames. ], tot_loss[loss=0.1781, simple_loss=0.2686, pruned_loss=0.04377, over 3202153.95 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:39:50,915 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9191, 3.2051, 3.3376, 1.9933, 2.8938, 2.2132, 3.4060, 3.4658], device='cuda:0'), covar=tensor([0.0269, 0.0847, 0.0722, 0.2197, 0.0923, 0.1053, 0.0666, 0.0905], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0170, 0.0156, 0.0148, 0.0132, 0.0145, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:40:14,414 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:19,463 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279030.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:40:51,498 INFO [train.py:904] (0/8) Epoch 28, batch 5000, loss[loss=0.1802, simple_loss=0.2829, pruned_loss=0.03874, over 15510.00 frames. ], tot_loss[loss=0.1792, simple_loss=0.2708, pruned_loss=0.04382, over 3199487.35 frames. ], batch size: 191, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:41:26,578 INFO [optim.py:368] (0/8) 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 13:41:27,924 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279078.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:41:39,668 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0419, 4.1341, 3.9237, 3.6589, 3.6464, 4.0557, 3.7148, 3.8347], device='cuda:0'), covar=tensor([0.0567, 0.0522, 0.0319, 0.0302, 0.0801, 0.0472, 0.1164, 0.0569], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0468, 0.0365, 0.0367, 0.0363, 0.0420, 0.0248, 0.0435], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:41:55,797 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5337, 4.6504, 4.8272, 4.5189, 4.6655, 5.1671, 4.6785, 4.3222], device='cuda:0'), covar=tensor([0.1323, 0.1783, 0.1861, 0.2104, 0.2431, 0.0954, 0.1520, 0.2419], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0632, 0.0693, 0.0516, 0.0685, 0.0721, 0.0541, 0.0687], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:42:04,202 INFO [train.py:904] (0/8) Epoch 28, batch 5050, loss[loss=0.176, simple_loss=0.2742, pruned_loss=0.03892, over 16530.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2718, pruned_loss=0.04361, over 3212827.84 frames. ], batch size: 75, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:42:45,685 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279131.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:17,187 INFO [train.py:904] (0/8) Epoch 28, batch 5100, loss[loss=0.1626, simple_loss=0.262, pruned_loss=0.03155, over 16828.00 frames. ], tot_loss[loss=0.1779, simple_loss=0.2699, pruned_loss=0.04295, over 3223916.52 frames. ], batch size: 102, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:43:37,521 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279167.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:43:52,653 INFO [optim.py:368] (0/8) 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 13:44:15,309 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:44:30,693 INFO [train.py:904] (0/8) Epoch 28, batch 5150, loss[loss=0.1869, simple_loss=0.2812, pruned_loss=0.04632, over 16885.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2698, pruned_loss=0.0426, over 3206661.68 frames. ], batch size: 116, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:44:48,996 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279215.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:44:51,749 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5104, 2.6333, 2.5408, 4.3339, 2.4934, 2.9827, 2.6367, 2.7569], device='cuda:0'), covar=tensor([0.1409, 0.3254, 0.2845, 0.0491, 0.3548, 0.2404, 0.3544, 0.2795], device='cuda:0'), in_proj_covar=tensor([0.0422, 0.0477, 0.0385, 0.0338, 0.0446, 0.0545, 0.0448, 0.0555], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:45:14,318 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0420, 3.3982, 3.6277, 2.0705, 3.0833, 2.5209, 3.4158, 3.7043], device='cuda:0'), covar=tensor([0.0357, 0.0846, 0.0601, 0.2081, 0.0883, 0.0925, 0.0777, 0.0832], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0158, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:45:42,233 INFO [train.py:904] (0/8) Epoch 28, batch 5200, loss[loss=0.1605, simple_loss=0.2538, pruned_loss=0.0336, over 16472.00 frames. ], tot_loss[loss=0.1757, simple_loss=0.2681, pruned_loss=0.04165, over 3207031.41 frames. ], batch size: 146, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:46:17,348 INFO [optim.py:368] (0/8) 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 13:46:21,101 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5668, 3.5814, 3.4148, 1.8171, 2.8954, 2.0556, 3.8274, 4.0008], device='cuda:0'), covar=tensor([0.0216, 0.0888, 0.0819, 0.2726, 0.1132, 0.1337, 0.0611, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0158, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:46:53,553 INFO [train.py:904] (0/8) Epoch 28, batch 5250, loss[loss=0.1744, simple_loss=0.2737, pruned_loss=0.03753, over 16233.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2656, pruned_loss=0.04145, over 3204842.61 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:48:06,865 INFO [train.py:904] (0/8) Epoch 28, batch 5300, loss[loss=0.1546, simple_loss=0.2427, pruned_loss=0.03327, over 16981.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2617, pruned_loss=0.04006, over 3212680.63 frames. ], batch size: 41, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:48:41,219 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.441e+02 1.963e+02 2.333e+02 2.759e+02 6.747e+02, threshold=4.665e+02, percent-clipped=2.0 2023-05-02 13:49:20,984 INFO [train.py:904] (0/8) Epoch 28, batch 5350, loss[loss=0.1851, simple_loss=0.2791, pruned_loss=0.04555, over 16721.00 frames. ], tot_loss[loss=0.17, simple_loss=0.2606, pruned_loss=0.03974, over 3218969.37 frames. ], batch size: 124, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:49:46,514 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4185, 2.9772, 2.6999, 2.3304, 2.2643, 2.3297, 2.9746, 2.8449], device='cuda:0'), covar=tensor([0.2639, 0.0645, 0.1685, 0.2536, 0.2290, 0.2092, 0.0497, 0.1252], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0275, 0.0312, 0.0327, 0.0305, 0.0276, 0.0305, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:50:16,881 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2041, 5.1887, 5.0807, 4.6482, 4.7097, 5.1105, 5.0277, 4.8365], device='cuda:0'), covar=tensor([0.0624, 0.0580, 0.0302, 0.0341, 0.1144, 0.0503, 0.0292, 0.0599], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0469, 0.0365, 0.0368, 0.0364, 0.0422, 0.0249, 0.0435], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:50:32,510 INFO [train.py:904] (0/8) Epoch 28, batch 5400, loss[loss=0.1787, simple_loss=0.2687, pruned_loss=0.04436, over 16539.00 frames. ], tot_loss[loss=0.1719, simple_loss=0.263, pruned_loss=0.04041, over 3220117.74 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:50:46,812 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4941, 3.3355, 3.9722, 2.0152, 4.1203, 4.0695, 3.0903, 2.9782], device='cuda:0'), covar=tensor([0.0898, 0.0320, 0.0151, 0.1157, 0.0064, 0.0137, 0.0401, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0137, 0.0086, 0.0129, 0.0128, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:50:51,014 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8259, 4.2295, 2.9802, 2.4774, 3.0083, 2.5808, 4.6424, 3.6537], device='cuda:0'), covar=tensor([0.3030, 0.0639, 0.2039, 0.2684, 0.2645, 0.2094, 0.0460, 0.1326], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0275, 0.0312, 0.0327, 0.0305, 0.0276, 0.0304, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 13:51:08,326 INFO [optim.py:368] (0/8) 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 13:51:24,277 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:51:50,354 INFO [train.py:904] (0/8) Epoch 28, batch 5450, loss[loss=0.2368, simple_loss=0.3217, pruned_loss=0.0759, over 15229.00 frames. ], tot_loss[loss=0.1742, simple_loss=0.2654, pruned_loss=0.04153, over 3199546.86 frames. ], batch size: 190, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:52:09,661 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3223, 3.4537, 3.5794, 3.5653, 3.5830, 3.4535, 3.4409, 3.4600], device='cuda:0'), covar=tensor([0.0455, 0.0855, 0.0614, 0.0545, 0.0584, 0.0756, 0.0975, 0.0621], device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0486, 0.0470, 0.0432, 0.0515, 0.0494, 0.0570, 0.0397], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 13:52:31,318 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279529.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:52:50,422 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.69 vs. limit=5.0 2023-05-02 13:52:52,443 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=279542.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:53:09,376 INFO [train.py:904] (0/8) Epoch 28, batch 5500, loss[loss=0.1873, simple_loss=0.2783, pruned_loss=0.04815, over 16555.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2725, pruned_loss=0.04554, over 3180037.86 frames. ], batch size: 62, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:53:14,904 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0408, 4.1219, 3.9258, 3.6806, 3.6777, 4.0516, 3.7033, 3.8498], device='cuda:0'), covar=tensor([0.0612, 0.0630, 0.0304, 0.0304, 0.0753, 0.0514, 0.1118, 0.0556], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0473, 0.0366, 0.0370, 0.0366, 0.0424, 0.0250, 0.0437], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:53:47,386 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.670e+02 2.730e+02 3.219e+02 3.961e+02 9.894e+02, threshold=6.439e+02, percent-clipped=12.0 2023-05-02 13:54:09,312 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279590.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:28,769 INFO [train.py:904] (0/8) Epoch 28, batch 5550, loss[loss=0.2763, simple_loss=0.3373, pruned_loss=0.1077, over 11536.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2794, pruned_loss=0.04989, over 3168953.39 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:54:29,358 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=279603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:54:34,467 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6302, 3.4152, 3.9653, 2.0532, 4.1208, 4.0866, 3.1009, 3.0705], device='cuda:0'), covar=tensor([0.0828, 0.0323, 0.0225, 0.1170, 0.0088, 0.0186, 0.0435, 0.0487], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0110, 0.0101, 0.0137, 0.0086, 0.0129, 0.0128, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 13:55:48,658 INFO [train.py:904] (0/8) Epoch 28, batch 5600, loss[loss=0.2029, simple_loss=0.2912, pruned_loss=0.05732, over 16718.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2833, pruned_loss=0.0534, over 3140007.10 frames. ], batch size: 76, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:56:28,893 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.204e+02 3.162e+02 3.859e+02 4.792e+02 1.536e+03, threshold=7.717e+02, percent-clipped=11.0 2023-05-02 13:57:11,936 INFO [train.py:904] (0/8) Epoch 28, batch 5650, loss[loss=0.2571, simple_loss=0.321, pruned_loss=0.09659, over 11298.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2883, pruned_loss=0.05777, over 3089173.48 frames. ], batch size: 247, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:58:01,641 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2069, 3.3201, 3.6278, 2.1399, 3.0398, 2.3225, 3.6473, 3.7137], device='cuda:0'), covar=tensor([0.0254, 0.0902, 0.0605, 0.2199, 0.0900, 0.1040, 0.0600, 0.0985], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0158, 0.0148, 0.0133, 0.0147, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 13:58:07,489 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0565, 2.2018, 2.1869, 3.4802, 2.1383, 2.5325, 2.2998, 2.3154], device='cuda:0'), covar=tensor([0.1486, 0.3508, 0.3212, 0.0714, 0.4103, 0.2327, 0.3461, 0.3369], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0474, 0.0384, 0.0336, 0.0444, 0.0542, 0.0445, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 13:58:29,305 INFO [train.py:904] (0/8) Epoch 28, batch 5700, loss[loss=0.1974, simple_loss=0.2971, pruned_loss=0.04886, over 16674.00 frames. ], tot_loss[loss=0.2036, simple_loss=0.2894, pruned_loss=0.05889, over 3079031.12 frames. ], batch size: 89, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 13:59:05,448 INFO [optim.py:368] (0/8) 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 13:59:22,122 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=279787.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 13:59:46,958 INFO [train.py:904] (0/8) Epoch 28, batch 5750, loss[loss=0.1894, simple_loss=0.2778, pruned_loss=0.05044, over 17038.00 frames. ], tot_loss[loss=0.2062, simple_loss=0.2923, pruned_loss=0.06004, over 3088684.00 frames. ], batch size: 53, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:00:39,196 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=279835.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:01:01,098 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7281, 2.5707, 2.6857, 4.6949, 3.3851, 4.1656, 1.6528, 3.0563], device='cuda:0'), covar=tensor([0.1490, 0.0943, 0.1275, 0.0159, 0.0283, 0.0408, 0.1800, 0.0853], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0180, 0.0201, 0.0204, 0.0207, 0.0218, 0.0210, 0.0199], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 14:01:07,318 INFO [train.py:904] (0/8) Epoch 28, batch 5800, loss[loss=0.1764, simple_loss=0.2739, pruned_loss=0.03945, over 17239.00 frames. ], tot_loss[loss=0.2053, simple_loss=0.2923, pruned_loss=0.05915, over 3076760.26 frames. ], batch size: 52, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:01:46,120 INFO [optim.py:368] (0/8) 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 14:01:57,438 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279885.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:09,980 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8650, 2.7065, 2.6244, 1.9291, 2.5676, 2.7136, 2.5713, 1.9197], device='cuda:0'), covar=tensor([0.0483, 0.0110, 0.0107, 0.0407, 0.0163, 0.0148, 0.0137, 0.0435], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0135, 0.0102, 0.0115, 0.0098, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 14:02:16,878 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=279898.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:02:25,216 INFO [train.py:904] (0/8) Epoch 28, batch 5850, loss[loss=0.1964, simple_loss=0.2845, pruned_loss=0.05416, over 16734.00 frames. ], tot_loss[loss=0.2026, simple_loss=0.2901, pruned_loss=0.05758, over 3091706.94 frames. ], batch size: 134, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:02:49,528 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 14:03:44,223 INFO [train.py:904] (0/8) Epoch 28, batch 5900, loss[loss=0.1979, simple_loss=0.2796, pruned_loss=0.05808, over 16250.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2896, pruned_loss=0.05712, over 3108180.58 frames. ], batch size: 165, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:04:10,145 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7737, 3.0600, 3.2904, 1.9670, 2.7895, 2.0854, 3.2989, 3.2827], device='cuda:0'), covar=tensor([0.0271, 0.0866, 0.0659, 0.2197, 0.0899, 0.1093, 0.0639, 0.0854], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0172, 0.0159, 0.0149, 0.0133, 0.0147, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 14:04:26,171 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.117e+02 2.561e+02 3.002e+02 3.588e+02 6.021e+02, threshold=6.005e+02, percent-clipped=0.0 2023-05-02 14:04:38,851 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4004, 3.2747, 2.6259, 2.1923, 2.2417, 2.3074, 3.3526, 3.0668], device='cuda:0'), covar=tensor([0.3170, 0.0677, 0.1960, 0.3055, 0.2493, 0.2284, 0.0615, 0.1325], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0274, 0.0312, 0.0327, 0.0305, 0.0276, 0.0304, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 14:05:00,801 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-280000.pt 2023-05-02 14:05:07,652 INFO [train.py:904] (0/8) Epoch 28, batch 5950, loss[loss=0.2085, simple_loss=0.2939, pruned_loss=0.06158, over 12025.00 frames. ], tot_loss[loss=0.2016, simple_loss=0.2909, pruned_loss=0.05609, over 3112003.32 frames. ], batch size: 248, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:20,850 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280051.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:06:23,834 INFO [train.py:904] (0/8) Epoch 28, batch 6000, loss[loss=0.184, simple_loss=0.2784, pruned_loss=0.04476, over 16870.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2904, pruned_loss=0.0566, over 3104024.77 frames. ], batch size: 96, lr: 2.40e-03, grad_scale: 8.0 2023-05-02 14:06:23,834 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 14:06:34,281 INFO [train.py:938] (0/8) Epoch 28, validation: loss=0.148, simple_loss=0.2602, pruned_loss=0.0179, over 944034.00 frames. 2023-05-02 14:06:34,282 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 14:06:47,172 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8805, 1.9353, 2.4976, 2.8144, 2.7114, 3.2338, 2.2356, 3.2131], device='cuda:0'), covar=tensor([0.0277, 0.0626, 0.0365, 0.0373, 0.0382, 0.0207, 0.0563, 0.0156], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0199, 0.0188, 0.0193, 0.0209, 0.0167, 0.0204, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:07:11,171 INFO [optim.py:368] (0/8) 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 14:07:25,945 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280088.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:07:51,274 INFO [train.py:904] (0/8) Epoch 28, batch 6050, loss[loss=0.2037, simple_loss=0.3045, pruned_loss=0.05139, over 16675.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2892, pruned_loss=0.05602, over 3119401.92 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:08:06,568 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:09:01,887 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280149.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:09:08,883 INFO [train.py:904] (0/8) Epoch 28, batch 6100, loss[loss=0.2124, simple_loss=0.2971, pruned_loss=0.06386, over 16635.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2888, pruned_loss=0.05563, over 3119245.53 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:09:30,706 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4242, 3.4196, 3.7267, 2.1461, 3.1559, 2.4116, 3.8210, 3.7786], device='cuda:0'), covar=tensor([0.0238, 0.0865, 0.0601, 0.2198, 0.0839, 0.1009, 0.0618, 0.0938], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 14:09:34,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3207, 3.0891, 3.4338, 1.8574, 3.5395, 3.5503, 2.8617, 2.7494], device='cuda:0'), covar=tensor([0.0889, 0.0319, 0.0217, 0.1216, 0.0107, 0.0212, 0.0472, 0.0501], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0110, 0.0102, 0.0138, 0.0086, 0.0130, 0.0129, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 14:09:51,403 INFO [optim.py:368] (0/8) 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 14:10:02,009 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280185.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:21,990 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:10:29,431 INFO [train.py:904] (0/8) Epoch 28, batch 6150, loss[loss=0.1765, simple_loss=0.2673, pruned_loss=0.04289, over 16796.00 frames. ], tot_loss[loss=0.1988, simple_loss=0.287, pruned_loss=0.05528, over 3110931.87 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:11:16,586 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280233.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:35,650 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280246.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:11:45,500 INFO [train.py:904] (0/8) Epoch 28, batch 6200, loss[loss=0.1683, simple_loss=0.2557, pruned_loss=0.04047, over 16535.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2855, pruned_loss=0.0553, over 3093557.33 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:12:01,263 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5148, 2.5481, 1.9233, 2.4050, 2.9713, 2.6410, 3.0840, 3.2459], device='cuda:0'), covar=tensor([0.0171, 0.0572, 0.0825, 0.0589, 0.0354, 0.0491, 0.0315, 0.0343], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0242, 0.0231, 0.0231, 0.0244, 0.0240, 0.0239, 0.0241], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:12:24,025 INFO [optim.py:368] (0/8) 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 14:12:41,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7003, 2.3705, 2.2491, 3.6664, 2.2831, 3.8325, 1.3762, 2.7793], device='cuda:0'), covar=tensor([0.1537, 0.1020, 0.1514, 0.0224, 0.0254, 0.0452, 0.2073, 0.0913], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0204, 0.0209, 0.0219, 0.0211, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 14:13:00,714 INFO [train.py:904] (0/8) Epoch 28, batch 6250, loss[loss=0.1744, simple_loss=0.2685, pruned_loss=0.04016, over 16706.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2849, pruned_loss=0.05473, over 3100666.66 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:13:36,611 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:13:38,258 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 14:14:14,458 INFO [train.py:904] (0/8) Epoch 28, batch 6300, loss[loss=0.1858, simple_loss=0.2756, pruned_loss=0.04804, over 16255.00 frames. ], tot_loss[loss=0.1964, simple_loss=0.2844, pruned_loss=0.05419, over 3100622.07 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:14:53,566 INFO [optim.py:368] (0/8) 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 14:15:09,863 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:15:31,110 INFO [train.py:904] (0/8) Epoch 28, batch 6350, loss[loss=0.1835, simple_loss=0.2725, pruned_loss=0.04721, over 17133.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2855, pruned_loss=0.05524, over 3103508.53 frames. ], batch size: 48, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:15:36,798 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280407.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:04,453 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6195, 2.5504, 1.8959, 2.6186, 2.1107, 2.7679, 2.1275, 2.3912], device='cuda:0'), covar=tensor([0.0317, 0.0380, 0.1244, 0.0284, 0.0687, 0.0469, 0.1166, 0.0602], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0180, 0.0195, 0.0172, 0.0179, 0.0219, 0.0203, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 14:16:13,861 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=5.01 vs. limit=5.0 2023-05-02 14:16:31,386 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280444.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:16:43,563 INFO [train.py:904] (0/8) Epoch 28, batch 6400, loss[loss=0.2007, simple_loss=0.2803, pruned_loss=0.0605, over 16396.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.2854, pruned_loss=0.05621, over 3087561.03 frames. ], batch size: 35, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:16:58,895 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6403, 2.6172, 1.8103, 2.1462, 3.0183, 2.6497, 3.2075, 3.2837], device='cuda:0'), covar=tensor([0.0136, 0.0525, 0.0807, 0.0680, 0.0331, 0.0486, 0.0328, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0228, 0.0242, 0.0231, 0.0231, 0.0244, 0.0241, 0.0239, 0.0241], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:16:59,169 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 14:17:19,319 INFO [optim.py:368] (0/8) 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 14:17:56,104 INFO [train.py:904] (0/8) Epoch 28, batch 6450, loss[loss=0.2012, simple_loss=0.2843, pruned_loss=0.05901, over 16792.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2857, pruned_loss=0.05587, over 3094024.46 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:18:13,085 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5283, 3.6189, 3.3282, 3.0078, 3.2262, 3.5059, 3.3273, 3.3344], device='cuda:0'), covar=tensor([0.0551, 0.0541, 0.0270, 0.0284, 0.0449, 0.0497, 0.1756, 0.0462], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0469, 0.0363, 0.0366, 0.0361, 0.0419, 0.0248, 0.0433], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:18:18,348 INFO [zipformer.py:625] (0/8) 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] (0/8) 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] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:19:12,200 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9088, 1.4473, 1.7982, 1.7571, 1.8608, 1.9733, 1.6621, 1.8371], device='cuda:0'), covar=tensor([0.0297, 0.0436, 0.0272, 0.0363, 0.0301, 0.0218, 0.0461, 0.0176], device='cuda:0'), in_proj_covar=tensor([0.0198, 0.0199, 0.0187, 0.0192, 0.0208, 0.0166, 0.0202, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:19:13,565 INFO [train.py:904] (0/8) Epoch 28, batch 6500, loss[loss=0.2258, simple_loss=0.2897, pruned_loss=0.08097, over 11976.00 frames. ], tot_loss[loss=0.1972, simple_loss=0.2837, pruned_loss=0.05539, over 3082555.48 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:19:49,979 INFO [optim.py:368] (0/8) 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 14:19:52,117 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280579.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:15,911 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:28,530 INFO [train.py:904] (0/8) Epoch 28, batch 6550, loss[loss=0.1921, simple_loss=0.2981, pruned_loss=0.04309, over 16648.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2856, pruned_loss=0.05583, over 3077709.02 frames. ], batch size: 76, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:20:37,410 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280608.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:20:45,320 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280614.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:21:44,262 INFO [train.py:904] (0/8) Epoch 28, batch 6600, loss[loss=0.1914, simple_loss=0.2921, pruned_loss=0.04536, over 16449.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.2878, pruned_loss=0.0563, over 3082340.79 frames. ], batch size: 146, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:22:18,031 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=280675.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:21,700 INFO [optim.py:368] (0/8) 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 14:22:26,532 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8706, 2.1994, 2.4557, 3.1092, 2.2411, 2.4283, 2.3801, 2.2913], device='cuda:0'), covar=tensor([0.1527, 0.3149, 0.2439, 0.0783, 0.3993, 0.2159, 0.3060, 0.3297], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0473, 0.0384, 0.0335, 0.0444, 0.0541, 0.0445, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:22:28,904 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:22:59,885 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=280702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:23:01,184 INFO [train.py:904] (0/8) Epoch 28, batch 6650, loss[loss=0.2536, simple_loss=0.3223, pruned_loss=0.09241, over 11598.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.289, pruned_loss=0.05798, over 3062401.40 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:23:07,205 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:02,821 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=280744.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:15,398 INFO [train.py:904] (0/8) Epoch 28, batch 6700, loss[loss=0.1917, simple_loss=0.2717, pruned_loss=0.05587, over 16622.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2875, pruned_loss=0.0581, over 3059012.17 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:24:18,764 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280755.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:24:30,581 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.083e+02 2.827e+02 3.469e+02 4.301e+02 6.613e+02, threshold=6.939e+02, percent-clipped=0.0 2023-05-02 14:25:12,496 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=280792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:25:29,256 INFO [train.py:904] (0/8) Epoch 28, batch 6750, loss[loss=0.1882, simple_loss=0.2742, pruned_loss=0.05107, over 16890.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.286, pruned_loss=0.05759, over 3062720.62 frames. ], batch size: 116, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:26:11,002 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.53 vs. limit=5.0 2023-05-02 14:26:12,058 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2671, 4.3081, 4.6132, 4.5553, 4.5887, 4.3081, 4.2930, 4.2638], device='cuda:0'), covar=tensor([0.0372, 0.0619, 0.0387, 0.0460, 0.0493, 0.0451, 0.0971, 0.0546], device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0491, 0.0475, 0.0440, 0.0522, 0.0501, 0.0578, 0.0402], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 14:26:41,175 INFO [train.py:904] (0/8) Epoch 28, batch 6800, loss[loss=0.2269, simple_loss=0.3022, pruned_loss=0.07582, over 11692.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2861, pruned_loss=0.05734, over 3078306.68 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:13,239 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:18,680 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.178e+02 2.733e+02 3.177e+02 3.942e+02 7.279e+02, threshold=6.355e+02, percent-clipped=1.0 2023-05-02 14:27:35,924 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280889.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:27:55,353 INFO [train.py:904] (0/8) Epoch 28, batch 6850, loss[loss=0.2048, simple_loss=0.3004, pruned_loss=0.05456, over 16745.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2869, pruned_loss=0.05734, over 3092001.95 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:27:55,593 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:28:23,228 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5044, 2.8679, 3.1877, 1.8687, 2.7963, 2.0552, 3.2230, 3.2130], device='cuda:0'), covar=tensor([0.0264, 0.0864, 0.0664, 0.2357, 0.0897, 0.1157, 0.0602, 0.0808], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0172, 0.0172, 0.0158, 0.0149, 0.0133, 0.0147, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 14:28:29,538 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2712, 4.1061, 4.2917, 4.4468, 4.5848, 4.1582, 4.5194, 4.6080], device='cuda:0'), covar=tensor([0.1779, 0.1182, 0.1465, 0.0755, 0.0852, 0.1201, 0.1176, 0.0726], device='cuda:0'), in_proj_covar=tensor([0.0667, 0.0813, 0.0943, 0.0829, 0.0634, 0.0662, 0.0695, 0.0800], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:28:58,626 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5478, 2.2862, 2.9490, 3.4545, 3.1574, 3.8674, 2.4892, 3.7878], device='cuda:0'), covar=tensor([0.0179, 0.0590, 0.0366, 0.0271, 0.0345, 0.0159, 0.0661, 0.0162], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0199, 0.0186, 0.0191, 0.0207, 0.0166, 0.0202, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:29:06,761 INFO [train.py:904] (0/8) Epoch 28, batch 6900, loss[loss=0.2158, simple_loss=0.2963, pruned_loss=0.06761, over 16913.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2892, pruned_loss=0.05677, over 3108596.76 frames. ], batch size: 109, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:29:33,816 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=280970.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:29:45,363 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.637e+02 2.748e+02 3.135e+02 3.737e+02 6.416e+02, threshold=6.270e+02, percent-clipped=1.0 2023-05-02 14:29:53,338 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 28, batch 6950, loss[loss=0.1952, simple_loss=0.2781, pruned_loss=0.05617, over 16731.00 frames. ], tot_loss[loss=0.2039, simple_loss=0.2908, pruned_loss=0.05845, over 3077227.34 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:30:32,582 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.43 vs. limit=2.0 2023-05-02 14:31:03,953 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:31:30,262 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281049.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:31:35,802 INFO [train.py:904] (0/8) Epoch 28, batch 7000, loss[loss=0.2086, simple_loss=0.3039, pruned_loss=0.05664, over 15574.00 frames. ], tot_loss[loss=0.2034, simple_loss=0.291, pruned_loss=0.05789, over 3080463.53 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:31:44,106 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281058.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:31:47,672 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7450, 2.3249, 1.9993, 2.0899, 2.6864, 2.3516, 2.4209, 2.8465], device='cuda:0'), covar=tensor([0.0305, 0.0497, 0.0632, 0.0565, 0.0327, 0.0451, 0.0265, 0.0342], device='cuda:0'), in_proj_covar=tensor([0.0229, 0.0243, 0.0233, 0.0233, 0.0244, 0.0243, 0.0240, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:32:12,985 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.569e+02 2.882e+02 3.440e+02 4.031e+02 6.341e+02, threshold=6.881e+02, percent-clipped=1.0 2023-05-02 14:32:50,411 INFO [train.py:904] (0/8) Epoch 28, batch 7050, loss[loss=0.185, simple_loss=0.2797, pruned_loss=0.04511, over 16901.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.2916, pruned_loss=0.05699, over 3102360.31 frames. ], batch size: 116, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:33:01,157 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281110.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:33:08,377 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3170, 4.4717, 4.2336, 3.9659, 3.8332, 4.3879, 4.1022, 4.0149], device='cuda:0'), covar=tensor([0.0750, 0.0641, 0.0350, 0.0367, 0.0981, 0.0571, 0.0811, 0.0721], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0464, 0.0359, 0.0362, 0.0358, 0.0415, 0.0246, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:33:38,542 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7343, 3.8682, 2.5028, 4.6530, 3.0008, 4.5404, 2.6008, 3.1112], device='cuda:0'), covar=tensor([0.0327, 0.0417, 0.1692, 0.0187, 0.0894, 0.0530, 0.1535, 0.0830], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0181, 0.0196, 0.0172, 0.0180, 0.0220, 0.0204, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 14:33:43,654 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 14:33:48,676 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-02 14:34:04,190 INFO [train.py:904] (0/8) Epoch 28, batch 7100, loss[loss=0.1698, simple_loss=0.2615, pruned_loss=0.0391, over 16748.00 frames. ], tot_loss[loss=0.2017, simple_loss=0.2897, pruned_loss=0.05681, over 3081961.85 frames. ], batch size: 89, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:34:09,537 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 14:34:29,297 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281169.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:37,220 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281174.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:34:43,395 INFO [optim.py:368] (0/8) 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 14:34:59,567 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 28, batch 7150, loss[loss=0.1973, simple_loss=0.2849, pruned_loss=0.05492, over 16766.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2884, pruned_loss=0.05699, over 3083971.70 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:35:20,994 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:00,724 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281230.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 14:36:09,137 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281237.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:29,668 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281251.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:36:31,502 INFO [train.py:904] (0/8) Epoch 28, batch 7200, loss[loss=0.1774, simple_loss=0.262, pruned_loss=0.04645, over 16611.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2861, pruned_loss=0.05567, over 3076101.36 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:36:56,570 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281270.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:37:07,760 INFO [optim.py:368] (0/8) 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 14:37:46,533 INFO [train.py:904] (0/8) Epoch 28, batch 7250, loss[loss=0.1828, simple_loss=0.267, pruned_loss=0.04933, over 15495.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2838, pruned_loss=0.05485, over 3062320.61 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:38:09,209 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:38:59,575 INFO [train.py:904] (0/8) Epoch 28, batch 7300, loss[loss=0.2069, simple_loss=0.2778, pruned_loss=0.06806, over 11092.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2837, pruned_loss=0.05479, over 3064785.00 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:39:08,088 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281358.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:39:11,441 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 14:39:39,521 INFO [optim.py:368] (0/8) 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 14:40:13,658 INFO [train.py:904] (0/8) Epoch 28, batch 7350, loss[loss=0.2098, simple_loss=0.2989, pruned_loss=0.06035, over 16448.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2859, pruned_loss=0.05633, over 3055603.39 frames. ], batch size: 68, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:40:16,308 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281405.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 14:40:17,488 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:40:36,439 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4980, 3.8117, 2.8017, 2.2332, 2.4480, 2.4259, 4.1247, 3.2626], device='cuda:0'), covar=tensor([0.3335, 0.0642, 0.2020, 0.2897, 0.2892, 0.2269, 0.0439, 0.1500], device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0274, 0.0313, 0.0328, 0.0305, 0.0277, 0.0305, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 14:40:49,391 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:40:52,568 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8377, 2.0887, 2.4038, 3.0567, 2.1430, 2.3066, 2.3347, 2.2653], device='cuda:0'), covar=tensor([0.1565, 0.3513, 0.2598, 0.0785, 0.4448, 0.2504, 0.3331, 0.3318], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0474, 0.0386, 0.0337, 0.0447, 0.0544, 0.0448, 0.0555], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:41:28,060 INFO [train.py:904] (0/8) Epoch 28, batch 7400, loss[loss=0.2313, simple_loss=0.3155, pruned_loss=0.07358, over 16283.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.286, pruned_loss=0.05611, over 3070155.02 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:41:33,676 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281456.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:42:08,141 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.796e+02 2.701e+02 3.025e+02 3.387e+02 6.725e+02, threshold=6.049e+02, percent-clipped=0.0 2023-05-02 14:42:18,969 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281487.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:42:42,679 INFO [train.py:904] (0/8) Epoch 28, batch 7450, loss[loss=0.1982, simple_loss=0.2877, pruned_loss=0.0544, over 16961.00 frames. ], tot_loss[loss=0.2009, simple_loss=0.287, pruned_loss=0.05733, over 3060309.38 frames. ], batch size: 116, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:42:51,185 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8834, 2.7006, 2.5764, 1.9518, 2.5601, 2.7104, 2.5403, 1.9431], device='cuda:0'), covar=tensor([0.0479, 0.0108, 0.0104, 0.0404, 0.0162, 0.0157, 0.0143, 0.0423], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0135, 0.0101, 0.0115, 0.0098, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 14:43:05,984 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281517.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:43:17,716 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281525.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:43:58,013 INFO [train.py:904] (0/8) Epoch 28, batch 7500, loss[loss=0.1812, simple_loss=0.2737, pruned_loss=0.04438, over 16759.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2869, pruned_loss=0.05626, over 3066098.63 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:44:27,156 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1287, 2.2435, 2.2413, 3.6203, 2.2337, 2.5633, 2.3122, 2.3707], device='cuda:0'), covar=tensor([0.1469, 0.3607, 0.3152, 0.0678, 0.4093, 0.2432, 0.3606, 0.3435], device='cuda:0'), in_proj_covar=tensor([0.0421, 0.0473, 0.0385, 0.0337, 0.0446, 0.0542, 0.0448, 0.0554], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:44:36,769 INFO [optim.py:368] (0/8) 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 14:45:11,674 INFO [train.py:904] (0/8) Epoch 28, batch 7550, loss[loss=0.1955, simple_loss=0.2855, pruned_loss=0.05275, over 16628.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2856, pruned_loss=0.05627, over 3073541.33 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 2.0 2023-05-02 14:46:25,958 INFO [train.py:904] (0/8) Epoch 28, batch 7600, loss[loss=0.2055, simple_loss=0.2941, pruned_loss=0.05851, over 16764.00 frames. ], tot_loss[loss=0.1997, simple_loss=0.2856, pruned_loss=0.05692, over 3064865.01 frames. ], batch size: 124, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:47:04,828 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 28, batch 7650, loss[loss=0.2789, simple_loss=0.3384, pruned_loss=0.1097, over 11176.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2859, pruned_loss=0.05769, over 3049522.91 frames. ], batch size: 248, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:47:43,678 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281705.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:48:33,127 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4594, 3.4212, 2.7234, 2.2004, 2.3143, 2.3266, 3.5404, 3.1195], device='cuda:0'), covar=tensor([0.3066, 0.0700, 0.1829, 0.2918, 0.2647, 0.2308, 0.0535, 0.1375], device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0273, 0.0312, 0.0327, 0.0304, 0.0277, 0.0304, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 14:48:43,002 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5259, 3.9851, 3.9674, 2.8093, 3.6113, 4.0419, 3.5833, 2.3940], device='cuda:0'), covar=tensor([0.0525, 0.0063, 0.0062, 0.0398, 0.0115, 0.0124, 0.0108, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0101, 0.0115, 0.0097, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 14:48:51,549 INFO [train.py:904] (0/8) Epoch 28, batch 7700, loss[loss=0.2036, simple_loss=0.2824, pruned_loss=0.06236, over 16584.00 frames. ], tot_loss[loss=0.201, simple_loss=0.286, pruned_loss=0.05802, over 3040774.97 frames. ], batch size: 57, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:48:51,891 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281753.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 14:48:59,706 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1921, 2.0340, 1.7123, 1.7621, 2.2828, 1.9337, 1.9298, 2.3843], device='cuda:0'), covar=tensor([0.0284, 0.0425, 0.0599, 0.0524, 0.0280, 0.0399, 0.0208, 0.0287], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0241, 0.0232, 0.0231, 0.0242, 0.0240, 0.0237, 0.0240], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:49:31,077 INFO [optim.py:368] (0/8) 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 14:49:35,135 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:07,535 INFO [train.py:904] (0/8) Epoch 28, batch 7750, loss[loss=0.2223, simple_loss=0.3009, pruned_loss=0.07185, over 15281.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2864, pruned_loss=0.05788, over 3032904.88 frames. ], batch size: 191, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:50:21,376 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=281812.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:50:24,388 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2692, 5.6143, 5.1455, 5.4747, 5.0965, 4.8144, 5.1431, 5.6811], device='cuda:0'), covar=tensor([0.2328, 0.1551, 0.2753, 0.1448, 0.1720, 0.1814, 0.2347, 0.1855], device='cuda:0'), in_proj_covar=tensor([0.0716, 0.0870, 0.0712, 0.0670, 0.0547, 0.0551, 0.0724, 0.0675], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:50:40,413 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=281825.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:07,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5054, 4.6338, 4.7925, 4.5449, 4.6589, 5.1726, 4.6315, 4.3622], device='cuda:0'), covar=tensor([0.1364, 0.1999, 0.2603, 0.2161, 0.2507, 0.1000, 0.1842, 0.2676], device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0637, 0.0700, 0.0515, 0.0690, 0.0723, 0.0546, 0.0692], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 14:51:21,127 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281852.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:51:21,941 INFO [train.py:904] (0/8) Epoch 28, batch 7800, loss[loss=0.1772, simple_loss=0.273, pruned_loss=0.04068, over 16505.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2861, pruned_loss=0.05724, over 3061630.45 frames. ], batch size: 75, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:51:52,697 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=281873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:52:03,258 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.973e+02 2.715e+02 3.377e+02 4.149e+02 7.003e+02, threshold=6.754e+02, percent-clipped=1.0 2023-05-02 14:52:23,347 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9069, 2.1851, 2.4555, 3.1084, 2.2453, 2.4085, 2.3883, 2.3219], device='cuda:0'), covar=tensor([0.1486, 0.3563, 0.2668, 0.0799, 0.4651, 0.2503, 0.3351, 0.3279], device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0473, 0.0384, 0.0335, 0.0444, 0.0540, 0.0446, 0.0552], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:52:37,340 INFO [train.py:904] (0/8) Epoch 28, batch 7850, loss[loss=0.196, simple_loss=0.2844, pruned_loss=0.05384, over 16620.00 frames. ], tot_loss[loss=0.2003, simple_loss=0.2867, pruned_loss=0.05691, over 3062284.54 frames. ], batch size: 62, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:52:41,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2853, 5.2828, 5.1599, 4.7527, 4.7810, 5.1510, 5.0303, 4.8465], device='cuda:0'), covar=tensor([0.0632, 0.0534, 0.0302, 0.0326, 0.1090, 0.0536, 0.0337, 0.0726], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0464, 0.0358, 0.0360, 0.0357, 0.0414, 0.0247, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:52:47,360 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281910.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 14:52:51,805 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281913.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:11,352 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 14:53:37,362 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=281945.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:53:48,943 INFO [train.py:904] (0/8) Epoch 28, batch 7900, loss[loss=0.1922, simple_loss=0.2808, pruned_loss=0.05181, over 15354.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2863, pruned_loss=0.05693, over 3051669.60 frames. ], batch size: 190, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:54:16,292 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=281971.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 14:54:29,914 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.540e+02 2.510e+02 2.970e+02 3.651e+02 5.631e+02, threshold=5.940e+02, percent-clipped=0.0 2023-05-02 14:54:42,284 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6369, 1.6639, 2.2177, 2.5197, 2.5959, 2.8587, 1.7142, 2.8556], device='cuda:0'), covar=tensor([0.0217, 0.0680, 0.0353, 0.0352, 0.0329, 0.0218, 0.0770, 0.0169], device='cuda:0'), in_proj_covar=tensor([0.0196, 0.0198, 0.0185, 0.0191, 0.0207, 0.0165, 0.0202, 0.0165], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:55:03,481 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-282000.pt 2023-05-02 14:55:09,797 INFO [train.py:904] (0/8) Epoch 28, batch 7950, loss[loss=0.2, simple_loss=0.2848, pruned_loss=0.05758, over 16804.00 frames. ], tot_loss[loss=0.2, simple_loss=0.2863, pruned_loss=0.05686, over 3059607.83 frames. ], batch size: 116, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 14:55:14,773 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282006.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:55:25,781 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6242, 2.5823, 1.9472, 2.6729, 2.1864, 2.7818, 2.1517, 2.4082], device='cuda:0'), covar=tensor([0.0326, 0.0341, 0.1189, 0.0291, 0.0609, 0.0487, 0.1227, 0.0610], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0181, 0.0197, 0.0173, 0.0181, 0.0221, 0.0206, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 14:55:33,439 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6895, 4.9414, 5.1150, 4.8586, 4.9802, 5.4857, 4.8702, 4.6235], device='cuda:0'), covar=tensor([0.1138, 0.1703, 0.2028, 0.1891, 0.2145, 0.0875, 0.1561, 0.2340], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0635, 0.0699, 0.0514, 0.0689, 0.0723, 0.0545, 0.0691], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 14:56:27,152 INFO [train.py:904] (0/8) Epoch 28, batch 8000, loss[loss=0.2602, simple_loss=0.3225, pruned_loss=0.09893, over 11201.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2871, pruned_loss=0.05772, over 3056979.01 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:07,769 INFO [optim.py:368] (0/8) 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 14:57:11,257 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282082.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:33,358 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282096.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:57:42,454 INFO [train.py:904] (0/8) Epoch 28, batch 8050, loss[loss=0.1926, simple_loss=0.2827, pruned_loss=0.05123, over 17174.00 frames. ], tot_loss[loss=0.2013, simple_loss=0.2873, pruned_loss=0.05763, over 3036386.03 frames. ], batch size: 46, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:57:53,995 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0280, 2.2489, 2.2641, 3.6568, 2.1627, 2.5490, 2.3410, 2.3992], device='cuda:0'), covar=tensor([0.1508, 0.3531, 0.3095, 0.0625, 0.4209, 0.2388, 0.3539, 0.3350], device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0475, 0.0385, 0.0337, 0.0446, 0.0542, 0.0448, 0.0555], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 14:57:56,113 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282112.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:22,177 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:58:57,924 INFO [train.py:904] (0/8) Epoch 28, batch 8100, loss[loss=0.1942, simple_loss=0.2827, pruned_loss=0.05284, over 16781.00 frames. ], tot_loss[loss=0.201, simple_loss=0.2874, pruned_loss=0.05727, over 3046928.65 frames. ], batch size: 83, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 14:59:03,727 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282157.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:08,427 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282160.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 14:59:38,415 INFO [optim.py:368] (0/8) 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 15:00:13,683 INFO [train.py:904] (0/8) Epoch 28, batch 8150, loss[loss=0.1955, simple_loss=0.2721, pruned_loss=0.05947, over 11762.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2849, pruned_loss=0.05626, over 3051488.70 frames. ], batch size: 247, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:00:21,757 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282208.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:00:46,932 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 15:01:29,255 INFO [train.py:904] (0/8) Epoch 28, batch 8200, loss[loss=0.1959, simple_loss=0.281, pruned_loss=0.05541, over 16906.00 frames. ], tot_loss[loss=0.1967, simple_loss=0.2824, pruned_loss=0.05548, over 3075730.17 frames. ], batch size: 109, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:01:30,194 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 15:01:40,776 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8386, 3.8794, 4.1576, 4.1221, 4.1141, 3.9009, 3.8984, 3.9546], device='cuda:0'), covar=tensor([0.0364, 0.0748, 0.0378, 0.0418, 0.0487, 0.0487, 0.0839, 0.0465], device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0487, 0.0470, 0.0434, 0.0517, 0.0495, 0.0572, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 15:01:50,075 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282266.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:02:13,183 INFO [optim.py:368] (0/8) 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 15:02:42,270 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6798, 3.7828, 2.8893, 2.2449, 2.3639, 2.4277, 3.9982, 3.2717], device='cuda:0'), covar=tensor([0.2946, 0.0517, 0.1829, 0.3167, 0.2913, 0.2260, 0.0385, 0.1387], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0273, 0.0313, 0.0328, 0.0306, 0.0278, 0.0305, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 15:02:46,949 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282301.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:02:50,332 INFO [train.py:904] (0/8) Epoch 28, batch 8250, loss[loss=0.1651, simple_loss=0.264, pruned_loss=0.03306, over 16782.00 frames. ], tot_loss[loss=0.1929, simple_loss=0.2807, pruned_loss=0.05252, over 3068279.93 frames. ], batch size: 89, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:03:30,351 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8592, 4.2626, 3.2362, 2.4071, 2.7567, 2.6788, 4.5576, 3.6326], device='cuda:0'), covar=tensor([0.2948, 0.0455, 0.1771, 0.3233, 0.2857, 0.2265, 0.0336, 0.1297], device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0272, 0.0312, 0.0327, 0.0305, 0.0277, 0.0304, 0.0350], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 15:04:07,382 INFO [train.py:904] (0/8) Epoch 28, batch 8300, loss[loss=0.1882, simple_loss=0.2926, pruned_loss=0.04191, over 16240.00 frames. ], tot_loss[loss=0.1884, simple_loss=0.2778, pruned_loss=0.04952, over 3047919.31 frames. ], batch size: 165, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:04:50,801 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.600e+02 2.126e+02 2.473e+02 2.954e+02 5.391e+02, threshold=4.946e+02, percent-clipped=0.0 2023-05-02 15:04:56,439 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0394, 4.1409, 3.9538, 3.6583, 3.6904, 4.0537, 3.7212, 3.8235], device='cuda:0'), covar=tensor([0.0593, 0.0663, 0.0334, 0.0324, 0.0659, 0.0508, 0.1164, 0.0631], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0465, 0.0359, 0.0359, 0.0356, 0.0414, 0.0248, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:05:26,440 INFO [train.py:904] (0/8) Epoch 28, batch 8350, loss[loss=0.1952, simple_loss=0.2758, pruned_loss=0.05731, over 11806.00 frames. ], tot_loss[loss=0.186, simple_loss=0.2767, pruned_loss=0.04764, over 3025333.66 frames. ], batch size: 246, lr: 2.39e-03, grad_scale: 4.0 2023-05-02 15:05:46,121 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0660, 4.1432, 3.9836, 3.6833, 3.7500, 4.0689, 3.7324, 3.8290], device='cuda:0'), covar=tensor([0.0593, 0.0705, 0.0307, 0.0309, 0.0700, 0.0563, 0.1129, 0.0655], device='cuda:0'), in_proj_covar=tensor([0.0306, 0.0465, 0.0358, 0.0359, 0.0355, 0.0414, 0.0247, 0.0428], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:05:58,005 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.14 vs. limit=2.0 2023-05-02 15:06:00,678 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9039, 2.6778, 2.5515, 1.9430, 2.5312, 2.7130, 2.6413, 1.9773], device='cuda:0'), covar=tensor([0.0409, 0.0097, 0.0093, 0.0363, 0.0149, 0.0137, 0.0118, 0.0446], device='cuda:0'), in_proj_covar=tensor([0.0136, 0.0088, 0.0090, 0.0133, 0.0100, 0.0114, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 15:06:43,095 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282452.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:06:43,825 INFO [train.py:904] (0/8) Epoch 28, batch 8400, loss[loss=0.1751, simple_loss=0.2715, pruned_loss=0.03937, over 16655.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2743, pruned_loss=0.04569, over 3020544.85 frames. ], batch size: 134, lr: 2.39e-03, grad_scale: 8.0 2023-05-02 15:07:26,991 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.421e+02 2.221e+02 2.637e+02 3.332e+02 6.864e+02, threshold=5.273e+02, percent-clipped=5.0 2023-05-02 15:07:40,031 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282489.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:08:02,092 INFO [train.py:904] (0/8) Epoch 28, batch 8450, loss[loss=0.1623, simple_loss=0.2572, pruned_loss=0.03375, over 15445.00 frames. ], tot_loss[loss=0.1801, simple_loss=0.2722, pruned_loss=0.04398, over 3022584.98 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:08:11,412 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282508.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:08:42,239 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7466, 3.9648, 3.0638, 2.2794, 2.4203, 2.5831, 4.2270, 3.4505], device='cuda:0'), covar=tensor([0.2913, 0.0526, 0.1736, 0.3260, 0.3188, 0.2228, 0.0335, 0.1273], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0269, 0.0308, 0.0323, 0.0301, 0.0274, 0.0300, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 15:09:17,249 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282550.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:17,590 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.96 vs. limit=2.0 2023-05-02 15:09:21,860 INFO [train.py:904] (0/8) Epoch 28, batch 8500, loss[loss=0.1608, simple_loss=0.2444, pruned_loss=0.03859, over 11844.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2688, pruned_loss=0.04185, over 3023402.03 frames. ], batch size: 247, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:09:27,450 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:09:43,367 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282566.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:10:07,072 INFO [optim.py:368] (0/8) 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 15:10:21,612 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9543, 2.3717, 2.0985, 2.1348, 2.6575, 2.3543, 2.3958, 2.8384], device='cuda:0'), covar=tensor([0.0229, 0.0439, 0.0560, 0.0500, 0.0327, 0.0442, 0.0301, 0.0298], device='cuda:0'), in_proj_covar=tensor([0.0221, 0.0236, 0.0227, 0.0228, 0.0237, 0.0236, 0.0233, 0.0235], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:10:42,369 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282601.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:10:42,631 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 15:10:45,539 INFO [train.py:904] (0/8) Epoch 28, batch 8550, loss[loss=0.1723, simple_loss=0.255, pruned_loss=0.04486, over 11764.00 frames. ], tot_loss[loss=0.1748, simple_loss=0.2671, pruned_loss=0.0413, over 3002719.93 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:11:05,482 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282614.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:11:21,502 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 15:11:37,469 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8673, 3.7618, 3.9199, 4.0047, 4.0945, 3.6828, 4.0679, 4.1180], device='cuda:0'), covar=tensor([0.1507, 0.1081, 0.1180, 0.0701, 0.0567, 0.1717, 0.0676, 0.0654], device='cuda:0'), in_proj_covar=tensor([0.0654, 0.0802, 0.0926, 0.0815, 0.0622, 0.0649, 0.0681, 0.0790], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:12:14,469 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:12:21,570 INFO [train.py:904] (0/8) Epoch 28, batch 8600, loss[loss=0.1786, simple_loss=0.2611, pruned_loss=0.0481, over 12463.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2678, pruned_loss=0.04039, over 3003361.78 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:12:32,774 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282658.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:11,863 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282677.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:13:11,971 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6601, 2.3971, 2.3280, 3.7661, 2.0765, 3.7432, 1.4198, 2.8249], device='cuda:0'), covar=tensor([0.1425, 0.0863, 0.1267, 0.0159, 0.0094, 0.0336, 0.1844, 0.0783], device='cuda:0'), in_proj_covar=tensor([0.0170, 0.0178, 0.0198, 0.0198, 0.0203, 0.0215, 0.0207, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 15:13:17,330 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.534e+02 2.220e+02 2.463e+02 3.072e+02 6.042e+02, threshold=4.927e+02, percent-clipped=1.0 2023-05-02 15:13:50,079 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282698.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:13:58,793 INFO [train.py:904] (0/8) Epoch 28, batch 8650, loss[loss=0.158, simple_loss=0.2651, pruned_loss=0.02545, over 16886.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2663, pruned_loss=0.03884, over 3013403.41 frames. ], batch size: 102, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:14:36,268 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282738.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:40,178 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=282752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:15:40,931 INFO [train.py:904] (0/8) Epoch 28, batch 8700, loss[loss=0.1557, simple_loss=0.2531, pruned_loss=0.02911, over 16153.00 frames. ], tot_loss[loss=0.1694, simple_loss=0.2634, pruned_loss=0.03767, over 3018894.90 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:15:53,204 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282759.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:15:56,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4499, 2.8427, 3.2275, 2.0720, 2.9018, 2.2118, 3.0984, 3.1026], device='cuda:0'), covar=tensor([0.0259, 0.1000, 0.0528, 0.2117, 0.0747, 0.0993, 0.0633, 0.0937], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0166, 0.0166, 0.0154, 0.0144, 0.0129, 0.0142, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 15:16:29,665 INFO [optim.py:368] (0/8) 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 15:17:06,891 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=282800.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:17:12,016 INFO [train.py:904] (0/8) Epoch 28, batch 8750, loss[loss=0.1869, simple_loss=0.2873, pruned_loss=0.04321, over 15253.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2638, pruned_loss=0.03762, over 3027535.35 frames. ], batch size: 190, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:17:18,920 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=282805.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:18:46,456 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=282845.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:19:02,838 INFO [train.py:904] (0/8) Epoch 28, batch 8800, loss[loss=0.1814, simple_loss=0.2705, pruned_loss=0.04615, over 12327.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2618, pruned_loss=0.03631, over 3041011.90 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:19:29,180 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=282866.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:20:00,726 INFO [optim.py:368] (0/8) 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 15:20:47,898 INFO [train.py:904] (0/8) Epoch 28, batch 8850, loss[loss=0.1411, simple_loss=0.2382, pruned_loss=0.02205, over 12711.00 frames. ], tot_loss[loss=0.1675, simple_loss=0.2637, pruned_loss=0.03566, over 3019971.02 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:22:03,967 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6797, 2.6567, 1.9495, 2.7978, 2.1302, 2.8305, 2.1398, 2.4145], device='cuda:0'), covar=tensor([0.0339, 0.0378, 0.1399, 0.0316, 0.0747, 0.0526, 0.1337, 0.0686], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0176, 0.0192, 0.0167, 0.0176, 0.0214, 0.0200, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 15:22:16,061 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 15:22:35,080 INFO [train.py:904] (0/8) Epoch 28, batch 8900, loss[loss=0.1464, simple_loss=0.2542, pruned_loss=0.01926, over 16857.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2651, pruned_loss=0.03515, over 3045093.09 frames. ], batch size: 96, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:23:38,049 INFO [optim.py:368] (0/8) 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 15:24:31,801 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8616, 4.9173, 4.7614, 4.3482, 4.4499, 4.8193, 4.6191, 4.5135], device='cuda:0'), covar=tensor([0.0589, 0.0599, 0.0315, 0.0335, 0.0905, 0.0545, 0.0414, 0.0760], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0457, 0.0354, 0.0354, 0.0349, 0.0408, 0.0244, 0.0421], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:24:39,877 INFO [train.py:904] (0/8) Epoch 28, batch 8950, loss[loss=0.1525, simple_loss=0.2452, pruned_loss=0.02992, over 15353.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2641, pruned_loss=0.03507, over 3048853.49 frames. ], batch size: 191, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:25:04,568 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283014.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:25:21,478 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5050, 3.2628, 3.5814, 1.8521, 3.7791, 3.8390, 3.0425, 2.9372], device='cuda:0'), covar=tensor([0.0747, 0.0314, 0.0255, 0.1279, 0.0084, 0.0161, 0.0417, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0108, 0.0099, 0.0135, 0.0083, 0.0126, 0.0126, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 15:25:45,377 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283033.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:26:27,364 INFO [train.py:904] (0/8) Epoch 28, batch 9000, loss[loss=0.1532, simple_loss=0.2495, pruned_loss=0.02842, over 15203.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2608, pruned_loss=0.03381, over 3058750.81 frames. ], batch size: 190, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:26:27,365 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 15:26:38,047 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 15:26:41,529 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283054.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 15:27:36,924 INFO [optim.py:368] (0/8) 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 15:28:21,101 INFO [train.py:904] (0/8) Epoch 28, batch 9050, loss[loss=0.1634, simple_loss=0.2488, pruned_loss=0.039, over 16619.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2623, pruned_loss=0.03457, over 3078418.24 frames. ], batch size: 134, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:28:48,288 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:46,404 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:29:48,404 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283145.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:04,167 INFO [train.py:904] (0/8) Epoch 28, batch 9100, loss[loss=0.1555, simple_loss=0.2569, pruned_loss=0.02708, over 16730.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2618, pruned_loss=0.03519, over 3070784.59 frames. ], batch size: 76, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:30:20,084 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283161.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:30:57,971 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:31:08,549 INFO [optim.py:368] (0/8) 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 15:31:21,213 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3581, 3.5087, 3.7292, 1.7737, 3.9351, 4.0870, 3.1613, 2.7862], device='cuda:0'), covar=tensor([0.1089, 0.0227, 0.0206, 0.1428, 0.0084, 0.0143, 0.0428, 0.0626], device='cuda:0'), in_proj_covar=tensor([0.0144, 0.0107, 0.0098, 0.0135, 0.0083, 0.0126, 0.0125, 0.0127], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 15:31:38,533 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283193.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:32:01,018 INFO [train.py:904] (0/8) Epoch 28, batch 9150, loss[loss=0.1506, simple_loss=0.2486, pruned_loss=0.02633, over 16863.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2625, pruned_loss=0.03499, over 3072394.41 frames. ], batch size: 96, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:32:06,384 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283205.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:32:11,104 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2451, 4.3803, 4.5156, 4.2553, 4.3843, 4.8381, 4.3814, 4.0863], device='cuda:0'), covar=tensor([0.1768, 0.1849, 0.1780, 0.2196, 0.2479, 0.0980, 0.1551, 0.2568], device='cuda:0'), in_proj_covar=tensor([0.0414, 0.0614, 0.0677, 0.0498, 0.0668, 0.0706, 0.0529, 0.0669], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 15:33:44,284 INFO [train.py:904] (0/8) Epoch 28, batch 9200, loss[loss=0.1482, simple_loss=0.2427, pruned_loss=0.0269, over 12108.00 frames. ], tot_loss[loss=0.163, simple_loss=0.258, pruned_loss=0.03403, over 3063500.27 frames. ], batch size: 247, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:34:16,696 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0938, 3.9806, 4.1491, 4.2591, 4.3874, 3.9784, 4.3681, 4.4095], device='cuda:0'), covar=tensor([0.1849, 0.1136, 0.1475, 0.0832, 0.0619, 0.1423, 0.0754, 0.0781], device='cuda:0'), in_proj_covar=tensor([0.0644, 0.0788, 0.0910, 0.0805, 0.0612, 0.0639, 0.0673, 0.0777], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:34:34,273 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.493e+02 2.118e+02 2.676e+02 3.077e+02 8.618e+02, threshold=5.351e+02, percent-clipped=1.0 2023-05-02 15:34:39,311 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:35:20,281 INFO [train.py:904] (0/8) Epoch 28, batch 9250, loss[loss=0.1548, simple_loss=0.2504, pruned_loss=0.02956, over 16193.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2575, pruned_loss=0.03411, over 3048553.12 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:35:44,247 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283314.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:28,207 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:36:34,462 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6461, 3.6830, 2.8018, 2.2354, 2.2929, 2.4337, 3.8794, 3.3126], device='cuda:0'), covar=tensor([0.3011, 0.0596, 0.1888, 0.3222, 0.3100, 0.2385, 0.0402, 0.1390], device='cuda:0'), in_proj_covar=tensor([0.0330, 0.0267, 0.0307, 0.0321, 0.0297, 0.0272, 0.0298, 0.0344], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 15:36:55,392 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 28, batch 9300, loss[loss=0.1599, simple_loss=0.2505, pruned_loss=0.03466, over 16639.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2565, pruned_loss=0.03376, over 3041066.16 frames. ], batch size: 148, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:37:16,900 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283354.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:37:34,851 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283362.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:16,510 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283381.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:38:47,476 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7031, 4.9919, 4.8014, 4.8075, 4.5372, 4.5709, 4.4168, 5.0615], device='cuda:0'), covar=tensor([0.1163, 0.0855, 0.0909, 0.0789, 0.0776, 0.1096, 0.1249, 0.0832], device='cuda:0'), in_proj_covar=tensor([0.0694, 0.0839, 0.0688, 0.0649, 0.0530, 0.0533, 0.0698, 0.0654], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:38:58,072 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283402.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:38:59,597 INFO [train.py:904] (0/8) Epoch 28, batch 9350, loss[loss=0.1728, simple_loss=0.2631, pruned_loss=0.04123, over 16886.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2565, pruned_loss=0.03367, over 3072914.71 frames. ], batch size: 116, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:39:51,008 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.54 vs. limit=2.0 2023-05-02 15:40:00,549 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.70 vs. limit=2.0 2023-05-02 15:40:41,032 INFO [train.py:904] (0/8) Epoch 28, batch 9400, loss[loss=0.145, simple_loss=0.2329, pruned_loss=0.02856, over 12789.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2562, pruned_loss=0.0333, over 3068176.81 frames. ], batch size: 248, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:41:00,128 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283461.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:11,273 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.73 vs. limit=2.0 2023-05-02 15:41:18,942 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283471.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:41:39,549 INFO [optim.py:368] (0/8) 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 15:41:47,360 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 15:42:20,566 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283500.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:42:24,720 INFO [train.py:904] (0/8) Epoch 28, batch 9450, loss[loss=0.1593, simple_loss=0.2522, pruned_loss=0.03325, over 16536.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.258, pruned_loss=0.03357, over 3054465.76 frames. ], batch size: 68, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:42:36,256 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283509.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:43:49,001 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283544.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 15:44:06,800 INFO [train.py:904] (0/8) Epoch 28, batch 9500, loss[loss=0.1447, simple_loss=0.2435, pruned_loss=0.02294, over 16917.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2566, pruned_loss=0.03304, over 3055429.54 frames. ], batch size: 96, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:44:24,821 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8924, 5.1855, 5.3113, 5.0999, 5.1437, 5.6984, 5.1765, 4.9224], device='cuda:0'), covar=tensor([0.0952, 0.1896, 0.2462, 0.1861, 0.2237, 0.0840, 0.1593, 0.2329], device='cuda:0'), in_proj_covar=tensor([0.0409, 0.0611, 0.0675, 0.0496, 0.0665, 0.0702, 0.0526, 0.0665], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 15:45:03,874 INFO [optim.py:368] (0/8) 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 15:45:53,239 INFO [train.py:904] (0/8) Epoch 28, batch 9550, loss[loss=0.1586, simple_loss=0.2554, pruned_loss=0.03093, over 16836.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2563, pruned_loss=0.03318, over 3067157.22 frames. ], batch size: 76, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:45:58,568 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283605.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:47:09,807 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283639.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:47:34,680 INFO [train.py:904] (0/8) Epoch 28, batch 9600, loss[loss=0.1655, simple_loss=0.254, pruned_loss=0.03848, over 12429.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2575, pruned_loss=0.03364, over 3074990.17 frames. ], batch size: 249, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:47:35,184 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9748, 4.2602, 4.1299, 4.1199, 3.7822, 3.8819, 3.8892, 4.2830], device='cuda:0'), covar=tensor([0.1223, 0.0997, 0.0936, 0.0809, 0.0825, 0.1722, 0.1054, 0.0880], device='cuda:0'), in_proj_covar=tensor([0.0694, 0.0840, 0.0685, 0.0649, 0.0530, 0.0532, 0.0698, 0.0653], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:48:02,100 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4645, 3.4144, 3.5115, 3.5752, 3.6120, 3.3274, 3.5788, 3.6522], device='cuda:0'), covar=tensor([0.1347, 0.0955, 0.1021, 0.0664, 0.0610, 0.2610, 0.1012, 0.0831], device='cuda:0'), in_proj_covar=tensor([0.0638, 0.0780, 0.0900, 0.0795, 0.0605, 0.0631, 0.0665, 0.0769], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:48:04,361 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.73 vs. limit=5.0 2023-05-02 15:48:29,447 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.462e+02 2.192e+02 2.621e+02 3.097e+02 5.622e+02, threshold=5.243e+02, percent-clipped=5.0 2023-05-02 15:48:55,400 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.89 vs. limit=5.0 2023-05-02 15:49:23,012 INFO [train.py:904] (0/8) Epoch 28, batch 9650, loss[loss=0.1622, simple_loss=0.2617, pruned_loss=0.03135, over 15437.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2591, pruned_loss=0.03382, over 3068044.83 frames. ], batch size: 190, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:50:39,261 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6880, 2.6365, 2.5012, 4.1180, 2.1594, 3.8895, 1.5754, 2.8653], device='cuda:0'), covar=tensor([0.1770, 0.0952, 0.1342, 0.0201, 0.0169, 0.0402, 0.2085, 0.0863], device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0175, 0.0195, 0.0195, 0.0198, 0.0211, 0.0205, 0.0194], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 15:51:09,580 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7911, 4.7575, 4.5154, 3.8927, 4.6563, 1.8431, 4.4517, 4.2940], device='cuda:0'), covar=tensor([0.0118, 0.0131, 0.0230, 0.0302, 0.0114, 0.2774, 0.0141, 0.0270], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0168, 0.0205, 0.0176, 0.0182, 0.0211, 0.0193, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:51:10,213 INFO [train.py:904] (0/8) Epoch 28, batch 9700, loss[loss=0.1695, simple_loss=0.2725, pruned_loss=0.03329, over 16896.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2581, pruned_loss=0.03362, over 3060942.70 frames. ], batch size: 102, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:51:47,264 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283771.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:08,996 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.467e+02 2.180e+02 2.423e+02 2.886e+02 5.203e+02, threshold=4.846e+02, percent-clipped=0.0 2023-05-02 15:52:40,457 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6059, 3.5479, 3.5284, 2.8401, 3.4361, 2.0404, 3.3090, 2.9440], device='cuda:0'), covar=tensor([0.0167, 0.0150, 0.0194, 0.0245, 0.0121, 0.2438, 0.0160, 0.0257], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0168, 0.0205, 0.0177, 0.0182, 0.0212, 0.0193, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0001, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:52:48,162 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283800.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:52:52,814 INFO [train.py:904] (0/8) Epoch 28, batch 9750, loss[loss=0.1533, simple_loss=0.243, pruned_loss=0.03178, over 16580.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2568, pruned_loss=0.03388, over 3051193.12 frames. ], batch size: 62, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:52:56,852 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5913, 4.6949, 4.4808, 4.1053, 4.2002, 4.6143, 4.3933, 4.2705], device='cuda:0'), covar=tensor([0.0619, 0.0526, 0.0343, 0.0360, 0.0937, 0.0447, 0.0406, 0.0759], device='cuda:0'), in_proj_covar=tensor([0.0299, 0.0449, 0.0350, 0.0350, 0.0345, 0.0404, 0.0241, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:53:24,762 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283819.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:12,846 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283842.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:24,074 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283848.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:54:31,914 INFO [train.py:904] (0/8) Epoch 28, batch 9800, loss[loss=0.1812, simple_loss=0.2846, pruned_loss=0.0389, over 15524.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2575, pruned_loss=0.03323, over 3064554.23 frames. ], batch size: 190, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:55:23,052 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.531e+02 2.061e+02 2.416e+02 3.000e+02 5.773e+02, threshold=4.831e+02, percent-clipped=3.0 2023-05-02 15:56:10,710 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=283900.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 15:56:14,718 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0007, 4.2388, 4.0868, 4.1171, 3.7907, 3.8611, 3.8486, 4.2478], device='cuda:0'), covar=tensor([0.1122, 0.0918, 0.0952, 0.0850, 0.0804, 0.1787, 0.1061, 0.0986], device='cuda:0'), in_proj_covar=tensor([0.0691, 0.0836, 0.0682, 0.0646, 0.0528, 0.0529, 0.0696, 0.0651], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:56:15,645 INFO [train.py:904] (0/8) Epoch 28, batch 9850, loss[loss=0.1455, simple_loss=0.2372, pruned_loss=0.02692, over 12170.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2587, pruned_loss=0.03305, over 3068464.64 frames. ], batch size: 246, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:56:17,406 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=283903.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:56:50,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8962, 2.6509, 2.8810, 2.0742, 2.6491, 2.1484, 2.6838, 2.8872], device='cuda:0'), covar=tensor([0.0314, 0.0992, 0.0543, 0.1953, 0.0842, 0.0994, 0.0612, 0.0785], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0163, 0.0166, 0.0153, 0.0143, 0.0129, 0.0141, 0.0176], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 15:57:36,918 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=283939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:58:06,909 INFO [train.py:904] (0/8) Epoch 28, batch 9900, loss[loss=0.167, simple_loss=0.2718, pruned_loss=0.03111, over 16419.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.259, pruned_loss=0.03343, over 3036844.43 frames. ], batch size: 146, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 15:58:13,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2254, 2.3281, 2.1164, 2.2070, 2.6921, 2.4463, 2.6056, 2.9319], device='cuda:0'), covar=tensor([0.0210, 0.0535, 0.0626, 0.0549, 0.0370, 0.0487, 0.0275, 0.0297], device='cuda:0'), in_proj_covar=tensor([0.0219, 0.0237, 0.0228, 0.0228, 0.0237, 0.0237, 0.0230, 0.0234], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 15:58:38,049 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=283966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:58:44,131 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8090, 2.7583, 2.6272, 1.8581, 2.4920, 2.7902, 2.6400, 1.8605], device='cuda:0'), covar=tensor([0.0535, 0.0087, 0.0092, 0.0419, 0.0182, 0.0108, 0.0117, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0134, 0.0086, 0.0087, 0.0131, 0.0099, 0.0110, 0.0094, 0.0126], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 15:59:13,296 INFO [optim.py:368] (0/8) 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 15:59:30,053 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=283987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 15:59:58,378 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-284000.pt 2023-05-02 16:00:07,769 INFO [train.py:904] (0/8) Epoch 28, batch 9950, loss[loss=0.1533, simple_loss=0.2596, pruned_loss=0.02354, over 16923.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2611, pruned_loss=0.03346, over 3045785.37 frames. ], batch size: 96, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:01:07,762 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:02:08,469 INFO [train.py:904] (0/8) Epoch 28, batch 10000, loss[loss=0.1547, simple_loss=0.2586, pruned_loss=0.02538, over 15547.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.26, pruned_loss=0.03323, over 3065930.93 frames. ], batch size: 192, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:03:03,891 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.562e+02 2.120e+02 2.490e+02 3.242e+02 7.022e+02, threshold=4.980e+02, percent-clipped=2.0 2023-05-02 16:03:50,393 INFO [train.py:904] (0/8) Epoch 28, batch 10050, loss[loss=0.1752, simple_loss=0.2763, pruned_loss=0.03699, over 16073.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2601, pruned_loss=0.03364, over 3038792.05 frames. ], batch size: 165, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:03:54,489 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2191, 4.2555, 4.5593, 4.5375, 4.5404, 4.2846, 4.2802, 4.2616], device='cuda:0'), covar=tensor([0.0329, 0.0530, 0.0390, 0.0383, 0.0431, 0.0341, 0.0842, 0.0404], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0469, 0.0455, 0.0419, 0.0502, 0.0479, 0.0549, 0.0386], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 16:05:24,434 INFO [train.py:904] (0/8) Epoch 28, batch 10100, loss[loss=0.1562, simple_loss=0.2505, pruned_loss=0.03096, over 16632.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2606, pruned_loss=0.03382, over 3037262.82 frames. ], batch size: 76, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:05:52,251 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.83 vs. limit=5.0 2023-05-02 16:06:20,722 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.509e+02 2.222e+02 2.734e+02 3.195e+02 6.018e+02, threshold=5.468e+02, percent-clipped=8.0 2023-05-02 16:06:39,386 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284198.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:06:42,047 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284200.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:06:44,470 INFO [train.py:904] (0/8) Epoch 28, batch 10150, loss[loss=0.1439, simple_loss=0.2346, pruned_loss=0.02665, over 12446.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2594, pruned_loss=0.03374, over 3033355.54 frames. ], batch size: 250, lr: 2.38e-03, grad_scale: 8.0 2023-05-02 16:06:46,553 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-28.pt 2023-05-02 16:07:10,316 INFO [train.py:904] (0/8) Epoch 29, batch 0, loss[loss=0.2267, simple_loss=0.2939, pruned_loss=0.0797, over 16359.00 frames. ], tot_loss[loss=0.2267, simple_loss=0.2939, pruned_loss=0.0797, over 16359.00 frames. ], batch size: 146, lr: 2.34e-03, grad_scale: 8.0 2023-05-02 16:07:10,317 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 16:07:17,746 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 16:08:18,233 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284248.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:08:26,902 INFO [train.py:904] (0/8) Epoch 29, batch 50, loss[loss=0.1711, simple_loss=0.2507, pruned_loss=0.04578, over 16709.00 frames. ], tot_loss[loss=0.1767, simple_loss=0.264, pruned_loss=0.04473, over 751852.13 frames. ], batch size: 124, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:08:54,132 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.45 vs. limit=2.0 2023-05-02 16:09:08,279 INFO [optim.py:368] (0/8) 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 16:09:24,120 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1535, 4.8543, 5.0708, 5.2789, 5.5123, 4.8105, 5.4406, 5.4940], device='cuda:0'), covar=tensor([0.2040, 0.1431, 0.2031, 0.0984, 0.0652, 0.0970, 0.0731, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0645, 0.0787, 0.0907, 0.0802, 0.0610, 0.0636, 0.0673, 0.0778], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:09:37,389 INFO [train.py:904] (0/8) Epoch 29, batch 100, loss[loss=0.1667, simple_loss=0.2728, pruned_loss=0.0303, over 17215.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2609, pruned_loss=0.04309, over 1314298.89 frames. ], batch size: 52, lr: 2.34e-03, grad_scale: 2.0 2023-05-02 16:10:02,069 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284322.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:10:16,001 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5977, 2.5040, 2.4655, 4.5187, 2.5358, 2.8620, 2.5629, 2.7138], device='cuda:0'), covar=tensor([0.1303, 0.3715, 0.3265, 0.0489, 0.4007, 0.2637, 0.3807, 0.3678], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0466, 0.0382, 0.0328, 0.0440, 0.0532, 0.0439, 0.0544], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:10:46,115 INFO [train.py:904] (0/8) Epoch 29, batch 150, loss[loss=0.1657, simple_loss=0.2448, pruned_loss=0.04328, over 16988.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2585, pruned_loss=0.04093, over 1769529.87 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:11:22,713 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-05-02 16:11:25,628 INFO [optim.py:368] (0/8) 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 16:11:48,921 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6670, 3.7613, 2.3578, 4.2218, 2.8746, 4.1771, 2.5315, 3.1841], device='cuda:0'), covar=tensor([0.0353, 0.0461, 0.1723, 0.0416, 0.0933, 0.0593, 0.1539, 0.0797], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0179, 0.0196, 0.0170, 0.0179, 0.0217, 0.0204, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 16:11:55,144 INFO [train.py:904] (0/8) Epoch 29, batch 200, loss[loss=0.1736, simple_loss=0.2671, pruned_loss=0.0401, over 16713.00 frames. ], tot_loss[loss=0.1708, simple_loss=0.2596, pruned_loss=0.04099, over 2117664.94 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:12:51,557 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284443.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:13:04,677 INFO [train.py:904] (0/8) Epoch 29, batch 250, loss[loss=0.1728, simple_loss=0.2703, pruned_loss=0.03766, over 17029.00 frames. ], tot_loss[loss=0.1697, simple_loss=0.2576, pruned_loss=0.04088, over 2376285.66 frames. ], batch size: 55, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:13:11,643 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2062, 5.7953, 5.8934, 5.5984, 5.7011, 6.2871, 5.7462, 5.4794], device='cuda:0'), covar=tensor([0.0865, 0.2095, 0.2591, 0.2197, 0.2516, 0.0926, 0.1568, 0.2272], device='cuda:0'), in_proj_covar=tensor([0.0417, 0.0625, 0.0693, 0.0510, 0.0680, 0.0716, 0.0538, 0.0679], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 16:13:47,581 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.584e+02 2.177e+02 2.500e+02 3.084e+02 6.085e+02, threshold=5.001e+02, percent-clipped=1.0 2023-05-02 16:13:54,869 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7139, 6.1050, 5.8230, 5.8891, 5.4621, 5.5476, 5.5251, 6.2395], device='cuda:0'), covar=tensor([0.1448, 0.0977, 0.1217, 0.0913, 0.0932, 0.0608, 0.1319, 0.0913], device='cuda:0'), in_proj_covar=tensor([0.0698, 0.0847, 0.0693, 0.0655, 0.0535, 0.0534, 0.0709, 0.0662], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:14:08,412 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284498.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:17,001 INFO [train.py:904] (0/8) Epoch 29, batch 300, loss[loss=0.1536, simple_loss=0.2473, pruned_loss=0.02996, over 17217.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2552, pruned_loss=0.04012, over 2576000.32 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:14:17,400 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284504.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:14:26,611 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2876, 5.2632, 5.0800, 4.5838, 5.0661, 2.0739, 4.9031, 4.9899], device='cuda:0'), covar=tensor([0.0132, 0.0120, 0.0254, 0.0452, 0.0125, 0.2766, 0.0194, 0.0237], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0170, 0.0207, 0.0178, 0.0184, 0.0214, 0.0195, 0.0174], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:14:56,173 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.8190, 6.1939, 5.9167, 6.0049, 5.5263, 5.6736, 5.6659, 6.3173], device='cuda:0'), covar=tensor([0.1335, 0.1010, 0.1069, 0.0912, 0.0901, 0.0589, 0.1271, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0702, 0.0851, 0.0697, 0.0658, 0.0538, 0.0537, 0.0712, 0.0665], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:15:14,273 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284546.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:15:23,228 INFO [train.py:904] (0/8) Epoch 29, batch 350, loss[loss=0.1432, simple_loss=0.2275, pruned_loss=0.02947, over 16795.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2525, pruned_loss=0.03919, over 2740036.51 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:15:32,735 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6851, 4.5956, 4.6060, 4.2639, 4.3231, 4.6040, 4.4910, 4.3927], device='cuda:0'), covar=tensor([0.0646, 0.0921, 0.0377, 0.0381, 0.0914, 0.0597, 0.0456, 0.0725], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0461, 0.0359, 0.0359, 0.0353, 0.0414, 0.0246, 0.0428], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:16:02,936 INFO [optim.py:368] (0/8) 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 16:16:10,872 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5893, 4.8529, 4.6578, 4.6371, 4.4476, 4.3943, 4.3529, 4.9171], device='cuda:0'), covar=tensor([0.1177, 0.0909, 0.1053, 0.0915, 0.0748, 0.1394, 0.1254, 0.0912], device='cuda:0'), in_proj_covar=tensor([0.0704, 0.0854, 0.0699, 0.0661, 0.0540, 0.0538, 0.0714, 0.0667], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:16:31,998 INFO [train.py:904] (0/8) Epoch 29, batch 400, loss[loss=0.1716, simple_loss=0.2554, pruned_loss=0.04389, over 16848.00 frames. ], tot_loss[loss=0.1647, simple_loss=0.2509, pruned_loss=0.03929, over 2865715.86 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:16:57,020 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=284622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:17:41,219 INFO [train.py:904] (0/8) Epoch 29, batch 450, loss[loss=0.1751, simple_loss=0.2515, pruned_loss=0.04934, over 16890.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2489, pruned_loss=0.03833, over 2963079.67 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:18:02,996 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=284670.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:18:18,526 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.454e+02 1.978e+02 2.312e+02 2.858e+02 4.933e+02, threshold=4.623e+02, percent-clipped=2.0 2023-05-02 16:18:36,744 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9152, 2.5545, 2.4980, 4.0035, 3.2462, 3.9943, 1.6362, 2.9334], device='cuda:0'), covar=tensor([0.1420, 0.0787, 0.1279, 0.0229, 0.0145, 0.0396, 0.1694, 0.0866], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0179, 0.0199, 0.0201, 0.0203, 0.0216, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 16:18:47,404 INFO [train.py:904] (0/8) Epoch 29, batch 500, loss[loss=0.173, simple_loss=0.2478, pruned_loss=0.04916, over 16758.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2481, pruned_loss=0.03786, over 3048782.93 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:19:48,365 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 16:19:56,205 INFO [train.py:904] (0/8) Epoch 29, batch 550, loss[loss=0.1787, simple_loss=0.2607, pruned_loss=0.04833, over 16825.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2482, pruned_loss=0.03784, over 3114836.26 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:20:07,438 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6935, 2.8395, 3.1542, 2.0422, 2.8366, 2.1963, 3.2694, 3.2593], device='cuda:0'), covar=tensor([0.0263, 0.1080, 0.0655, 0.2082, 0.0930, 0.1094, 0.0573, 0.0978], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0169, 0.0170, 0.0158, 0.0148, 0.0133, 0.0145, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 16:20:35,026 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3000, 5.2758, 5.0263, 4.5256, 5.0589, 2.0660, 4.8636, 4.9007], device='cuda:0'), covar=tensor([0.0085, 0.0087, 0.0233, 0.0422, 0.0118, 0.2849, 0.0145, 0.0248], device='cuda:0'), in_proj_covar=tensor([0.0179, 0.0173, 0.0211, 0.0181, 0.0186, 0.0217, 0.0198, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:20:35,719 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.399e+02 1.886e+02 2.233e+02 2.634e+02 5.354e+02, threshold=4.465e+02, percent-clipped=1.0 2023-05-02 16:20:57,924 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=284799.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:21:04,188 INFO [train.py:904] (0/8) Epoch 29, batch 600, loss[loss=0.1577, simple_loss=0.2387, pruned_loss=0.03836, over 16693.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2472, pruned_loss=0.0377, over 3156837.07 frames. ], batch size: 76, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:21:21,367 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284816.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:21:26,747 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8164, 3.6537, 4.0232, 2.2347, 4.1052, 4.1518, 3.2863, 3.0936], device='cuda:0'), covar=tensor([0.0826, 0.0274, 0.0224, 0.1230, 0.0131, 0.0230, 0.0461, 0.0510], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0139, 0.0086, 0.0130, 0.0129, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 16:22:12,365 INFO [train.py:904] (0/8) Epoch 29, batch 650, loss[loss=0.1842, simple_loss=0.2627, pruned_loss=0.05289, over 15400.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2467, pruned_loss=0.0376, over 3194740.60 frames. ], batch size: 190, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:22:46,190 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=284877.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:22:53,843 INFO [optim.py:368] (0/8) 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 16:23:22,619 INFO [train.py:904] (0/8) Epoch 29, batch 700, loss[loss=0.1576, simple_loss=0.2493, pruned_loss=0.0329, over 17179.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2468, pruned_loss=0.03684, over 3236640.00 frames. ], batch size: 46, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:24:12,292 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284939.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:24:13,299 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2360, 5.7510, 5.8138, 5.5812, 5.6556, 6.2170, 5.6798, 5.4435], device='cuda:0'), covar=tensor([0.0853, 0.1928, 0.2632, 0.2259, 0.2738, 0.1011, 0.1599, 0.2249], device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0639, 0.0711, 0.0522, 0.0699, 0.0731, 0.0549, 0.0694], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 16:24:30,130 INFO [train.py:904] (0/8) Epoch 29, batch 750, loss[loss=0.1471, simple_loss=0.2266, pruned_loss=0.03382, over 15897.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2469, pruned_loss=0.0372, over 3256096.10 frames. ], batch size: 35, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:25:13,141 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.375e+02 2.091e+02 2.322e+02 2.629e+02 4.322e+02, threshold=4.644e+02, percent-clipped=0.0 2023-05-02 16:25:30,101 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=284996.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:25:37,980 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 29, batch 800, loss[loss=0.1456, simple_loss=0.2357, pruned_loss=0.02776, over 17244.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.246, pruned_loss=0.03652, over 3272614.19 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:26:33,707 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285042.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:26:49,076 INFO [train.py:904] (0/8) Epoch 29, batch 850, loss[loss=0.1546, simple_loss=0.2324, pruned_loss=0.03843, over 15458.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2459, pruned_loss=0.03673, over 3277661.95 frames. ], batch size: 190, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:26:53,042 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285057.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:31,501 INFO [optim.py:368] (0/8) 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 16:27:49,589 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:27:54,271 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285103.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 16:27:55,641 INFO [train.py:904] (0/8) Epoch 29, batch 900, loss[loss=0.1648, simple_loss=0.2539, pruned_loss=0.03783, over 16438.00 frames. ], tot_loss[loss=0.1583, simple_loss=0.2449, pruned_loss=0.03582, over 3288489.99 frames. ], batch size: 75, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:28:15,817 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.21 vs. limit=2.0 2023-05-02 16:28:34,049 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.40 vs. limit=5.0 2023-05-02 16:28:55,570 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285147.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:05,603 INFO [train.py:904] (0/8) Epoch 29, batch 950, loss[loss=0.1607, simple_loss=0.2413, pruned_loss=0.04012, over 16759.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.2443, pruned_loss=0.03592, over 3285117.69 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:29:30,491 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285172.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:29:47,172 INFO [optim.py:368] (0/8) 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 16:30:14,208 INFO [train.py:904] (0/8) Epoch 29, batch 1000, loss[loss=0.1695, simple_loss=0.2586, pruned_loss=0.04021, over 16641.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2433, pruned_loss=0.03566, over 3301568.95 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:31:02,846 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8296, 2.9233, 3.2200, 2.0316, 2.8486, 2.2325, 3.3830, 3.2947], device='cuda:0'), covar=tensor([0.0258, 0.1038, 0.0623, 0.2043, 0.0888, 0.1032, 0.0549, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 16:31:21,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3672, 3.6041, 4.0201, 2.1938, 3.1654, 2.5620, 3.7590, 3.8163], device='cuda:0'), covar=tensor([0.0285, 0.1019, 0.0464, 0.2171, 0.0905, 0.1011, 0.0624, 0.1136], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0171, 0.0170, 0.0158, 0.0148, 0.0133, 0.0146, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 16:31:24,212 INFO [train.py:904] (0/8) Epoch 29, batch 1050, loss[loss=0.1466, simple_loss=0.2345, pruned_loss=0.02937, over 16860.00 frames. ], tot_loss[loss=0.1574, simple_loss=0.2429, pruned_loss=0.03593, over 3311496.54 frames. ], batch size: 42, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:32:05,304 INFO [optim.py:368] (0/8) 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 16:32:19,765 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 29, batch 1100, loss[loss=0.1439, simple_loss=0.2435, pruned_loss=0.02217, over 17252.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.243, pruned_loss=0.03567, over 3315241.46 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:33:37,336 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285352.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:33:37,738 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 16:33:40,187 INFO [train.py:904] (0/8) Epoch 29, batch 1150, loss[loss=0.1406, simple_loss=0.2266, pruned_loss=0.02726, over 16782.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2428, pruned_loss=0.03505, over 3323968.07 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 2.0 2023-05-02 16:34:00,649 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2798, 5.6171, 5.3669, 5.4241, 5.1200, 5.1386, 5.0580, 5.7261], device='cuda:0'), covar=tensor([0.1342, 0.1036, 0.1121, 0.0983, 0.0902, 0.0902, 0.1282, 0.0989], device='cuda:0'), in_proj_covar=tensor([0.0725, 0.0877, 0.0717, 0.0681, 0.0555, 0.0553, 0.0738, 0.0687], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:34:07,014 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3166, 2.4157, 2.5047, 4.1922, 2.3571, 2.7725, 2.4770, 2.5895], device='cuda:0'), covar=tensor([0.1474, 0.3681, 0.3154, 0.0600, 0.4102, 0.2719, 0.3702, 0.3681], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0477, 0.0391, 0.0338, 0.0449, 0.0546, 0.0450, 0.0560], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:34:22,233 INFO [optim.py:368] (0/8) 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 16:34:40,266 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=285398.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 16:34:47,469 INFO [train.py:904] (0/8) Epoch 29, batch 1200, loss[loss=0.1501, simple_loss=0.2338, pruned_loss=0.03318, over 16821.00 frames. ], tot_loss[loss=0.1553, simple_loss=0.2417, pruned_loss=0.03447, over 3324928.59 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:35:22,708 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 16:35:56,666 INFO [train.py:904] (0/8) Epoch 29, batch 1250, loss[loss=0.1382, simple_loss=0.228, pruned_loss=0.02418, over 17234.00 frames. ], tot_loss[loss=0.156, simple_loss=0.2424, pruned_loss=0.0348, over 3328501.48 frames. ], batch size: 43, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:36:21,873 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:36:38,702 INFO [optim.py:368] (0/8) 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 16:37:05,022 INFO [train.py:904] (0/8) Epoch 29, batch 1300, loss[loss=0.1351, simple_loss=0.2224, pruned_loss=0.02388, over 16784.00 frames. ], tot_loss[loss=0.1559, simple_loss=0.242, pruned_loss=0.03491, over 3325778.75 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:37:17,133 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5558, 4.2089, 4.5951, 2.7167, 4.8291, 4.8196, 3.7020, 4.0481], device='cuda:0'), covar=tensor([0.0595, 0.0245, 0.0210, 0.1049, 0.0078, 0.0192, 0.0357, 0.0330], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0141, 0.0087, 0.0133, 0.0131, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 16:37:27,305 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285520.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:38:00,072 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7507, 5.0338, 4.8328, 4.8348, 4.5710, 4.5922, 4.4600, 5.1305], device='cuda:0'), covar=tensor([0.1270, 0.0918, 0.1064, 0.0927, 0.0833, 0.1238, 0.1285, 0.0915], device='cuda:0'), in_proj_covar=tensor([0.0727, 0.0878, 0.0719, 0.0681, 0.0556, 0.0554, 0.0738, 0.0689], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:38:13,888 INFO [train.py:904] (0/8) Epoch 29, batch 1350, loss[loss=0.1494, simple_loss=0.2374, pruned_loss=0.03067, over 17006.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2427, pruned_loss=0.03512, over 3326731.40 frames. ], batch size: 41, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:38:58,372 INFO [optim.py:368] (0/8) 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 16:39:11,797 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285595.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:39:17,480 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4572, 2.4666, 2.4143, 4.2144, 2.4066, 2.8171, 2.4900, 2.6124], device='cuda:0'), covar=tensor([0.1452, 0.3681, 0.3384, 0.0628, 0.4331, 0.2791, 0.3849, 0.3740], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0479, 0.0392, 0.0339, 0.0450, 0.0549, 0.0452, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:39:24,839 INFO [train.py:904] (0/8) Epoch 29, batch 1400, loss[loss=0.1442, simple_loss=0.2352, pruned_loss=0.02664, over 17241.00 frames. ], tot_loss[loss=0.1568, simple_loss=0.2432, pruned_loss=0.03522, over 3334365.29 frames. ], batch size: 44, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:39:25,173 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0547, 5.2343, 5.5806, 5.5637, 5.6000, 5.2779, 5.1487, 5.0180], device='cuda:0'), covar=tensor([0.0502, 0.0745, 0.0489, 0.0570, 0.0665, 0.0509, 0.1329, 0.0496], device='cuda:0'), in_proj_covar=tensor([0.0445, 0.0503, 0.0485, 0.0446, 0.0532, 0.0508, 0.0584, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 16:39:31,923 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 16:39:49,586 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=285622.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:19,789 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:29,901 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4324, 4.4136, 4.3615, 3.8360, 4.3777, 1.9372, 4.1545, 3.9335], device='cuda:0'), covar=tensor([0.0145, 0.0141, 0.0188, 0.0290, 0.0105, 0.2731, 0.0140, 0.0262], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0177, 0.0215, 0.0185, 0.0192, 0.0220, 0.0203, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:40:30,947 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:40:33,557 INFO [train.py:904] (0/8) Epoch 29, batch 1450, loss[loss=0.153, simple_loss=0.2283, pruned_loss=0.03882, over 16886.00 frames. ], tot_loss[loss=0.1572, simple_loss=0.2429, pruned_loss=0.03571, over 3323745.27 frames. ], batch size: 96, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:41:13,935 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=285683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:16,390 INFO [optim.py:368] (0/8) 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 16:41:34,308 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=285698.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:41:37,508 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:41:43,190 INFO [train.py:904] (0/8) Epoch 29, batch 1500, loss[loss=0.1893, simple_loss=0.2561, pruned_loss=0.06126, over 16887.00 frames. ], tot_loss[loss=0.1571, simple_loss=0.243, pruned_loss=0.03565, over 3323771.89 frames. ], batch size: 109, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:42:40,638 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=285746.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:42:51,747 INFO [train.py:904] (0/8) Epoch 29, batch 1550, loss[loss=0.1593, simple_loss=0.2389, pruned_loss=0.03982, over 16902.00 frames. ], tot_loss[loss=0.1579, simple_loss=0.2437, pruned_loss=0.03606, over 3329670.23 frames. ], batch size: 90, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:43:34,670 INFO [optim.py:368] (0/8) 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 16:43:44,829 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.51 vs. limit=2.0 2023-05-02 16:44:00,607 INFO [train.py:904] (0/8) Epoch 29, batch 1600, loss[loss=0.1464, simple_loss=0.2284, pruned_loss=0.03217, over 16769.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2461, pruned_loss=0.03722, over 3322267.17 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:44:30,955 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 16:45:09,523 INFO [train.py:904] (0/8) Epoch 29, batch 1650, loss[loss=0.1878, simple_loss=0.2589, pruned_loss=0.05831, over 16727.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2465, pruned_loss=0.03696, over 3327839.38 frames. ], batch size: 89, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:45:50,990 INFO [optim.py:368] (0/8) 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 16:46:16,890 INFO [train.py:904] (0/8) Epoch 29, batch 1700, loss[loss=0.178, simple_loss=0.262, pruned_loss=0.04703, over 16881.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2479, pruned_loss=0.03667, over 3328323.04 frames. ], batch size: 116, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:46:58,560 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.22 vs. limit=5.0 2023-05-02 16:47:24,311 INFO [train.py:904] (0/8) Epoch 29, batch 1750, loss[loss=0.1537, simple_loss=0.2569, pruned_loss=0.02525, over 17251.00 frames. ], tot_loss[loss=0.1616, simple_loss=0.2488, pruned_loss=0.03722, over 3319123.90 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:47:24,887 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8317, 2.7279, 2.9165, 5.0082, 3.8941, 4.3640, 1.8092, 3.1961], device='cuda:0'), covar=tensor([0.1501, 0.0928, 0.1143, 0.0177, 0.0238, 0.0421, 0.1750, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0180, 0.0199, 0.0204, 0.0205, 0.0218, 0.0210, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 16:47:57,689 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 16:48:28,349 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-286000.pt 2023-05-02 16:48:36,769 INFO [train.py:904] (0/8) Epoch 29, batch 1800, loss[loss=0.164, simple_loss=0.257, pruned_loss=0.03546, over 17252.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2501, pruned_loss=0.03737, over 3318460.99 frames. ], batch size: 52, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:49:43,960 INFO [train.py:904] (0/8) Epoch 29, batch 1850, loss[loss=0.1718, simple_loss=0.2494, pruned_loss=0.0471, over 16751.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.251, pruned_loss=0.03758, over 3323182.00 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:50:28,095 INFO [optim.py:368] (0/8) 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 16:50:53,107 INFO [train.py:904] (0/8) Epoch 29, batch 1900, loss[loss=0.2121, simple_loss=0.2976, pruned_loss=0.06329, over 11792.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2503, pruned_loss=0.03708, over 3316364.05 frames. ], batch size: 247, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:51:18,917 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5861, 4.5929, 4.8274, 4.6075, 4.6935, 5.2666, 4.7453, 4.4173], device='cuda:0'), covar=tensor([0.1577, 0.2262, 0.2350, 0.2316, 0.2718, 0.1150, 0.1810, 0.2702], device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0647, 0.0721, 0.0529, 0.0709, 0.0740, 0.0557, 0.0705], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 16:51:20,192 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286123.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 16:51:29,284 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:52:04,253 INFO [train.py:904] (0/8) Epoch 29, batch 1950, loss[loss=0.1682, simple_loss=0.2643, pruned_loss=0.03605, over 17172.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2507, pruned_loss=0.03677, over 3313911.21 frames. ], batch size: 46, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:52:28,959 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1128, 4.5510, 4.5145, 3.4204, 3.7184, 4.4360, 3.9673, 2.8112], device='cuda:0'), covar=tensor([0.0428, 0.0081, 0.0054, 0.0326, 0.0176, 0.0114, 0.0111, 0.0446], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0091, 0.0092, 0.0137, 0.0104, 0.0117, 0.0099, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 16:52:42,614 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5885, 3.5456, 3.4510, 2.7182, 3.2125, 2.0406, 3.0775, 2.6687], device='cuda:0'), covar=tensor([0.0169, 0.0175, 0.0192, 0.0251, 0.0134, 0.2614, 0.0155, 0.0303], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0177, 0.0216, 0.0186, 0.0193, 0.0220, 0.0204, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:52:46,859 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286184.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 16:52:48,713 INFO [optim.py:368] (0/8) 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 16:52:55,358 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286190.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:53:13,098 INFO [train.py:904] (0/8) Epoch 29, batch 2000, loss[loss=0.1738, simple_loss=0.2716, pruned_loss=0.03796, over 16654.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2506, pruned_loss=0.03672, over 3308651.95 frames. ], batch size: 57, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:21,795 INFO [train.py:904] (0/8) Epoch 29, batch 2050, loss[loss=0.1568, simple_loss=0.2572, pruned_loss=0.02819, over 17070.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2512, pruned_loss=0.03753, over 3310210.80 frames. ], batch size: 50, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:54:54,871 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286278.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:55:04,678 INFO [optim.py:368] (0/8) 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 16:55:18,568 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286295.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:55:30,416 INFO [train.py:904] (0/8) Epoch 29, batch 2100, loss[loss=0.1929, simple_loss=0.2703, pruned_loss=0.05777, over 16319.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2521, pruned_loss=0.03818, over 3314440.73 frames. ], batch size: 165, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:00,348 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286326.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:26,333 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:40,455 INFO [train.py:904] (0/8) Epoch 29, batch 2150, loss[loss=0.213, simple_loss=0.2876, pruned_loss=0.06921, over 11759.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.2533, pruned_loss=0.03903, over 3308210.22 frames. ], batch size: 247, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 16:56:44,719 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286356.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:56:52,398 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 16:56:57,442 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6952, 2.7187, 2.4018, 2.6577, 2.9686, 2.7931, 3.2574, 3.2101], device='cuda:0'), covar=tensor([0.0195, 0.0526, 0.0610, 0.0494, 0.0362, 0.0488, 0.0307, 0.0348], device='cuda:0'), in_proj_covar=tensor([0.0240, 0.0254, 0.0241, 0.0242, 0.0254, 0.0252, 0.0250, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:57:06,931 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9314, 5.0955, 5.3997, 5.3880, 5.4690, 5.1497, 5.0156, 4.8730], device='cuda:0'), covar=tensor([0.0505, 0.0733, 0.0603, 0.0644, 0.0679, 0.0631, 0.1307, 0.0545], device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0510, 0.0492, 0.0454, 0.0541, 0.0516, 0.0592, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 16:57:11,640 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8526, 4.6447, 4.8994, 5.0677, 5.2079, 4.6954, 5.2196, 5.2210], device='cuda:0'), covar=tensor([0.1929, 0.1469, 0.1729, 0.0774, 0.0576, 0.1021, 0.0745, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0704, 0.0864, 0.0998, 0.0876, 0.0666, 0.0691, 0.0733, 0.0845], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 16:57:24,983 INFO [optim.py:368] (0/8) 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 16:57:50,678 INFO [train.py:904] (0/8) Epoch 29, batch 2200, loss[loss=0.1556, simple_loss=0.249, pruned_loss=0.03112, over 17240.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2535, pruned_loss=0.03906, over 3311968.80 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:57:52,972 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:58:59,517 INFO [train.py:904] (0/8) Epoch 29, batch 2250, loss[loss=0.1343, simple_loss=0.2234, pruned_loss=0.02264, over 16762.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2534, pruned_loss=0.03934, over 3309497.86 frames. ], batch size: 39, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 16:59:16,133 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0325, 2.3139, 2.7381, 2.9594, 2.8659, 3.5548, 2.5937, 3.4705], device='cuda:0'), covar=tensor([0.0306, 0.0536, 0.0375, 0.0403, 0.0393, 0.0200, 0.0527, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0203, 0.0190, 0.0197, 0.0214, 0.0170, 0.0207, 0.0170], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 16:59:33,869 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286479.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 16:59:43,677 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286485.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 16:59:45,657 INFO [optim.py:368] (0/8) 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 17:00:08,675 INFO [train.py:904] (0/8) Epoch 29, batch 2300, loss[loss=0.1544, simple_loss=0.2502, pruned_loss=0.02932, over 17024.00 frames. ], tot_loss[loss=0.1659, simple_loss=0.2533, pruned_loss=0.03928, over 3317861.08 frames. ], batch size: 50, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:00:11,545 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.74 vs. limit=5.0 2023-05-02 17:00:26,993 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3523, 3.9084, 4.3469, 2.3042, 4.5792, 4.6779, 3.4283, 3.5585], device='cuda:0'), covar=tensor([0.0631, 0.0278, 0.0250, 0.1136, 0.0088, 0.0188, 0.0473, 0.0430], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0113, 0.0104, 0.0141, 0.0088, 0.0134, 0.0132, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 17:01:17,622 INFO [train.py:904] (0/8) Epoch 29, batch 2350, loss[loss=0.1591, simple_loss=0.2444, pruned_loss=0.03688, over 16490.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2537, pruned_loss=0.03966, over 3317356.35 frames. ], batch size: 146, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:02:03,070 INFO [optim.py:368] (0/8) 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 17:02:04,845 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286588.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:02:16,979 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0782, 3.2807, 3.5944, 2.2044, 3.0097, 2.3878, 3.6060, 3.5932], device='cuda:0'), covar=tensor([0.0285, 0.1011, 0.0631, 0.2110, 0.0918, 0.1088, 0.0587, 0.1003], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0172, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 17:02:27,454 INFO [train.py:904] (0/8) Epoch 29, batch 2400, loss[loss=0.1701, simple_loss=0.2481, pruned_loss=0.04605, over 16814.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2543, pruned_loss=0.03941, over 3319343.58 frames. ], batch size: 102, lr: 2.33e-03, grad_scale: 8.0 2023-05-02 17:03:30,976 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286649.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:33,615 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286651.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:03:37,420 INFO [train.py:904] (0/8) Epoch 29, batch 2450, loss[loss=0.1523, simple_loss=0.2486, pruned_loss=0.02798, over 17272.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2542, pruned_loss=0.03891, over 3326253.08 frames. ], batch size: 45, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:03:59,896 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 17:04:25,677 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 17:04:41,086 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286700.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:04:46,157 INFO [train.py:904] (0/8) Epoch 29, batch 2500, loss[loss=0.1631, simple_loss=0.2434, pruned_loss=0.04142, over 16794.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2552, pruned_loss=0.03906, over 3323495.72 frames. ], batch size: 83, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:04:47,924 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8138, 2.9341, 2.6506, 5.0139, 3.8027, 4.3516, 1.6343, 3.1426], device='cuda:0'), covar=tensor([0.1564, 0.0926, 0.1415, 0.0218, 0.0206, 0.0434, 0.1904, 0.0875], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0207, 0.0206, 0.0221, 0.0211, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 17:05:17,526 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.68 vs. limit=2.0 2023-05-02 17:05:24,969 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286732.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:05:41,486 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5014, 4.3567, 4.5827, 4.7105, 4.8219, 4.3679, 4.6985, 4.8084], device='cuda:0'), covar=tensor([0.1836, 0.1371, 0.1328, 0.0693, 0.0637, 0.1146, 0.1824, 0.0904], device='cuda:0'), in_proj_covar=tensor([0.0704, 0.0864, 0.0995, 0.0875, 0.0667, 0.0693, 0.0734, 0.0848], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:05:55,673 INFO [train.py:904] (0/8) Epoch 29, batch 2550, loss[loss=0.1672, simple_loss=0.2452, pruned_loss=0.04458, over 16765.00 frames. ], tot_loss[loss=0.1674, simple_loss=0.2559, pruned_loss=0.03943, over 3323767.90 frames. ], batch size: 124, lr: 2.33e-03, grad_scale: 4.0 2023-05-02 17:06:19,074 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=286770.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:32,389 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286779.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:06:40,965 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:06:44,659 INFO [optim.py:368] (0/8) 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 17:07:07,738 INFO [train.py:904] (0/8) Epoch 29, batch 2600, loss[loss=0.1667, simple_loss=0.2621, pruned_loss=0.03565, over 16773.00 frames. ], tot_loss[loss=0.1665, simple_loss=0.2555, pruned_loss=0.03877, over 3326863.49 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:07:39,075 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286827.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:07:44,531 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=286831.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:07:47,331 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286833.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:08:15,902 INFO [train.py:904] (0/8) Epoch 29, batch 2650, loss[loss=0.1669, simple_loss=0.2686, pruned_loss=0.03261, over 17128.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2559, pruned_loss=0.03883, over 3334885.32 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:08:49,584 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.76 vs. limit=5.0 2023-05-02 17:09:00,291 INFO [optim.py:368] (0/8) 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 17:09:22,334 INFO [train.py:904] (0/8) Epoch 29, batch 2700, loss[loss=0.1599, simple_loss=0.248, pruned_loss=0.03588, over 17210.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2554, pruned_loss=0.03814, over 3337700.77 frames. ], batch size: 45, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:09:40,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6788, 4.7313, 5.0179, 4.9946, 5.0672, 4.7621, 4.7376, 4.5650], device='cuda:0'), covar=tensor([0.0365, 0.0554, 0.0392, 0.0418, 0.0524, 0.0430, 0.0951, 0.0583], device='cuda:0'), in_proj_covar=tensor([0.0451, 0.0510, 0.0493, 0.0451, 0.0539, 0.0516, 0.0594, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 17:10:17,474 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=286944.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:26,223 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=286951.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:10:30,044 INFO [train.py:904] (0/8) Epoch 29, batch 2750, loss[loss=0.1546, simple_loss=0.2539, pruned_loss=0.02766, over 17248.00 frames. ], tot_loss[loss=0.1657, simple_loss=0.256, pruned_loss=0.03772, over 3337856.79 frames. ], batch size: 52, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:17,995 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=286999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:35,325 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287000.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:11:40,743 INFO [train.py:904] (0/8) Epoch 29, batch 2800, loss[loss=0.1598, simple_loss=0.2504, pruned_loss=0.03462, over 16536.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2547, pruned_loss=0.0377, over 3337175.63 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:11:56,096 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5508, 3.8053, 3.9740, 2.8569, 3.5438, 4.0091, 3.6415, 2.5138], device='cuda:0'), covar=tensor([0.0511, 0.0260, 0.0068, 0.0372, 0.0144, 0.0123, 0.0117, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0091, 0.0093, 0.0137, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 17:12:12,983 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:19,170 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0897, 5.0429, 4.9774, 4.5035, 4.6738, 4.9938, 4.9018, 4.6855], device='cuda:0'), covar=tensor([0.0554, 0.0535, 0.0300, 0.0349, 0.0922, 0.0482, 0.0400, 0.0744], device='cuda:0'), in_proj_covar=tensor([0.0324, 0.0488, 0.0380, 0.0381, 0.0375, 0.0438, 0.0260, 0.0454], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 17:12:42,703 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287048.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:12:46,999 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2266, 5.5294, 5.2805, 5.2944, 5.0747, 4.9928, 4.9248, 5.6336], device='cuda:0'), covar=tensor([0.1278, 0.0952, 0.1159, 0.0909, 0.0825, 0.0985, 0.1358, 0.0951], device='cuda:0'), in_proj_covar=tensor([0.0733, 0.0888, 0.0727, 0.0688, 0.0561, 0.0559, 0.0745, 0.0698], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:12:50,055 INFO [train.py:904] (0/8) Epoch 29, batch 2850, loss[loss=0.1761, simple_loss=0.2683, pruned_loss=0.04191, over 16684.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.255, pruned_loss=0.03833, over 3326978.28 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:13:39,986 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.623e+02 2.226e+02 2.685e+02 3.371e+02 7.491e+02, threshold=5.370e+02, percent-clipped=5.0 2023-05-02 17:14:00,573 INFO [train.py:904] (0/8) Epoch 29, batch 2900, loss[loss=0.1449, simple_loss=0.2322, pruned_loss=0.02881, over 17215.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2546, pruned_loss=0.03884, over 3318725.62 frames. ], batch size: 45, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:14:23,298 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2816, 5.2577, 4.9962, 4.4468, 5.0828, 1.9261, 4.8255, 4.7633], device='cuda:0'), covar=tensor([0.0116, 0.0097, 0.0248, 0.0454, 0.0124, 0.3095, 0.0169, 0.0298], device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0179, 0.0217, 0.0189, 0.0195, 0.0222, 0.0205, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:14:31,475 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:15:10,998 INFO [train.py:904] (0/8) Epoch 29, batch 2950, loss[loss=0.1803, simple_loss=0.2668, pruned_loss=0.04695, over 16511.00 frames. ], tot_loss[loss=0.166, simple_loss=0.2538, pruned_loss=0.03908, over 3320044.44 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:15:59,394 INFO [optim.py:368] (0/8) 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 17:16:12,078 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.13 vs. limit=2.0 2023-05-02 17:16:20,173 INFO [train.py:904] (0/8) Epoch 29, batch 3000, loss[loss=0.1585, simple_loss=0.2558, pruned_loss=0.03058, over 17127.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2535, pruned_loss=0.03946, over 3323333.05 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:16:20,174 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 17:16:28,748 INFO [train.py:938] (0/8) Epoch 29, validation: loss=0.1336, simple_loss=0.2385, pruned_loss=0.01438, over 944034.00 frames. 2023-05-02 17:16:28,749 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 17:17:25,962 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287244.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:17:40,003 INFO [train.py:904] (0/8) Epoch 29, batch 3050, loss[loss=0.1459, simple_loss=0.234, pruned_loss=0.02895, over 17245.00 frames. ], tot_loss[loss=0.1651, simple_loss=0.2527, pruned_loss=0.03878, over 3329145.67 frames. ], batch size: 45, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:17:58,040 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6345, 4.7028, 5.0111, 5.0056, 5.0597, 4.7232, 4.7103, 4.5421], device='cuda:0'), covar=tensor([0.0406, 0.0661, 0.0456, 0.0458, 0.0570, 0.0457, 0.0953, 0.0660], device='cuda:0'), in_proj_covar=tensor([0.0452, 0.0512, 0.0495, 0.0453, 0.0541, 0.0518, 0.0597, 0.0417], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 17:18:29,496 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.465e+02 2.177e+02 2.556e+02 3.004e+02 5.914e+02, threshold=5.112e+02, percent-clipped=3.0 2023-05-02 17:18:34,331 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287292.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:18:49,728 INFO [train.py:904] (0/8) Epoch 29, batch 3100, loss[loss=0.1484, simple_loss=0.2343, pruned_loss=0.03128, over 16496.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2524, pruned_loss=0.0383, over 3332999.86 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:19:15,260 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4464, 1.8164, 2.2053, 2.3351, 2.5237, 2.4863, 1.9357, 2.5673], device='cuda:0'), covar=tensor([0.0260, 0.0556, 0.0346, 0.0346, 0.0360, 0.0379, 0.0613, 0.0247], device='cuda:0'), in_proj_covar=tensor([0.0204, 0.0204, 0.0192, 0.0198, 0.0216, 0.0172, 0.0208, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 17:19:22,448 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:00,325 INFO [train.py:904] (0/8) Epoch 29, batch 3150, loss[loss=0.1436, simple_loss=0.2349, pruned_loss=0.02616, over 17156.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2519, pruned_loss=0.03867, over 3320490.04 frames. ], batch size: 46, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:20:30,292 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287375.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:20:49,322 INFO [optim.py:368] (0/8) 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 17:21:05,023 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1224, 5.1629, 5.5263, 5.4936, 5.5546, 5.2049, 5.1456, 4.9686], device='cuda:0'), covar=tensor([0.0388, 0.0592, 0.0443, 0.0496, 0.0546, 0.0429, 0.1016, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0511, 0.0494, 0.0451, 0.0540, 0.0517, 0.0595, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 17:21:06,476 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2299, 4.3211, 4.2027, 3.9983, 3.7215, 4.3804, 4.0493, 4.0895], device='cuda:0'), covar=tensor([0.1027, 0.1167, 0.0461, 0.0458, 0.1423, 0.0642, 0.1084, 0.0733], device='cuda:0'), in_proj_covar=tensor([0.0327, 0.0493, 0.0383, 0.0385, 0.0379, 0.0442, 0.0263, 0.0458], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 17:21:10,113 INFO [train.py:904] (0/8) Epoch 29, batch 3200, loss[loss=0.1811, simple_loss=0.2757, pruned_loss=0.04325, over 16763.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2508, pruned_loss=0.03822, over 3318047.68 frames. ], batch size: 62, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:21:20,060 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287411.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:21:40,284 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=287426.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:00,042 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.90 vs. limit=5.0 2023-05-02 17:22:16,560 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2894, 5.6365, 5.3749, 5.4257, 5.1077, 5.0697, 5.0700, 5.7346], device='cuda:0'), covar=tensor([0.1373, 0.0927, 0.1006, 0.0975, 0.0912, 0.0906, 0.1292, 0.0878], device='cuda:0'), in_proj_covar=tensor([0.0741, 0.0895, 0.0732, 0.0696, 0.0567, 0.0564, 0.0752, 0.0702], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:22:18,331 INFO [train.py:904] (0/8) Epoch 29, batch 3250, loss[loss=0.1749, simple_loss=0.2556, pruned_loss=0.04706, over 16349.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2512, pruned_loss=0.03826, over 3315473.23 frames. ], batch size: 145, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:22:21,940 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8041, 3.4784, 3.9398, 2.1153, 4.0343, 4.0317, 3.2437, 3.0049], device='cuda:0'), covar=tensor([0.0761, 0.0280, 0.0168, 0.1199, 0.0112, 0.0208, 0.0401, 0.0485], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0139, 0.0087, 0.0133, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 17:22:43,508 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287472.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:45,676 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=287474.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:22:52,925 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.35 vs. limit=2.0 2023-05-02 17:23:05,924 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.595e+02 2.139e+02 2.558e+02 3.030e+02 4.692e+02, threshold=5.116e+02, percent-clipped=0.0 2023-05-02 17:23:26,861 INFO [train.py:904] (0/8) Epoch 29, batch 3300, loss[loss=0.2055, simple_loss=0.2819, pruned_loss=0.06454, over 16693.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2515, pruned_loss=0.03829, over 3312613.93 frames. ], batch size: 134, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:23:40,971 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 17:24:02,536 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.2031, 3.8809, 4.3646, 2.3009, 4.5800, 4.6477, 3.3398, 3.5447], device='cuda:0'), covar=tensor([0.0718, 0.0279, 0.0246, 0.1186, 0.0100, 0.0208, 0.0478, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0149, 0.0112, 0.0103, 0.0139, 0.0087, 0.0133, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 17:24:11,480 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287536.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:24:34,317 INFO [train.py:904] (0/8) Epoch 29, batch 3350, loss[loss=0.1736, simple_loss=0.2606, pruned_loss=0.04324, over 15627.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2521, pruned_loss=0.03826, over 3313964.10 frames. ], batch size: 191, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:24:45,924 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287562.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:14,592 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 17:25:22,982 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5796, 4.6468, 4.9645, 4.9329, 4.9798, 4.6538, 4.6395, 4.5359], device='cuda:0'), covar=tensor([0.0395, 0.0676, 0.0414, 0.0455, 0.0554, 0.0466, 0.0892, 0.0592], device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0510, 0.0493, 0.0452, 0.0540, 0.0517, 0.0595, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 17:25:33,277 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287597.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:25:43,307 INFO [train.py:904] (0/8) Epoch 29, batch 3400, loss[loss=0.1681, simple_loss=0.2477, pruned_loss=0.04424, over 16340.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2525, pruned_loss=0.03826, over 3314405.36 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:26:10,427 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:40,121 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287644.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:26:53,544 INFO [train.py:904] (0/8) Epoch 29, batch 3450, loss[loss=0.1497, simple_loss=0.2355, pruned_loss=0.03198, over 16292.00 frames. ], tot_loss[loss=0.1629, simple_loss=0.2503, pruned_loss=0.03778, over 3319936.00 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:27:43,292 INFO [optim.py:368] (0/8) 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 17:28:03,409 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0064, 5.3579, 5.1036, 5.1350, 4.8627, 4.8933, 4.8177, 5.4236], device='cuda:0'), covar=tensor([0.1325, 0.0857, 0.1041, 0.0920, 0.0836, 0.0950, 0.1307, 0.0935], device='cuda:0'), in_proj_covar=tensor([0.0745, 0.0898, 0.0736, 0.0697, 0.0570, 0.0567, 0.0754, 0.0705], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:28:04,239 INFO [train.py:904] (0/8) Epoch 29, batch 3500, loss[loss=0.1599, simple_loss=0.2371, pruned_loss=0.04135, over 16853.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2496, pruned_loss=0.03758, over 3324881.84 frames. ], batch size: 96, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:28:23,739 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3319, 5.7264, 5.4726, 5.5582, 5.1641, 5.2495, 5.1880, 5.8059], device='cuda:0'), covar=tensor([0.1558, 0.1051, 0.1120, 0.0885, 0.0965, 0.0818, 0.1367, 0.0999], device='cuda:0'), in_proj_covar=tensor([0.0744, 0.0897, 0.0735, 0.0697, 0.0570, 0.0567, 0.0753, 0.0704], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:28:26,769 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:28:43,861 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 17:29:09,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1480, 2.4174, 2.4681, 3.9472, 3.0889, 3.8978, 1.7996, 2.7919], device='cuda:0'), covar=tensor([0.1225, 0.0915, 0.1369, 0.0254, 0.0193, 0.0607, 0.1550, 0.0983], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0182, 0.0202, 0.0209, 0.0208, 0.0222, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 17:29:14,567 INFO [train.py:904] (0/8) Epoch 29, batch 3550, loss[loss=0.17, simple_loss=0.2454, pruned_loss=0.04732, over 16278.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2494, pruned_loss=0.03777, over 3321721.09 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:29:33,114 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287767.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:42,538 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=287773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:29:53,937 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287781.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:30:07,192 INFO [optim.py:368] (0/8) 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 17:30:26,891 INFO [train.py:904] (0/8) Epoch 29, batch 3600, loss[loss=0.1351, simple_loss=0.2241, pruned_loss=0.02304, over 17019.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2488, pruned_loss=0.03748, over 3314133.95 frames. ], batch size: 41, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:31:10,896 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=287834.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:31:40,150 INFO [train.py:904] (0/8) Epoch 29, batch 3650, loss[loss=0.1579, simple_loss=0.2326, pruned_loss=0.04159, over 16452.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2474, pruned_loss=0.03732, over 3304638.23 frames. ], batch size: 68, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:32:21,880 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5452, 4.5157, 4.4977, 3.9941, 4.5171, 2.0549, 4.2919, 4.1463], device='cuda:0'), covar=tensor([0.0142, 0.0127, 0.0191, 0.0321, 0.0106, 0.2614, 0.0149, 0.0232], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0182, 0.0221, 0.0192, 0.0198, 0.0224, 0.0209, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:32:34,204 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3423, 3.5681, 3.6129, 3.5882, 3.6082, 3.4844, 3.4982, 3.5063], device='cuda:0'), covar=tensor([0.0422, 0.0596, 0.0470, 0.0457, 0.0609, 0.0495, 0.0765, 0.0582], device='cuda:0'), in_proj_covar=tensor([0.0454, 0.0514, 0.0496, 0.0456, 0.0545, 0.0522, 0.0600, 0.0419], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 17:32:35,504 INFO [optim.py:368] (0/8) 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 17:32:38,199 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287892.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:32:57,095 INFO [train.py:904] (0/8) Epoch 29, batch 3700, loss[loss=0.1752, simple_loss=0.2501, pruned_loss=0.05012, over 16718.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2466, pruned_loss=0.03883, over 3279836.17 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:33:18,790 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=287918.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:33:48,733 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 29, batch 3750, loss[loss=0.1695, simple_loss=0.2421, pruned_loss=0.04841, over 16879.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2469, pruned_loss=0.03993, over 3269571.21 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:34:48,898 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2396, 3.3221, 3.5865, 2.3176, 3.1725, 2.4453, 3.6900, 3.7677], device='cuda:0'), covar=tensor([0.0237, 0.0883, 0.0618, 0.1952, 0.0797, 0.1062, 0.0496, 0.0800], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0173, 0.0171, 0.0158, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 17:35:05,561 INFO [optim.py:368] (0/8) 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 17:35:16,464 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1246, 2.2888, 2.4505, 3.9477, 2.2881, 2.5923, 2.3831, 2.4560], device='cuda:0'), covar=tensor([0.1716, 0.4058, 0.3093, 0.0661, 0.4035, 0.2778, 0.4184, 0.3142], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0483, 0.0392, 0.0344, 0.0451, 0.0554, 0.0454, 0.0567], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:35:21,673 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-288000.pt 2023-05-02 17:35:29,229 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288003.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:35:30,046 INFO [train.py:904] (0/8) Epoch 29, batch 3800, loss[loss=0.1735, simple_loss=0.2503, pruned_loss=0.04835, over 16193.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2484, pruned_loss=0.04117, over 3263046.33 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:40,386 INFO [train.py:904] (0/8) Epoch 29, batch 3850, loss[loss=0.1784, simple_loss=0.2572, pruned_loss=0.04986, over 16725.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2482, pruned_loss=0.04166, over 3275527.27 frames. ], batch size: 124, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:36:46,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 17:36:55,418 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288064.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:36:59,121 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288067.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:37:01,220 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.67 vs. limit=2.0 2023-05-02 17:37:12,442 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288076.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:37:30,815 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.513e+02 2.131e+02 2.477e+02 2.774e+02 5.510e+02, threshold=4.954e+02, percent-clipped=1.0 2023-05-02 17:37:50,143 INFO [train.py:904] (0/8) Epoch 29, batch 3900, loss[loss=0.1577, simple_loss=0.2376, pruned_loss=0.03894, over 16938.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.2484, pruned_loss=0.04241, over 3272996.74 frames. ], batch size: 90, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:37:53,626 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288106.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:38:05,839 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:27,307 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:38:41,132 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4607, 1.8096, 2.2207, 2.3160, 2.5074, 2.4704, 1.9197, 2.5964], device='cuda:0'), covar=tensor([0.0228, 0.0579, 0.0359, 0.0420, 0.0369, 0.0404, 0.0611, 0.0234], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0203, 0.0191, 0.0199, 0.0215, 0.0171, 0.0208, 0.0172], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 17:38:47,120 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0437, 5.3597, 5.1276, 5.1597, 4.8798, 4.8510, 4.8123, 5.4630], device='cuda:0'), covar=tensor([0.1234, 0.0867, 0.1048, 0.1003, 0.0860, 0.1094, 0.1268, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0744, 0.0896, 0.0733, 0.0697, 0.0571, 0.0567, 0.0754, 0.0704], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:38:58,597 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8991, 4.8188, 4.9353, 5.1134, 5.2615, 4.7037, 5.1981, 5.3085], device='cuda:0'), covar=tensor([0.1801, 0.1163, 0.1528, 0.0674, 0.0587, 0.1018, 0.0702, 0.0619], device='cuda:0'), in_proj_covar=tensor([0.0707, 0.0863, 0.0994, 0.0876, 0.0668, 0.0692, 0.0731, 0.0848], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:39:00,655 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288152.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:39:03,105 INFO [train.py:904] (0/8) Epoch 29, batch 3950, loss[loss=0.171, simple_loss=0.2409, pruned_loss=0.05055, over 16911.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2477, pruned_loss=0.043, over 3277080.63 frames. ], batch size: 96, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:39:22,329 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288167.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:39:55,386 INFO [optim.py:368] (0/8) 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 17:39:58,791 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:40:16,242 INFO [train.py:904] (0/8) Epoch 29, batch 4000, loss[loss=0.1589, simple_loss=0.2439, pruned_loss=0.03694, over 16784.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2475, pruned_loss=0.04336, over 3280637.74 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:40:30,224 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288213.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:40:36,628 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:07,393 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288240.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:41:08,596 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4752, 3.5733, 2.2903, 3.9250, 2.7249, 3.9599, 2.3278, 2.8968], device='cuda:0'), covar=tensor([0.0329, 0.0400, 0.1679, 0.0248, 0.0882, 0.0550, 0.1628, 0.0811], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0184, 0.0199, 0.0179, 0.0183, 0.0226, 0.0207, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 17:41:29,254 INFO [train.py:904] (0/8) Epoch 29, batch 4050, loss[loss=0.1628, simple_loss=0.244, pruned_loss=0.04075, over 16707.00 frames. ], tot_loss[loss=0.167, simple_loss=0.2488, pruned_loss=0.04265, over 3281404.29 frames. ], batch size: 57, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:41:33,154 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.31 vs. limit=5.0 2023-05-02 17:41:47,214 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288266.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:17,221 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288287.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:42:21,066 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.369e+02 1.858e+02 2.084e+02 2.415e+02 5.689e+02, threshold=4.167e+02, percent-clipped=1.0 2023-05-02 17:42:38,937 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1278, 3.7711, 3.7321, 2.3594, 3.3766, 3.7992, 3.4413, 2.0974], device='cuda:0'), covar=tensor([0.0614, 0.0059, 0.0063, 0.0472, 0.0120, 0.0099, 0.0112, 0.0493], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0091, 0.0093, 0.0137, 0.0104, 0.0118, 0.0100, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 17:42:42,072 INFO [train.py:904] (0/8) Epoch 29, batch 4100, loss[loss=0.177, simple_loss=0.2702, pruned_loss=0.04192, over 16815.00 frames. ], tot_loss[loss=0.1679, simple_loss=0.251, pruned_loss=0.04242, over 3272364.87 frames. ], batch size: 102, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:43:09,859 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 17:43:17,812 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288327.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:31,245 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1281, 2.4135, 2.3883, 3.8317, 2.2284, 2.7221, 2.3859, 2.4942], device='cuda:0'), covar=tensor([0.1480, 0.3220, 0.2909, 0.0601, 0.3883, 0.2275, 0.3386, 0.3329], device='cuda:0'), in_proj_covar=tensor([0.0427, 0.0482, 0.0391, 0.0342, 0.0449, 0.0554, 0.0454, 0.0565], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:43:45,780 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:43:59,220 INFO [train.py:904] (0/8) Epoch 29, batch 4150, loss[loss=0.1964, simple_loss=0.2799, pruned_loss=0.05644, over 17001.00 frames. ], tot_loss[loss=0.1741, simple_loss=0.2581, pruned_loss=0.04501, over 3229650.26 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:44:07,907 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:33,314 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:51,067 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288388.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:44:52,912 INFO [optim.py:368] (0/8) 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 17:45:02,657 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288396.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:13,941 INFO [train.py:904] (0/8) Epoch 29, batch 4200, loss[loss=0.2166, simple_loss=0.3169, pruned_loss=0.05815, over 16877.00 frames. ], tot_loss[loss=0.179, simple_loss=0.265, pruned_loss=0.04652, over 3204697.03 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:45:18,052 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:45,256 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:45:53,486 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:06,841 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3604, 2.4307, 2.3730, 4.1184, 2.3015, 2.7360, 2.4023, 2.5459], device='cuda:0'), covar=tensor([0.1453, 0.3898, 0.3277, 0.0545, 0.4428, 0.2581, 0.3977, 0.3645], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0479, 0.0389, 0.0340, 0.0447, 0.0551, 0.0451, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:46:29,466 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.29 vs. limit=2.0 2023-05-02 17:46:30,280 INFO [train.py:904] (0/8) Epoch 29, batch 4250, loss[loss=0.1699, simple_loss=0.2695, pruned_loss=0.0352, over 16491.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.2688, pruned_loss=0.04678, over 3158993.76 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:46:36,726 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288457.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:46:42,595 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288462.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 17:46:45,606 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5816, 3.6639, 3.4208, 3.0700, 3.2692, 3.5465, 3.3392, 3.3325], device='cuda:0'), covar=tensor([0.0616, 0.0866, 0.0328, 0.0317, 0.0589, 0.0555, 0.1270, 0.0492], device='cuda:0'), in_proj_covar=tensor([0.0321, 0.0486, 0.0377, 0.0379, 0.0373, 0.0435, 0.0258, 0.0450], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 17:47:05,116 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288477.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:47:22,847 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 17:47:24,432 INFO [optim.py:368] (0/8) 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 17:47:39,278 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.62 vs. limit=5.0 2023-05-02 17:47:45,993 INFO [train.py:904] (0/8) Epoch 29, batch 4300, loss[loss=0.1942, simple_loss=0.2835, pruned_loss=0.05249, over 11827.00 frames. ], tot_loss[loss=0.1812, simple_loss=0.27, pruned_loss=0.04623, over 3138881.12 frames. ], batch size: 246, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:47:51,121 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288508.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:48:11,893 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8918, 2.2814, 1.9728, 2.0655, 2.6468, 2.2946, 2.5169, 2.8053], device='cuda:0'), covar=tensor([0.0197, 0.0502, 0.0596, 0.0536, 0.0303, 0.0423, 0.0245, 0.0302], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0248, 0.0237, 0.0238, 0.0249, 0.0248, 0.0246, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:48:59,345 INFO [train.py:904] (0/8) Epoch 29, batch 4350, loss[loss=0.2077, simple_loss=0.2996, pruned_loss=0.05787, over 16566.00 frames. ], tot_loss[loss=0.1828, simple_loss=0.2726, pruned_loss=0.04654, over 3165300.34 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:49:53,415 INFO [optim.py:368] (0/8) 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 17:50:13,848 INFO [train.py:904] (0/8) Epoch 29, batch 4400, loss[loss=0.1888, simple_loss=0.275, pruned_loss=0.05137, over 16979.00 frames. ], tot_loss[loss=0.1852, simple_loss=0.2748, pruned_loss=0.04773, over 3175827.88 frames. ], batch size: 55, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:50:22,276 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:26,311 INFO [train.py:904] (0/8) Epoch 29, batch 4450, loss[loss=0.2011, simple_loss=0.2977, pruned_loss=0.05228, over 16984.00 frames. ], tot_loss[loss=0.1883, simple_loss=0.2783, pruned_loss=0.04912, over 3182229.89 frames. ], batch size: 50, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:51:30,123 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5784, 2.4424, 2.4880, 3.5260, 2.6108, 3.7031, 1.4741, 2.7804], device='cuda:0'), covar=tensor([0.1400, 0.0869, 0.1165, 0.0146, 0.0219, 0.0337, 0.1808, 0.0810], device='cuda:0'), in_proj_covar=tensor([0.0171, 0.0180, 0.0199, 0.0205, 0.0206, 0.0217, 0.0209, 0.0198], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 17:51:32,328 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4087, 5.2713, 5.4362, 5.5838, 5.7155, 5.1465, 5.7139, 5.7139], device='cuda:0'), covar=tensor([0.1579, 0.1028, 0.1259, 0.0554, 0.0392, 0.0680, 0.0389, 0.0466], device='cuda:0'), in_proj_covar=tensor([0.0693, 0.0845, 0.0975, 0.0860, 0.0655, 0.0678, 0.0716, 0.0831], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:51:33,498 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288659.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:51:39,673 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6726, 4.8345, 5.0475, 4.7254, 4.8849, 5.4185, 4.8462, 4.4493], device='cuda:0'), covar=tensor([0.1148, 0.1813, 0.1956, 0.1908, 0.2269, 0.0834, 0.1463, 0.2293], device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0646, 0.0713, 0.0524, 0.0701, 0.0734, 0.0553, 0.0697], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 17:51:48,969 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288670.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:52:08,530 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288683.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:18,792 INFO [optim.py:368] (0/8) 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 17:52:34,529 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288701.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:52:37,605 INFO [train.py:904] (0/8) Epoch 29, batch 4500, loss[loss=0.2009, simple_loss=0.2807, pruned_loss=0.06052, over 17226.00 frames. ], tot_loss[loss=0.1893, simple_loss=0.2788, pruned_loss=0.04988, over 3198897.01 frames. ], batch size: 45, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:52:42,934 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288707.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:49,136 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288752.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:53:51,862 INFO [train.py:904] (0/8) Epoch 29, batch 4550, loss[loss=0.2122, simple_loss=0.3075, pruned_loss=0.05846, over 16330.00 frames. ], tot_loss[loss=0.1909, simple_loss=0.2798, pruned_loss=0.05105, over 3202975.88 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:54:03,013 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288762.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 17:54:07,811 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 17:54:39,963 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1954, 3.3414, 3.6397, 2.1531, 3.2525, 2.3094, 3.5187, 3.6064], device='cuda:0'), covar=tensor([0.0217, 0.0899, 0.0576, 0.2238, 0.0816, 0.1131, 0.0574, 0.0901], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0174, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 17:54:44,112 INFO [optim.py:368] (0/8) 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 17:54:53,671 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=288797.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:55:04,701 INFO [train.py:904] (0/8) Epoch 29, batch 4600, loss[loss=0.2002, simple_loss=0.2876, pruned_loss=0.05639, over 16385.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2805, pruned_loss=0.05091, over 3222840.29 frames. ], batch size: 35, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:55:11,541 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288808.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 17:55:14,424 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288810.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:56:18,398 INFO [train.py:904] (0/8) Epoch 29, batch 4650, loss[loss=0.1781, simple_loss=0.2695, pruned_loss=0.04338, over 16879.00 frames. ], tot_loss[loss=0.192, simple_loss=0.2807, pruned_loss=0.05169, over 3217158.33 frames. ], batch size: 116, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 17:56:21,004 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=288856.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:56:24,254 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=288858.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:57:10,671 INFO [optim.py:368] (0/8) 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 17:57:29,736 INFO [train.py:904] (0/8) Epoch 29, batch 4700, loss[loss=0.1613, simple_loss=0.2587, pruned_loss=0.0319, over 16499.00 frames. ], tot_loss[loss=0.1894, simple_loss=0.2779, pruned_loss=0.05046, over 3213618.02 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:57:38,407 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6302, 5.9266, 5.6870, 5.7868, 5.4262, 5.2637, 5.3545, 6.0758], device='cuda:0'), covar=tensor([0.1337, 0.0801, 0.0903, 0.0823, 0.0842, 0.0724, 0.1143, 0.0857], device='cuda:0'), in_proj_covar=tensor([0.0724, 0.0871, 0.0714, 0.0677, 0.0555, 0.0551, 0.0730, 0.0686], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 17:58:41,607 INFO [train.py:904] (0/8) Epoch 29, batch 4750, loss[loss=0.1632, simple_loss=0.2511, pruned_loss=0.03769, over 17102.00 frames. ], tot_loss[loss=0.1846, simple_loss=0.2731, pruned_loss=0.04811, over 3213234.80 frames. ], batch size: 49, lr: 2.32e-03, grad_scale: 4.0 2023-05-02 17:58:57,752 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=288965.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 17:59:23,583 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=288983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:35,276 INFO [optim.py:368] (0/8) 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 17:59:43,666 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.28 vs. limit=5.0 2023-05-02 17:59:50,309 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289001.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 17:59:54,518 INFO [train.py:904] (0/8) Epoch 29, batch 4800, loss[loss=0.1893, simple_loss=0.2921, pruned_loss=0.04321, over 16333.00 frames. ], tot_loss[loss=0.1802, simple_loss=0.269, pruned_loss=0.04572, over 3213107.88 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:00:37,056 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289031.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:00:59,802 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289046.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:01:05,207 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289049.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:08,953 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289052.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:01:12,138 INFO [train.py:904] (0/8) Epoch 29, batch 4850, loss[loss=0.1889, simple_loss=0.292, pruned_loss=0.04286, over 16236.00 frames. ], tot_loss[loss=0.1797, simple_loss=0.2695, pruned_loss=0.04492, over 3199811.49 frames. ], batch size: 165, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:02,873 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9046, 5.1960, 5.3512, 5.0812, 5.1502, 5.7238, 5.1793, 4.9103], device='cuda:0'), covar=tensor([0.0938, 0.1753, 0.2033, 0.1930, 0.2385, 0.0884, 0.1455, 0.2252], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0636, 0.0701, 0.0516, 0.0691, 0.0725, 0.0545, 0.0688], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 18:02:08,680 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.294e+02 1.829e+02 2.187e+02 2.631e+02 3.612e+02, threshold=4.373e+02, percent-clipped=0.0 2023-05-02 18:02:22,360 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289100.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:02:28,901 INFO [train.py:904] (0/8) Epoch 29, batch 4900, loss[loss=0.165, simple_loss=0.2617, pruned_loss=0.03415, over 16709.00 frames. ], tot_loss[loss=0.1776, simple_loss=0.2683, pruned_loss=0.04343, over 3183484.50 frames. ], batch size: 83, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:02:33,635 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289107.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:02:56,654 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3745, 3.4535, 2.0846, 3.8752, 2.6351, 3.8514, 2.2272, 2.7860], device='cuda:0'), covar=tensor([0.0350, 0.0399, 0.1814, 0.0208, 0.0877, 0.0631, 0.1640, 0.0867], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0182, 0.0197, 0.0175, 0.0181, 0.0223, 0.0205, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 18:03:16,433 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289136.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:42,157 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:03:43,052 INFO [train.py:904] (0/8) Epoch 29, batch 4950, loss[loss=0.1806, simple_loss=0.2797, pruned_loss=0.04076, over 16435.00 frames. ], tot_loss[loss=0.1769, simple_loss=0.268, pruned_loss=0.04293, over 3190070.91 frames. ], batch size: 75, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:04:38,082 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.471e+02 2.040e+02 2.385e+02 2.853e+02 5.286e+02, threshold=4.771e+02, percent-clipped=2.0 2023-05-02 18:04:46,878 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289197.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:04:56,970 INFO [train.py:904] (0/8) Epoch 29, batch 5000, loss[loss=0.1655, simple_loss=0.2624, pruned_loss=0.03427, over 17201.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2693, pruned_loss=0.04289, over 3191659.58 frames. ], batch size: 44, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:06:09,974 INFO [train.py:904] (0/8) Epoch 29, batch 5050, loss[loss=0.1838, simple_loss=0.2806, pruned_loss=0.04348, over 16415.00 frames. ], tot_loss[loss=0.1777, simple_loss=0.2699, pruned_loss=0.04274, over 3198237.10 frames. ], batch size: 146, lr: 2.32e-03, grad_scale: 8.0 2023-05-02 18:06:26,281 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289265.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:06:56,113 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4421, 4.6957, 4.4966, 4.5587, 4.2942, 4.2469, 4.2148, 4.7345], device='cuda:0'), covar=tensor([0.1277, 0.0794, 0.0990, 0.0792, 0.0837, 0.1432, 0.1071, 0.0924], device='cuda:0'), in_proj_covar=tensor([0.0724, 0.0869, 0.0714, 0.0675, 0.0555, 0.0550, 0.0729, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:07:03,097 INFO [optim.py:368] (0/8) 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 18:07:15,190 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5377, 3.7568, 2.7364, 2.2731, 2.4442, 2.4458, 3.9914, 3.2944], device='cuda:0'), covar=tensor([0.3016, 0.0623, 0.1993, 0.2787, 0.2535, 0.2140, 0.0483, 0.1299], device='cuda:0'), in_proj_covar=tensor([0.0335, 0.0275, 0.0313, 0.0327, 0.0306, 0.0278, 0.0305, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 18:07:22,110 INFO [train.py:904] (0/8) Epoch 29, batch 5100, loss[loss=0.1684, simple_loss=0.2544, pruned_loss=0.04117, over 16862.00 frames. ], tot_loss[loss=0.1764, simple_loss=0.2684, pruned_loss=0.04223, over 3197012.85 frames. ], batch size: 109, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:07:34,857 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:08:19,108 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0630, 3.9140, 4.1098, 4.2360, 4.3716, 4.0166, 4.3287, 4.4113], device='cuda:0'), covar=tensor([0.1749, 0.1262, 0.1582, 0.0790, 0.0563, 0.1265, 0.0789, 0.0667], device='cuda:0'), in_proj_covar=tensor([0.0684, 0.0833, 0.0962, 0.0849, 0.0646, 0.0667, 0.0704, 0.0820], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:08:30,113 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0442, 2.8428, 3.0537, 1.7035, 3.1689, 3.2718, 2.7385, 2.5134], device='cuda:0'), covar=tensor([0.0997, 0.0286, 0.0223, 0.1422, 0.0150, 0.0195, 0.0436, 0.0562], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0112, 0.0103, 0.0139, 0.0088, 0.0132, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 18:08:36,378 INFO [train.py:904] (0/8) Epoch 29, batch 5150, loss[loss=0.1762, simple_loss=0.2762, pruned_loss=0.03812, over 15273.00 frames. ], tot_loss[loss=0.1756, simple_loss=0.2681, pruned_loss=0.0416, over 3194331.51 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:09:29,048 INFO [optim.py:368] (0/8) 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 18:09:45,676 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=289402.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:09:47,655 INFO [train.py:904] (0/8) Epoch 29, batch 5200, loss[loss=0.1674, simple_loss=0.2562, pruned_loss=0.0393, over 16683.00 frames. ], tot_loss[loss=0.1751, simple_loss=0.2673, pruned_loss=0.04144, over 3171891.49 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:10:30,764 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0752, 2.3138, 2.2179, 3.7219, 2.1595, 2.6845, 2.3569, 2.4590], device='cuda:0'), covar=tensor([0.1587, 0.3534, 0.3187, 0.0583, 0.4073, 0.2477, 0.3672, 0.2916], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0480, 0.0389, 0.0340, 0.0448, 0.0552, 0.0452, 0.0561], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:10:59,572 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289453.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:11:00,521 INFO [train.py:904] (0/8) Epoch 29, batch 5250, loss[loss=0.1637, simple_loss=0.261, pruned_loss=0.03322, over 16871.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2652, pruned_loss=0.04102, over 3170710.34 frames. ], batch size: 116, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:11:23,118 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 18:11:55,945 INFO [optim.py:368] (0/8) 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 18:11:58,121 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289501.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:14,371 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289503.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:12:15,681 INFO [train.py:904] (0/8) Epoch 29, batch 5300, loss[loss=0.1696, simple_loss=0.2546, pruned_loss=0.04227, over 12349.00 frames. ], tot_loss[loss=0.1709, simple_loss=0.2617, pruned_loss=0.04005, over 3175575.11 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:12:28,778 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9832, 4.2302, 4.0646, 4.1384, 3.8201, 3.8461, 3.8416, 4.2248], device='cuda:0'), covar=tensor([0.1242, 0.0922, 0.1002, 0.0756, 0.0805, 0.1755, 0.0956, 0.1021], device='cuda:0'), in_proj_covar=tensor([0.0728, 0.0875, 0.0717, 0.0677, 0.0558, 0.0553, 0.0732, 0.0688], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:13:26,908 INFO [train.py:904] (0/8) Epoch 29, batch 5350, loss[loss=0.1717, simple_loss=0.269, pruned_loss=0.03721, over 16717.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.26, pruned_loss=0.03953, over 3178707.24 frames. ], batch size: 89, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:13:42,077 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289564.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:14:21,416 INFO [optim.py:368] (0/8) 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 18:14:30,972 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7443, 4.9909, 5.1600, 4.8943, 4.9453, 5.5304, 5.0415, 4.7373], device='cuda:0'), covar=tensor([0.1123, 0.1712, 0.2043, 0.1955, 0.2389, 0.0848, 0.1526, 0.2223], device='cuda:0'), in_proj_covar=tensor([0.0428, 0.0634, 0.0696, 0.0514, 0.0688, 0.0724, 0.0543, 0.0685], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 18:14:41,101 INFO [train.py:904] (0/8) Epoch 29, batch 5400, loss[loss=0.166, simple_loss=0.2646, pruned_loss=0.03368, over 16892.00 frames. ], tot_loss[loss=0.1715, simple_loss=0.2627, pruned_loss=0.04017, over 3177855.87 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:15:57,441 INFO [train.py:904] (0/8) Epoch 29, batch 5450, loss[loss=0.177, simple_loss=0.2743, pruned_loss=0.03985, over 16725.00 frames. ], tot_loss[loss=0.1735, simple_loss=0.2649, pruned_loss=0.04107, over 3188974.46 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:16:53,642 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.432e+02 2.272e+02 2.915e+02 3.734e+02 7.781e+02, threshold=5.831e+02, percent-clipped=14.0 2023-05-02 18:17:11,563 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289702.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:17:14,653 INFO [train.py:904] (0/8) Epoch 29, batch 5500, loss[loss=0.2659, simple_loss=0.3378, pruned_loss=0.09702, over 11812.00 frames. ], tot_loss[loss=0.1813, simple_loss=0.2721, pruned_loss=0.04526, over 3157428.58 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:17:48,981 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7284, 3.1784, 3.0152, 1.7030, 2.6286, 1.8470, 3.2293, 3.5231], device='cuda:0'), covar=tensor([0.0267, 0.0833, 0.0831, 0.2712, 0.1159, 0.1364, 0.0716, 0.0918], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 18:18:26,455 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289750.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:18:31,719 INFO [train.py:904] (0/8) Epoch 29, batch 5550, loss[loss=0.2706, simple_loss=0.328, pruned_loss=0.1066, over 11106.00 frames. ], tot_loss[loss=0.1891, simple_loss=0.2787, pruned_loss=0.04977, over 3133804.77 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:19:30,746 INFO [optim.py:368] (0/8) 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 18:19:33,664 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=289792.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:19:53,208 INFO [train.py:904] (0/8) Epoch 29, batch 5600, loss[loss=0.1919, simple_loss=0.2793, pruned_loss=0.05229, over 16650.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2827, pruned_loss=0.0531, over 3107229.71 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:20:53,922 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=289840.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:21:16,896 INFO [train.py:904] (0/8) Epoch 29, batch 5650, loss[loss=0.1937, simple_loss=0.2827, pruned_loss=0.05236, over 16775.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2868, pruned_loss=0.05607, over 3095550.08 frames. ], batch size: 102, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:21:25,657 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 18:22:35,920 INFO [train.py:904] (0/8) Epoch 29, batch 5700, loss[loss=0.179, simple_loss=0.279, pruned_loss=0.03954, over 16807.00 frames. ], tot_loss[loss=0.2021, simple_loss=0.2886, pruned_loss=0.05777, over 3077158.97 frames. ], batch size: 83, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:22:38,481 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.78 vs. limit=5.0 2023-05-02 18:23:01,291 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289920.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:23:04,634 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0134, 3.3978, 3.4870, 2.1659, 2.9832, 2.3274, 3.3191, 3.6624], device='cuda:0'), covar=tensor([0.0367, 0.0917, 0.0599, 0.2161, 0.0944, 0.1034, 0.0913, 0.1079], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0172, 0.0171, 0.0157, 0.0148, 0.0133, 0.0146, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 18:23:55,713 INFO [train.py:904] (0/8) Epoch 29, batch 5750, loss[loss=0.2279, simple_loss=0.294, pruned_loss=0.08083, over 10934.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2919, pruned_loss=0.05965, over 3049841.70 frames. ], batch size: 246, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:24:40,784 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=289981.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 18:24:56,872 INFO [optim.py:368] (0/8) 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 18:25:08,734 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=289998.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:25:10,968 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-290000.pt 2023-05-02 18:25:20,970 INFO [train.py:904] (0/8) Epoch 29, batch 5800, loss[loss=0.1848, simple_loss=0.2714, pruned_loss=0.04905, over 16224.00 frames. ], tot_loss[loss=0.2056, simple_loss=0.2924, pruned_loss=0.0594, over 3044107.82 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:26:34,271 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290050.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:26:39,157 INFO [train.py:904] (0/8) Epoch 29, batch 5850, loss[loss=0.1905, simple_loss=0.2763, pruned_loss=0.05239, over 16646.00 frames. ], tot_loss[loss=0.2025, simple_loss=0.2899, pruned_loss=0.05752, over 3079654.22 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:26:48,250 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290059.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:27:14,304 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.38 vs. limit=2.0 2023-05-02 18:27:39,838 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.636e+02 2.628e+02 3.224e+02 3.685e+02 6.132e+02, threshold=6.447e+02, percent-clipped=0.0 2023-05-02 18:27:51,264 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6258, 3.7685, 2.8698, 2.2814, 2.4011, 2.4890, 4.0369, 3.2809], device='cuda:0'), covar=tensor([0.3090, 0.0565, 0.1938, 0.3068, 0.2787, 0.2204, 0.0490, 0.1383], device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0275, 0.0314, 0.0329, 0.0307, 0.0279, 0.0307, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 18:28:01,356 INFO [train.py:904] (0/8) Epoch 29, batch 5900, loss[loss=0.1837, simple_loss=0.2853, pruned_loss=0.04103, over 16730.00 frames. ], tot_loss[loss=0.2023, simple_loss=0.29, pruned_loss=0.05729, over 3093334.54 frames. ], batch size: 89, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:28:15,100 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290111.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:28:54,781 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.76 vs. limit=5.0 2023-05-02 18:29:20,960 INFO [train.py:904] (0/8) Epoch 29, batch 5950, loss[loss=0.175, simple_loss=0.2707, pruned_loss=0.0396, over 16852.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2909, pruned_loss=0.05611, over 3096635.25 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:29:30,122 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290159.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:29:50,675 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3883, 4.2481, 4.4283, 4.5824, 4.7370, 4.2814, 4.6850, 4.7718], device='cuda:0'), covar=tensor([0.1983, 0.1384, 0.1618, 0.0773, 0.0654, 0.1250, 0.0861, 0.0743], device='cuda:0'), in_proj_covar=tensor([0.0685, 0.0836, 0.0963, 0.0848, 0.0646, 0.0669, 0.0707, 0.0822], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:30:18,756 INFO [optim.py:368] (0/8) 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 18:30:40,229 INFO [train.py:904] (0/8) Epoch 29, batch 6000, loss[loss=0.195, simple_loss=0.2835, pruned_loss=0.05329, over 16344.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2901, pruned_loss=0.05574, over 3101888.98 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:30:40,230 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 18:30:50,248 INFO [train.py:938] (0/8) Epoch 29, validation: loss=0.1475, simple_loss=0.2594, pruned_loss=0.01778, over 944034.00 frames. 2023-05-02 18:30:50,249 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 18:30:55,887 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290207.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:32:11,212 INFO [train.py:904] (0/8) Epoch 29, batch 6050, loss[loss=0.2339, simple_loss=0.3025, pruned_loss=0.08265, over 11906.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2889, pruned_loss=0.05566, over 3090315.03 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:32:42,708 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290276.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:33:04,447 INFO [optim.py:368] (0/8) 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 18:33:22,103 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8798, 1.9997, 2.4891, 2.7911, 2.8019, 3.1958, 2.1739, 3.2208], device='cuda:0'), covar=tensor([0.0274, 0.0626, 0.0394, 0.0381, 0.0376, 0.0251, 0.0615, 0.0192], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0198, 0.0187, 0.0194, 0.0211, 0.0168, 0.0205, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:33:29,593 INFO [train.py:904] (0/8) Epoch 29, batch 6100, loss[loss=0.2185, simple_loss=0.2892, pruned_loss=0.07395, over 11851.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2885, pruned_loss=0.05471, over 3107735.67 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:33:30,414 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3350, 1.6404, 2.0260, 2.1995, 2.3484, 2.5499, 1.7190, 2.5343], device='cuda:0'), covar=tensor([0.0285, 0.0659, 0.0379, 0.0450, 0.0400, 0.0259, 0.0677, 0.0179], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0199, 0.0188, 0.0194, 0.0211, 0.0168, 0.0205, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 18:33:53,039 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=290318.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:34:17,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3977, 1.7234, 2.1212, 2.3218, 2.4716, 2.7061, 1.8964, 2.6517], device='cuda:0'), covar=tensor([0.0268, 0.0585, 0.0363, 0.0455, 0.0384, 0.0241, 0.0600, 0.0195], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0199, 0.0188, 0.0194, 0.0211, 0.0168, 0.0205, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 18:34:31,499 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1361, 2.1820, 2.2910, 3.9434, 2.0921, 2.4485, 2.2482, 2.3687], device='cuda:0'), covar=tensor([0.1862, 0.4224, 0.3442, 0.0723, 0.5329, 0.3191, 0.3914, 0.4135], device='cuda:0'), in_proj_covar=tensor([0.0423, 0.0476, 0.0387, 0.0338, 0.0446, 0.0547, 0.0449, 0.0559], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:34:46,617 INFO [train.py:904] (0/8) Epoch 29, batch 6150, loss[loss=0.1806, simple_loss=0.2645, pruned_loss=0.04836, over 16186.00 frames. ], tot_loss[loss=0.1969, simple_loss=0.2859, pruned_loss=0.05395, over 3118416.79 frames. ], batch size: 35, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:34:47,116 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290354.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:27,356 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=290379.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:35:44,984 INFO [optim.py:368] (0/8) 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 18:36:04,352 INFO [train.py:904] (0/8) Epoch 29, batch 6200, loss[loss=0.2009, simple_loss=0.2833, pruned_loss=0.05924, over 15468.00 frames. ], tot_loss[loss=0.1956, simple_loss=0.2841, pruned_loss=0.05351, over 3118002.97 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:36:08,278 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290406.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:36:48,243 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7429, 3.5125, 4.0084, 2.0558, 4.1860, 4.1347, 3.1560, 3.1844], device='cuda:0'), covar=tensor([0.0799, 0.0305, 0.0206, 0.1221, 0.0071, 0.0201, 0.0421, 0.0502], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0111, 0.0103, 0.0139, 0.0088, 0.0133, 0.0130, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 18:37:21,107 INFO [train.py:904] (0/8) Epoch 29, batch 6250, loss[loss=0.1765, simple_loss=0.2781, pruned_loss=0.03741, over 16669.00 frames. ], tot_loss[loss=0.1949, simple_loss=0.2832, pruned_loss=0.05333, over 3105135.16 frames. ], batch size: 62, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:38:07,145 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 18:38:15,681 INFO [optim.py:368] (0/8) 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 18:38:34,384 INFO [train.py:904] (0/8) Epoch 29, batch 6300, loss[loss=0.1812, simple_loss=0.269, pruned_loss=0.04669, over 16204.00 frames. ], tot_loss[loss=0.195, simple_loss=0.2834, pruned_loss=0.05334, over 3102744.23 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:39:34,492 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-02 18:39:52,810 INFO [train.py:904] (0/8) Epoch 29, batch 6350, loss[loss=0.1843, simple_loss=0.2672, pruned_loss=0.0507, over 16974.00 frames. ], tot_loss[loss=0.1966, simple_loss=0.2844, pruned_loss=0.05438, over 3092784.91 frames. ], batch size: 41, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:40:28,168 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290576.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 18:40:50,036 INFO [optim.py:368] (0/8) 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 18:41:02,375 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5092, 3.5712, 3.3279, 3.0132, 3.2237, 3.4515, 3.3166, 3.3052], device='cuda:0'), covar=tensor([0.0602, 0.0696, 0.0297, 0.0296, 0.0509, 0.0505, 0.1232, 0.0515], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0473, 0.0365, 0.0366, 0.0361, 0.0422, 0.0250, 0.0437], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:41:09,303 INFO [train.py:904] (0/8) Epoch 29, batch 6400, loss[loss=0.1897, simple_loss=0.2825, pruned_loss=0.04848, over 16560.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2853, pruned_loss=0.05576, over 3089097.56 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:41:24,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4178, 3.4764, 2.1129, 3.8902, 2.6418, 3.8637, 2.2504, 2.7256], device='cuda:0'), covar=tensor([0.0343, 0.0454, 0.1827, 0.0223, 0.0922, 0.0559, 0.1682, 0.0899], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0180, 0.0196, 0.0174, 0.0180, 0.0221, 0.0204, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 18:41:39,259 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290624.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 18:42:23,291 INFO [train.py:904] (0/8) Epoch 29, batch 6450, loss[loss=0.1844, simple_loss=0.2826, pruned_loss=0.04311, over 16895.00 frames. ], tot_loss[loss=0.1984, simple_loss=0.2857, pruned_loss=0.0556, over 3085151.77 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:42:23,815 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290654.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:42:30,235 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.84 vs. limit=5.0 2023-05-02 18:42:53,371 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=290674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:19,916 INFO [optim.py:368] (0/8) 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 18:43:37,904 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290702.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:43:40,050 INFO [train.py:904] (0/8) Epoch 29, batch 6500, loss[loss=0.2245, simple_loss=0.29, pruned_loss=0.07952, over 11543.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2845, pruned_loss=0.05547, over 3100677.29 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:43:40,694 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8561, 4.9067, 4.7324, 4.3950, 4.4218, 4.8298, 4.6184, 4.5132], device='cuda:0'), covar=tensor([0.0660, 0.0770, 0.0333, 0.0347, 0.0990, 0.0556, 0.0575, 0.0730], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0473, 0.0365, 0.0366, 0.0362, 0.0422, 0.0251, 0.0439], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:43:44,310 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290706.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:44:59,517 INFO [train.py:904] (0/8) Epoch 29, batch 6550, loss[loss=0.2054, simple_loss=0.3018, pruned_loss=0.05449, over 15518.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.287, pruned_loss=0.05577, over 3104795.09 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:44:59,830 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=290754.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:45:23,573 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0223, 5.5417, 5.7420, 5.4000, 5.5105, 6.0588, 5.5265, 5.2477], device='cuda:0'), covar=tensor([0.1030, 0.1958, 0.2679, 0.2174, 0.2410, 0.0972, 0.1675, 0.2430], device='cuda:0'), in_proj_covar=tensor([0.0435, 0.0645, 0.0715, 0.0521, 0.0699, 0.0733, 0.0554, 0.0699], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 18:45:23,692 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2501, 5.2837, 5.0800, 4.7065, 4.7410, 5.1753, 4.9921, 4.8686], device='cuda:0'), covar=tensor([0.0698, 0.1061, 0.0343, 0.0357, 0.0971, 0.0535, 0.0589, 0.0710], device='cuda:0'), in_proj_covar=tensor([0.0311, 0.0473, 0.0365, 0.0366, 0.0362, 0.0422, 0.0250, 0.0438], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:45:40,992 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0543, 2.3199, 2.4355, 2.8120, 1.8775, 3.1809, 1.8816, 2.7263], device='cuda:0'), covar=tensor([0.1213, 0.0672, 0.1006, 0.0224, 0.0111, 0.0394, 0.1484, 0.0724], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0182, 0.0203, 0.0206, 0.0208, 0.0220, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 18:45:58,162 INFO [optim.py:368] (0/8) 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 18:46:19,474 INFO [train.py:904] (0/8) Epoch 29, batch 6600, loss[loss=0.2354, simple_loss=0.3067, pruned_loss=0.08202, over 11463.00 frames. ], tot_loss[loss=0.2006, simple_loss=0.2886, pruned_loss=0.05625, over 3097498.78 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:47:38,823 INFO [train.py:904] (0/8) Epoch 29, batch 6650, loss[loss=0.1928, simple_loss=0.2838, pruned_loss=0.05092, over 16191.00 frames. ], tot_loss[loss=0.2012, simple_loss=0.2886, pruned_loss=0.05687, over 3088409.29 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:48:04,693 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6616, 3.7573, 2.8857, 2.3048, 2.4318, 2.4449, 4.0422, 3.2824], device='cuda:0'), covar=tensor([0.3025, 0.0676, 0.1893, 0.2732, 0.2743, 0.2227, 0.0450, 0.1447], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0276, 0.0314, 0.0330, 0.0308, 0.0279, 0.0306, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 18:48:35,971 INFO [optim.py:368] (0/8) 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 18:48:36,760 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 18:48:55,757 INFO [train.py:904] (0/8) Epoch 29, batch 6700, loss[loss=0.195, simple_loss=0.2819, pruned_loss=0.05402, over 16810.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.287, pruned_loss=0.05671, over 3081284.37 frames. ], batch size: 96, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:49:01,057 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6038, 2.4621, 2.3325, 3.6037, 2.3025, 3.7144, 1.4033, 2.6948], device='cuda:0'), covar=tensor([0.1587, 0.0946, 0.1470, 0.0237, 0.0249, 0.0465, 0.2089, 0.1002], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0182, 0.0202, 0.0206, 0.0208, 0.0220, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 18:50:12,974 INFO [train.py:904] (0/8) Epoch 29, batch 6750, loss[loss=0.194, simple_loss=0.2816, pruned_loss=0.05321, over 16311.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.286, pruned_loss=0.05637, over 3092429.55 frames. ], batch size: 165, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:50:43,280 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=290974.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:51:10,933 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.985e+02 2.623e+02 3.095e+02 3.882e+02 8.048e+02, threshold=6.190e+02, percent-clipped=2.0 2023-05-02 18:51:28,430 INFO [train.py:904] (0/8) Epoch 29, batch 6800, loss[loss=0.2363, simple_loss=0.3043, pruned_loss=0.08416, over 11449.00 frames. ], tot_loss[loss=0.2004, simple_loss=0.2867, pruned_loss=0.05701, over 3085911.00 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:51:57,166 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291022.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:52:33,791 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 18:52:46,137 INFO [train.py:904] (0/8) Epoch 29, batch 6850, loss[loss=0.2345, simple_loss=0.3034, pruned_loss=0.08284, over 11493.00 frames. ], tot_loss[loss=0.2014, simple_loss=0.2879, pruned_loss=0.05742, over 3087203.67 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:53:04,549 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8751, 2.6678, 2.5832, 1.8964, 2.5710, 2.6912, 2.5770, 1.7910], device='cuda:0'), covar=tensor([0.0515, 0.0122, 0.0124, 0.0441, 0.0173, 0.0187, 0.0157, 0.0523], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0103, 0.0117, 0.0099, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 18:53:09,363 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291069.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:53:42,658 INFO [optim.py:368] (0/8) 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 18:54:02,737 INFO [train.py:904] (0/8) Epoch 29, batch 6900, loss[loss=0.2024, simple_loss=0.288, pruned_loss=0.05835, over 16689.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.2899, pruned_loss=0.05689, over 3098104.31 frames. ], batch size: 134, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:54:44,507 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291130.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 18:54:49,550 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 18:55:20,068 INFO [train.py:904] (0/8) Epoch 29, batch 6950, loss[loss=0.2, simple_loss=0.2875, pruned_loss=0.05626, over 16444.00 frames. ], tot_loss[loss=0.2043, simple_loss=0.2914, pruned_loss=0.05857, over 3075167.89 frames. ], batch size: 146, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:56:18,598 INFO [optim.py:368] (0/8) 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 18:56:36,633 INFO [train.py:904] (0/8) Epoch 29, batch 7000, loss[loss=0.2117, simple_loss=0.2886, pruned_loss=0.06738, over 11480.00 frames. ], tot_loss[loss=0.2028, simple_loss=0.291, pruned_loss=0.05726, over 3081194.25 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:57:13,566 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7646, 4.6203, 4.7876, 4.9713, 5.1516, 4.6451, 5.1513, 5.1663], device='cuda:0'), covar=tensor([0.2028, 0.1384, 0.1767, 0.0851, 0.0709, 0.1075, 0.0776, 0.0695], device='cuda:0'), in_proj_covar=tensor([0.0678, 0.0829, 0.0954, 0.0840, 0.0641, 0.0662, 0.0704, 0.0816], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:57:15,500 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7433, 2.7831, 2.3405, 2.5299, 3.1885, 2.7915, 3.2336, 3.3833], device='cuda:0'), covar=tensor([0.0167, 0.0521, 0.0647, 0.0567, 0.0318, 0.0477, 0.0300, 0.0298], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0243, 0.0233, 0.0233, 0.0244, 0.0242, 0.0238, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:57:25,085 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7645, 4.0885, 3.0369, 2.4534, 2.8114, 2.6438, 4.3922, 3.6611], device='cuda:0'), covar=tensor([0.2954, 0.0614, 0.1909, 0.2741, 0.2497, 0.2045, 0.0459, 0.1209], device='cuda:0'), in_proj_covar=tensor([0.0337, 0.0275, 0.0315, 0.0329, 0.0307, 0.0279, 0.0306, 0.0352], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 18:57:50,290 INFO [train.py:904] (0/8) Epoch 29, batch 7050, loss[loss=0.1946, simple_loss=0.2829, pruned_loss=0.05317, over 16447.00 frames. ], tot_loss[loss=0.2027, simple_loss=0.2916, pruned_loss=0.05687, over 3091715.44 frames. ], batch size: 68, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 18:57:59,129 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5863, 4.5513, 4.4282, 3.1331, 4.4851, 1.5113, 4.1366, 3.9647], device='cuda:0'), covar=tensor([0.0207, 0.0187, 0.0308, 0.0848, 0.0180, 0.4138, 0.0265, 0.0533], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0177, 0.0215, 0.0187, 0.0192, 0.0219, 0.0203, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 18:58:03,570 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 18:58:49,446 INFO [optim.py:368] (0/8) 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 18:59:07,333 INFO [train.py:904] (0/8) Epoch 29, batch 7100, loss[loss=0.1823, simple_loss=0.2807, pruned_loss=0.04194, over 16544.00 frames. ], tot_loss[loss=0.2018, simple_loss=0.29, pruned_loss=0.05675, over 3062364.09 frames. ], batch size: 75, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:00:25,901 INFO [train.py:904] (0/8) Epoch 29, batch 7150, loss[loss=0.2339, simple_loss=0.2985, pruned_loss=0.08465, over 11452.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2883, pruned_loss=0.05689, over 3061237.66 frames. ], batch size: 247, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:01:23,567 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.798e+02 3.396e+02 4.128e+02 9.099e+02, threshold=6.792e+02, percent-clipped=1.0 2023-05-02 19:01:41,836 INFO [train.py:904] (0/8) Epoch 29, batch 7200, loss[loss=0.1691, simple_loss=0.2633, pruned_loss=0.03742, over 15445.00 frames. ], tot_loss[loss=0.1995, simple_loss=0.2868, pruned_loss=0.05611, over 3041099.48 frames. ], batch size: 191, lr: 2.31e-03, grad_scale: 8.0 2023-05-02 19:02:14,456 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=291425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:03:00,076 INFO [train.py:904] (0/8) Epoch 29, batch 7250, loss[loss=0.1802, simple_loss=0.2642, pruned_loss=0.04813, over 16800.00 frames. ], tot_loss[loss=0.197, simple_loss=0.2843, pruned_loss=0.05481, over 3044063.06 frames. ], batch size: 124, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:03:58,892 INFO [optim.py:368] (0/8) 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 19:04:15,823 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.33 vs. limit=5.0 2023-05-02 19:04:16,117 INFO [train.py:904] (0/8) Epoch 29, batch 7300, loss[loss=0.2254, simple_loss=0.2951, pruned_loss=0.07787, over 11673.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2836, pruned_loss=0.0544, over 3034625.36 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:05:03,069 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.92 vs. limit=5.0 2023-05-02 19:05:32,038 INFO [train.py:904] (0/8) Epoch 29, batch 7350, loss[loss=0.2211, simple_loss=0.2921, pruned_loss=0.07501, over 11332.00 frames. ], tot_loss[loss=0.198, simple_loss=0.285, pruned_loss=0.05552, over 3016453.33 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:06:32,146 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.793e+02 2.867e+02 3.369e+02 3.998e+02 9.840e+02, threshold=6.738e+02, percent-clipped=8.0 2023-05-02 19:06:49,564 INFO [train.py:904] (0/8) Epoch 29, batch 7400, loss[loss=0.1726, simple_loss=0.264, pruned_loss=0.04066, over 16777.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2851, pruned_loss=0.05553, over 3021790.92 frames. ], batch size: 39, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:08:07,280 INFO [train.py:904] (0/8) Epoch 29, batch 7450, loss[loss=0.2212, simple_loss=0.2931, pruned_loss=0.07463, over 11533.00 frames. ], tot_loss[loss=0.1992, simple_loss=0.2857, pruned_loss=0.05632, over 3036681.95 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:08:07,787 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3200, 3.7108, 3.8226, 2.3202, 3.4708, 3.8505, 3.5250, 2.0955], device='cuda:0'), covar=tensor([0.0583, 0.0068, 0.0062, 0.0505, 0.0117, 0.0113, 0.0103, 0.0532], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0090, 0.0092, 0.0136, 0.0103, 0.0117, 0.0099, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 19:09:10,903 INFO [optim.py:368] (0/8) 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 19:09:28,125 INFO [train.py:904] (0/8) Epoch 29, batch 7500, loss[loss=0.1902, simple_loss=0.2748, pruned_loss=0.05283, over 15300.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2856, pruned_loss=0.05534, over 3038102.42 frames. ], batch size: 190, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:10:01,627 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=291725.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:10:29,642 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 19:10:45,772 INFO [train.py:904] (0/8) Epoch 29, batch 7550, loss[loss=0.2241, simple_loss=0.2972, pruned_loss=0.07548, over 11600.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2856, pruned_loss=0.05627, over 3026160.93 frames. ], batch size: 248, lr: 2.31e-03, grad_scale: 4.0 2023-05-02 19:11:15,344 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=291773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:19,699 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=291776.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:11:44,718 INFO [optim.py:368] (0/8) 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 19:12:01,406 INFO [train.py:904] (0/8) Epoch 29, batch 7600, loss[loss=0.2077, simple_loss=0.2926, pruned_loss=0.0614, over 15279.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2844, pruned_loss=0.05598, over 3046839.47 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:12:52,167 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0328, 4.1227, 2.7040, 4.9794, 3.3813, 4.8332, 2.9792, 3.4862], device='cuda:0'), covar=tensor([0.0301, 0.0411, 0.1595, 0.0227, 0.0739, 0.0514, 0.1386, 0.0752], device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0181, 0.0197, 0.0174, 0.0181, 0.0222, 0.0205, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 19:12:52,190 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=291837.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:13:18,977 INFO [train.py:904] (0/8) Epoch 29, batch 7650, loss[loss=0.1825, simple_loss=0.2781, pruned_loss=0.04348, over 16777.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2839, pruned_loss=0.0554, over 3072980.15 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:13:57,746 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.7123, 6.0898, 5.7825, 5.8840, 5.4132, 5.4327, 5.3942, 6.1957], device='cuda:0'), covar=tensor([0.1370, 0.0793, 0.0998, 0.0925, 0.0907, 0.0670, 0.1355, 0.0871], device='cuda:0'), in_proj_covar=tensor([0.0722, 0.0866, 0.0715, 0.0675, 0.0553, 0.0556, 0.0726, 0.0682], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:14:08,711 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 19:14:20,911 INFO [optim.py:368] (0/8) 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 19:14:36,029 INFO [train.py:904] (0/8) Epoch 29, batch 7700, loss[loss=0.1739, simple_loss=0.2715, pruned_loss=0.03817, over 16715.00 frames. ], tot_loss[loss=0.1981, simple_loss=0.2841, pruned_loss=0.05604, over 3037994.81 frames. ], batch size: 89, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:15:53,503 INFO [train.py:904] (0/8) Epoch 29, batch 7750, loss[loss=0.1894, simple_loss=0.2785, pruned_loss=0.05012, over 16736.00 frames. ], tot_loss[loss=0.1975, simple_loss=0.2842, pruned_loss=0.05539, over 3062824.07 frames. ], batch size: 39, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:16:42,236 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8483, 1.4328, 1.7101, 1.6364, 1.8286, 1.8605, 1.5832, 1.7918], device='cuda:0'), covar=tensor([0.0267, 0.0425, 0.0233, 0.0309, 0.0304, 0.0197, 0.0488, 0.0177], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0196, 0.0185, 0.0190, 0.0208, 0.0166, 0.0203, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:16:55,758 INFO [optim.py:368] (0/8) 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 19:17:05,173 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-292000.pt 2023-05-02 19:17:14,582 INFO [train.py:904] (0/8) Epoch 29, batch 7800, loss[loss=0.2441, simple_loss=0.3101, pruned_loss=0.08905, over 11538.00 frames. ], tot_loss[loss=0.1979, simple_loss=0.2847, pruned_loss=0.05556, over 3066440.39 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:17:19,410 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5214, 3.5132, 3.4867, 2.6976, 3.3692, 2.1690, 3.1724, 2.8145], device='cuda:0'), covar=tensor([0.0205, 0.0182, 0.0214, 0.0241, 0.0133, 0.2430, 0.0168, 0.0284], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0175, 0.0215, 0.0186, 0.0190, 0.0219, 0.0201, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:18:02,373 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 19:18:28,640 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1372, 4.9060, 5.1134, 5.3016, 5.5329, 4.8914, 5.5144, 5.5103], device='cuda:0'), covar=tensor([0.2069, 0.1455, 0.1904, 0.0848, 0.0629, 0.1033, 0.0658, 0.0734], device='cuda:0'), in_proj_covar=tensor([0.0675, 0.0823, 0.0947, 0.0837, 0.0639, 0.0656, 0.0701, 0.0812], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:18:30,018 INFO [train.py:904] (0/8) Epoch 29, batch 7850, loss[loss=0.2042, simple_loss=0.2955, pruned_loss=0.05644, over 16912.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2859, pruned_loss=0.05564, over 3074822.68 frames. ], batch size: 42, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:18:30,453 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292054.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:19:02,722 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7839, 1.8422, 1.6143, 1.4658, 1.9630, 1.6061, 1.5940, 1.9695], device='cuda:0'), covar=tensor([0.0226, 0.0347, 0.0495, 0.0421, 0.0275, 0.0317, 0.0202, 0.0245], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0244, 0.0234, 0.0234, 0.0245, 0.0242, 0.0240, 0.0243], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:19:30,101 INFO [optim.py:368] (0/8) 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 19:19:40,171 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7542, 5.0645, 4.8430, 4.8225, 4.5774, 4.5646, 4.4682, 5.1379], device='cuda:0'), covar=tensor([0.1278, 0.0830, 0.0938, 0.0912, 0.0830, 0.1153, 0.1158, 0.0832], device='cuda:0'), in_proj_covar=tensor([0.0721, 0.0865, 0.0714, 0.0674, 0.0553, 0.0555, 0.0725, 0.0682], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:19:43,456 INFO [train.py:904] (0/8) Epoch 29, batch 7900, loss[loss=0.2354, simple_loss=0.2981, pruned_loss=0.08637, over 11437.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2848, pruned_loss=0.05489, over 3079026.34 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:19:59,934 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=292115.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:20:26,296 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292132.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:21:01,250 INFO [train.py:904] (0/8) Epoch 29, batch 7950, loss[loss=0.2586, simple_loss=0.3087, pruned_loss=0.1043, over 11569.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.286, pruned_loss=0.05606, over 3074017.02 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 2.0 2023-05-02 19:21:47,571 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2275, 4.2154, 4.1069, 3.2254, 4.1581, 1.9023, 3.9481, 3.7182], device='cuda:0'), covar=tensor([0.0144, 0.0136, 0.0229, 0.0384, 0.0113, 0.2904, 0.0172, 0.0320], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0175, 0.0214, 0.0185, 0.0190, 0.0218, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:22:03,663 INFO [optim.py:368] (0/8) 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 19:22:18,484 INFO [train.py:904] (0/8) Epoch 29, batch 8000, loss[loss=0.1914, simple_loss=0.2768, pruned_loss=0.05298, over 15263.00 frames. ], tot_loss[loss=0.2005, simple_loss=0.287, pruned_loss=0.057, over 3060056.16 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:22:44,586 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.23 vs. limit=5.0 2023-05-02 19:22:48,248 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5072, 4.6908, 4.8338, 4.6251, 4.6646, 5.1771, 4.7158, 4.4601], device='cuda:0'), covar=tensor([0.1422, 0.1849, 0.2182, 0.2003, 0.2345, 0.1015, 0.1641, 0.2425], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0640, 0.0708, 0.0519, 0.0693, 0.0730, 0.0550, 0.0693], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 19:22:52,227 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.32 vs. limit=2.0 2023-05-02 19:23:19,912 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8767, 3.1781, 3.4036, 2.0776, 2.9311, 2.2454, 3.3349, 3.5014], device='cuda:0'), covar=tensor([0.0386, 0.0948, 0.0653, 0.2327, 0.0988, 0.1128, 0.0875, 0.1078], device='cuda:0'), in_proj_covar=tensor([0.0160, 0.0171, 0.0170, 0.0157, 0.0148, 0.0133, 0.0145, 0.0183], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 19:23:31,121 INFO [train.py:904] (0/8) Epoch 29, batch 8050, loss[loss=0.2566, simple_loss=0.3205, pruned_loss=0.09642, over 11839.00 frames. ], tot_loss[loss=0.2001, simple_loss=0.2868, pruned_loss=0.05674, over 3070988.90 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:24:32,484 INFO [optim.py:368] (0/8) 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 19:24:46,387 INFO [train.py:904] (0/8) Epoch 29, batch 8100, loss[loss=0.1868, simple_loss=0.2849, pruned_loss=0.04439, over 16733.00 frames. ], tot_loss[loss=0.1985, simple_loss=0.2857, pruned_loss=0.05563, over 3090151.03 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:25:34,157 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.80 vs. limit=2.0 2023-05-02 19:26:01,432 INFO [train.py:904] (0/8) Epoch 29, batch 8150, loss[loss=0.1728, simple_loss=0.2553, pruned_loss=0.04516, over 17025.00 frames. ], tot_loss[loss=0.1959, simple_loss=0.283, pruned_loss=0.05439, over 3114231.35 frames. ], batch size: 50, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:27:01,242 INFO [optim.py:368] (0/8) 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 19:27:15,071 INFO [train.py:904] (0/8) Epoch 29, batch 8200, loss[loss=0.2148, simple_loss=0.291, pruned_loss=0.06925, over 11799.00 frames. ], tot_loss[loss=0.194, simple_loss=0.2806, pruned_loss=0.05374, over 3119538.41 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:27:25,895 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=292410.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:28:00,110 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292432.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:28:21,793 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4059, 3.3437, 3.4420, 3.5232, 3.5787, 3.3168, 3.5329, 3.6363], device='cuda:0'), covar=tensor([0.1493, 0.1097, 0.1218, 0.0731, 0.0747, 0.2441, 0.1090, 0.1003], device='cuda:0'), in_proj_covar=tensor([0.0672, 0.0820, 0.0943, 0.0835, 0.0636, 0.0655, 0.0697, 0.0812], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:28:34,805 INFO [train.py:904] (0/8) Epoch 29, batch 8250, loss[loss=0.1719, simple_loss=0.2647, pruned_loss=0.03955, over 16428.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2791, pruned_loss=0.05141, over 3090611.62 frames. ], batch size: 75, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:29:17,990 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292480.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:29:18,198 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8262, 1.4582, 1.7451, 1.7137, 1.9052, 1.8724, 1.7624, 1.7792], device='cuda:0'), covar=tensor([0.0277, 0.0425, 0.0248, 0.0322, 0.0330, 0.0234, 0.0458, 0.0175], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0196, 0.0185, 0.0191, 0.0207, 0.0165, 0.0203, 0.0167], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:29:41,438 INFO [optim.py:368] (0/8) 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 19:29:43,408 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9955, 4.2599, 4.0976, 4.1357, 3.7864, 3.8786, 3.8543, 4.2601], device='cuda:0'), covar=tensor([0.1143, 0.0869, 0.1015, 0.0804, 0.0857, 0.1649, 0.0955, 0.0974], device='cuda:0'), in_proj_covar=tensor([0.0718, 0.0861, 0.0711, 0.0670, 0.0549, 0.0552, 0.0723, 0.0677], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:29:55,812 INFO [train.py:904] (0/8) Epoch 29, batch 8300, loss[loss=0.1571, simple_loss=0.2551, pruned_loss=0.02957, over 16522.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.2764, pruned_loss=0.04854, over 3074436.24 frames. ], batch size: 62, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:30:07,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6145, 2.6870, 2.3422, 2.5408, 3.0075, 2.7207, 3.0554, 3.2208], device='cuda:0'), covar=tensor([0.0170, 0.0468, 0.0608, 0.0479, 0.0310, 0.0419, 0.0303, 0.0282], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0241, 0.0231, 0.0231, 0.0241, 0.0239, 0.0237, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:31:15,569 INFO [train.py:904] (0/8) Epoch 29, batch 8350, loss[loss=0.2013, simple_loss=0.285, pruned_loss=0.05882, over 12352.00 frames. ], tot_loss[loss=0.1848, simple_loss=0.2761, pruned_loss=0.04679, over 3078089.20 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 4.0 2023-05-02 19:32:07,427 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4984, 4.4366, 4.2938, 3.4870, 4.3392, 1.7221, 4.1091, 3.9864], device='cuda:0'), covar=tensor([0.0115, 0.0108, 0.0216, 0.0345, 0.0110, 0.3068, 0.0145, 0.0272], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0175, 0.0214, 0.0185, 0.0190, 0.0219, 0.0202, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:32:20,504 INFO [optim.py:368] (0/8) 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 19:32:31,740 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2981, 1.6756, 2.0338, 2.2084, 2.4074, 2.4214, 1.8167, 2.4647], device='cuda:0'), covar=tensor([0.0283, 0.0561, 0.0389, 0.0381, 0.0368, 0.0285, 0.0614, 0.0186], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0196, 0.0185, 0.0190, 0.0206, 0.0165, 0.0202, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:32:36,091 INFO [train.py:904] (0/8) Epoch 29, batch 8400, loss[loss=0.1888, simple_loss=0.272, pruned_loss=0.05277, over 12357.00 frames. ], tot_loss[loss=0.1818, simple_loss=0.2739, pruned_loss=0.04488, over 3071438.93 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:33:56,335 INFO [train.py:904] (0/8) Epoch 29, batch 8450, loss[loss=0.1701, simple_loss=0.2684, pruned_loss=0.03592, over 16744.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2722, pruned_loss=0.04337, over 3072427.78 frames. ], batch size: 83, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:34:04,238 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4710, 3.6795, 3.6861, 2.7040, 3.3565, 3.7589, 3.5218, 2.3038], device='cuda:0'), covar=tensor([0.0499, 0.0078, 0.0067, 0.0364, 0.0123, 0.0107, 0.0089, 0.0504], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0133, 0.0102, 0.0115, 0.0098, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 19:34:28,920 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1930, 2.4900, 2.5833, 2.0353, 2.7885, 2.7597, 2.6058, 2.5463], device='cuda:0'), covar=tensor([0.0604, 0.0273, 0.0299, 0.0971, 0.0131, 0.0315, 0.0419, 0.0428], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0102, 0.0138, 0.0087, 0.0130, 0.0129, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 19:34:40,773 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0590, 4.1455, 3.9398, 3.6299, 3.7031, 4.0571, 3.7354, 3.8417], device='cuda:0'), covar=tensor([0.0595, 0.0735, 0.0350, 0.0339, 0.0639, 0.0544, 0.1097, 0.0638], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0468, 0.0361, 0.0361, 0.0356, 0.0416, 0.0249, 0.0432], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:35:03,485 INFO [optim.py:368] (0/8) 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 19:35:19,249 INFO [train.py:904] (0/8) Epoch 29, batch 8500, loss[loss=0.1602, simple_loss=0.2517, pruned_loss=0.0343, over 15310.00 frames. ], tot_loss[loss=0.176, simple_loss=0.269, pruned_loss=0.04146, over 3066558.80 frames. ], batch size: 190, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:35:28,942 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=292710.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:36:06,631 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9074, 2.7317, 2.6098, 1.9666, 2.4586, 2.7749, 2.6467, 1.9667], device='cuda:0'), covar=tensor([0.0478, 0.0115, 0.0103, 0.0378, 0.0186, 0.0134, 0.0121, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0133, 0.0101, 0.0114, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 19:36:26,519 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4452, 4.4049, 4.2177, 3.3694, 4.3040, 1.7493, 4.0635, 3.9565], device='cuda:0'), covar=tensor([0.0131, 0.0134, 0.0263, 0.0456, 0.0149, 0.3129, 0.0194, 0.0364], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0174, 0.0213, 0.0184, 0.0189, 0.0217, 0.0200, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 19:36:43,009 INFO [train.py:904] (0/8) Epoch 29, batch 8550, loss[loss=0.1768, simple_loss=0.2766, pruned_loss=0.03853, over 16794.00 frames. ], tot_loss[loss=0.1743, simple_loss=0.2673, pruned_loss=0.04067, over 3052337.94 frames. ], batch size: 76, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:36:52,470 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=292758.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:37:07,596 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9300, 2.7689, 2.6349, 1.9856, 2.5441, 2.8169, 2.6905, 1.9791], device='cuda:0'), covar=tensor([0.0437, 0.0113, 0.0097, 0.0371, 0.0175, 0.0126, 0.0118, 0.0445], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0133, 0.0101, 0.0114, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 19:37:07,690 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9259, 2.6731, 2.9809, 2.1507, 2.7241, 2.2014, 2.6585, 2.8009], device='cuda:0'), covar=tensor([0.0342, 0.0984, 0.0463, 0.1963, 0.0801, 0.0996, 0.0634, 0.0912], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0168, 0.0168, 0.0155, 0.0146, 0.0132, 0.0143, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 19:38:04,316 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.654e+02 2.162e+02 2.570e+02 3.257e+02 5.024e+02, threshold=5.139e+02, percent-clipped=0.0 2023-05-02 19:38:21,510 INFO [train.py:904] (0/8) Epoch 29, batch 8600, loss[loss=0.1722, simple_loss=0.2695, pruned_loss=0.03742, over 16201.00 frames. ], tot_loss[loss=0.1733, simple_loss=0.2671, pruned_loss=0.03978, over 3042514.46 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:40:01,876 INFO [train.py:904] (0/8) Epoch 29, batch 8650, loss[loss=0.1562, simple_loss=0.2548, pruned_loss=0.02881, over 12260.00 frames. ], tot_loss[loss=0.1716, simple_loss=0.2657, pruned_loss=0.03877, over 3030349.02 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:41:05,786 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5426, 4.5610, 4.8593, 4.8431, 4.8331, 4.5985, 4.5583, 4.5296], device='cuda:0'), covar=tensor([0.0337, 0.0596, 0.0402, 0.0398, 0.0479, 0.0400, 0.0827, 0.0431], device='cuda:0'), in_proj_covar=tensor([0.0432, 0.0491, 0.0471, 0.0435, 0.0517, 0.0496, 0.0572, 0.0399], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 19:41:31,360 INFO [optim.py:368] (0/8) 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 19:41:48,485 INFO [train.py:904] (0/8) Epoch 29, batch 8700, loss[loss=0.166, simple_loss=0.2521, pruned_loss=0.03996, over 12182.00 frames. ], tot_loss[loss=0.1688, simple_loss=0.2628, pruned_loss=0.03742, over 3046122.02 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:43:26,856 INFO [train.py:904] (0/8) Epoch 29, batch 8750, loss[loss=0.1501, simple_loss=0.2425, pruned_loss=0.02888, over 12822.00 frames. ], tot_loss[loss=0.1684, simple_loss=0.2626, pruned_loss=0.03705, over 3035953.54 frames. ], batch size: 250, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:43:40,085 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.74 vs. limit=2.0 2023-05-02 19:44:01,875 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=292967.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:44:12,397 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5240, 2.8942, 3.1885, 1.9219, 2.7404, 2.1140, 3.0267, 3.0817], device='cuda:0'), covar=tensor([0.0399, 0.1104, 0.0631, 0.2449, 0.1002, 0.1171, 0.0824, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0158, 0.0167, 0.0168, 0.0154, 0.0145, 0.0131, 0.0143, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 19:45:00,945 INFO [optim.py:368] (0/8) 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 19:45:20,136 INFO [train.py:904] (0/8) Epoch 29, batch 8800, loss[loss=0.1689, simple_loss=0.273, pruned_loss=0.03242, over 16866.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2607, pruned_loss=0.03587, over 3030484.48 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:46:11,636 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293028.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:47:05,851 INFO [train.py:904] (0/8) Epoch 29, batch 8850, loss[loss=0.179, simple_loss=0.2797, pruned_loss=0.03919, over 17039.00 frames. ], tot_loss[loss=0.1668, simple_loss=0.2634, pruned_loss=0.03513, over 3040126.56 frames. ], batch size: 55, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:48:05,766 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293081.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:48:07,734 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.27 vs. limit=2.0 2023-05-02 19:48:35,606 INFO [optim.py:368] (0/8) 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 19:48:55,723 INFO [train.py:904] (0/8) Epoch 29, batch 8900, loss[loss=0.1465, simple_loss=0.2489, pruned_loss=0.02208, over 16857.00 frames. ], tot_loss[loss=0.1667, simple_loss=0.2644, pruned_loss=0.03454, over 3056156.58 frames. ], batch size: 90, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:49:23,383 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293117.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:50:34,682 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293142.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:51:01,970 INFO [train.py:904] (0/8) Epoch 29, batch 8950, loss[loss=0.1682, simple_loss=0.2576, pruned_loss=0.03939, over 12774.00 frames. ], tot_loss[loss=0.1662, simple_loss=0.2633, pruned_loss=0.03451, over 3056842.18 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:51:52,419 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293178.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:52:32,342 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 29, batch 9000, loss[loss=0.1513, simple_loss=0.2471, pruned_loss=0.02773, over 16114.00 frames. ], tot_loss[loss=0.1633, simple_loss=0.26, pruned_loss=0.03337, over 3064394.92 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:52:53,147 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 19:53:02,751 INFO [train.py:938] (0/8) 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] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 19:54:47,965 INFO [train.py:904] (0/8) Epoch 29, batch 9050, loss[loss=0.1512, simple_loss=0.2454, pruned_loss=0.02852, over 15476.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2605, pruned_loss=0.03363, over 3067016.80 frames. ], batch size: 193, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:56:15,031 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.590e+02 2.113e+02 2.508e+02 2.955e+02 4.534e+02, threshold=5.015e+02, percent-clipped=0.0 2023-05-02 19:56:34,930 INFO [train.py:904] (0/8) Epoch 29, batch 9100, loss[loss=0.1647, simple_loss=0.2509, pruned_loss=0.03924, over 11994.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2602, pruned_loss=0.03415, over 3065594.90 frames. ], batch size: 249, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:57:17,193 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:34,122 INFO [train.py:904] (0/8) Epoch 29, batch 9150, loss[loss=0.1747, simple_loss=0.2603, pruned_loss=0.04457, over 12323.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2604, pruned_loss=0.03385, over 3049551.53 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 19:58:51,365 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293361.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:58:56,691 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293364.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 19:59:02,462 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9457, 2.0738, 2.5723, 2.8679, 2.7294, 3.2019, 2.3213, 3.3046], device='cuda:0'), covar=tensor([0.0258, 0.0623, 0.0378, 0.0372, 0.0409, 0.0274, 0.0597, 0.0190], device='cuda:0'), in_proj_covar=tensor([0.0193, 0.0193, 0.0182, 0.0187, 0.0205, 0.0162, 0.0200, 0.0163], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:00:04,476 INFO [optim.py:368] (0/8) 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 20:00:20,601 INFO [train.py:904] (0/8) Epoch 29, batch 9200, loss[loss=0.1523, simple_loss=0.2428, pruned_loss=0.03095, over 12257.00 frames. ], tot_loss[loss=0.1612, simple_loss=0.2563, pruned_loss=0.03308, over 3065352.90 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:00:54,553 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293422.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:00,072 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=293425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:23,103 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293437.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:01:58,362 INFO [train.py:904] (0/8) Epoch 29, batch 9250, loss[loss=0.171, simple_loss=0.2508, pruned_loss=0.04564, over 12059.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2559, pruned_loss=0.03283, over 3072930.62 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:02:15,146 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7036, 2.6045, 1.9459, 2.7710, 2.2038, 2.8152, 2.2228, 2.3899], device='cuda:0'), covar=tensor([0.0359, 0.0388, 0.1298, 0.0247, 0.0710, 0.0472, 0.1238, 0.0647], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0175, 0.0192, 0.0167, 0.0176, 0.0214, 0.0201, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 20:02:39,023 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:03:01,618 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.47 vs. limit=2.0 2023-05-02 20:03:28,486 INFO [optim.py:368] (0/8) 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 20:03:49,272 INFO [train.py:904] (0/8) Epoch 29, batch 9300, loss[loss=0.1462, simple_loss=0.2428, pruned_loss=0.02485, over 16427.00 frames. ], tot_loss[loss=0.1597, simple_loss=0.2544, pruned_loss=0.03245, over 3073924.17 frames. ], batch size: 146, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:05:33,323 INFO [train.py:904] (0/8) Epoch 29, batch 9350, loss[loss=0.1598, simple_loss=0.2542, pruned_loss=0.03271, over 15346.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2551, pruned_loss=0.03291, over 3083008.99 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:05:54,840 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6172, 2.5085, 2.4753, 4.4183, 2.3424, 2.9724, 2.5858, 2.7038], device='cuda:0'), covar=tensor([0.1194, 0.3560, 0.3227, 0.0452, 0.4235, 0.2622, 0.3698, 0.3421], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0467, 0.0380, 0.0329, 0.0438, 0.0533, 0.0440, 0.0546], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:06:56,974 INFO [optim.py:368] (0/8) 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 20:07:15,897 INFO [train.py:904] (0/8) Epoch 29, batch 9400, loss[loss=0.1609, simple_loss=0.2649, pruned_loss=0.02845, over 16604.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.255, pruned_loss=0.03293, over 3055825.66 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:07:53,836 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293623.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:08:55,061 INFO [train.py:904] (0/8) Epoch 29, batch 9450, loss[loss=0.1631, simple_loss=0.2618, pruned_loss=0.0322, over 16656.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2564, pruned_loss=0.03306, over 3038415.77 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:09:11,951 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 20:09:29,100 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:10:05,377 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.79 vs. limit=5.0 2023-05-02 20:10:18,288 INFO [optim.py:368] (0/8) 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 20:10:34,122 INFO [train.py:904] (0/8) Epoch 29, batch 9500, loss[loss=0.1582, simple_loss=0.2574, pruned_loss=0.02947, over 16240.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2559, pruned_loss=0.03284, over 3051902.36 frames. ], batch size: 165, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:11:02,448 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293717.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:08,424 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=293720.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:41,943 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293737.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:11:43,664 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7173, 3.0159, 3.3948, 2.0116, 2.9618, 2.2095, 3.2032, 3.1782], device='cuda:0'), covar=tensor([0.0300, 0.0890, 0.0556, 0.2179, 0.0768, 0.1088, 0.0670, 0.1037], device='cuda:0'), in_proj_covar=tensor([0.0156, 0.0164, 0.0166, 0.0153, 0.0144, 0.0130, 0.0141, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 20:12:19,487 INFO [train.py:904] (0/8) Epoch 29, batch 9550, loss[loss=0.1792, simple_loss=0.2596, pruned_loss=0.04941, over 12156.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2556, pruned_loss=0.03285, over 3041264.77 frames. ], batch size: 246, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:12:23,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0681, 2.1728, 2.1058, 3.8241, 2.1068, 2.5075, 2.2432, 2.3738], device='cuda:0'), covar=tensor([0.1522, 0.4134, 0.3636, 0.0615, 0.4664, 0.2930, 0.4215, 0.3836], device='cuda:0'), in_proj_covar=tensor([0.0413, 0.0466, 0.0380, 0.0328, 0.0438, 0.0533, 0.0441, 0.0546], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:12:54,745 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8660, 3.7661, 3.9077, 4.0145, 4.0993, 3.6804, 4.0830, 4.1434], device='cuda:0'), covar=tensor([0.1614, 0.1104, 0.1344, 0.0704, 0.0595, 0.1880, 0.0747, 0.0769], device='cuda:0'), in_proj_covar=tensor([0.0654, 0.0798, 0.0916, 0.0816, 0.0620, 0.0639, 0.0680, 0.0791], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:12:59,188 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=293773.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:24,954 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293785.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:13:35,633 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9501, 4.2257, 4.2396, 3.1513, 3.7074, 4.2588, 3.8881, 2.5222], device='cuda:0'), covar=tensor([0.0475, 0.0049, 0.0048, 0.0342, 0.0112, 0.0089, 0.0080, 0.0522], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0090, 0.0134, 0.0102, 0.0114, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 20:13:43,638 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.327e+02 2.273e+02 2.766e+02 3.522e+02 6.419e+02, threshold=5.532e+02, percent-clipped=5.0 2023-05-02 20:13:59,142 INFO [train.py:904] (0/8) Epoch 29, batch 9600, loss[loss=0.1729, simple_loss=0.2528, pruned_loss=0.04649, over 12093.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2577, pruned_loss=0.03372, over 3035644.78 frames. ], batch size: 247, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:14:32,483 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=293821.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:15:45,754 INFO [train.py:904] (0/8) Epoch 29, batch 9650, loss[loss=0.1655, simple_loss=0.2619, pruned_loss=0.03451, over 16711.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2596, pruned_loss=0.03399, over 3046005.07 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:17:16,479 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.571e+02 2.271e+02 2.644e+02 3.266e+02 5.877e+02, threshold=5.288e+02, percent-clipped=1.0 2023-05-02 20:17:35,797 INFO [train.py:904] (0/8) Epoch 29, batch 9700, loss[loss=0.1554, simple_loss=0.2511, pruned_loss=0.02987, over 16806.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2587, pruned_loss=0.03372, over 3049899.23 frames. ], batch size: 124, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:17:45,544 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 20:18:21,151 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0185, 4.2916, 4.1354, 4.1476, 3.8045, 3.8417, 3.8405, 4.2754], device='cuda:0'), covar=tensor([0.1068, 0.0868, 0.0913, 0.0839, 0.0838, 0.1831, 0.1067, 0.0934], device='cuda:0'), in_proj_covar=tensor([0.0704, 0.0848, 0.0696, 0.0659, 0.0540, 0.0541, 0.0709, 0.0664], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:19:17,223 INFO [train.py:904] (0/8) Epoch 29, batch 9750, loss[loss=0.13, simple_loss=0.2076, pruned_loss=0.02616, over 16290.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.257, pruned_loss=0.03371, over 3035120.99 frames. ], batch size: 35, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:20:03,130 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-02 20:20:24,556 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=293987.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:20:38,044 INFO [optim.py:368] (0/8) 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 20:20:47,055 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-294000.pt 2023-05-02 20:20:56,842 INFO [train.py:904] (0/8) Epoch 29, batch 9800, loss[loss=0.1509, simple_loss=0.2576, pruned_loss=0.02205, over 16904.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.2567, pruned_loss=0.03301, over 3032292.96 frames. ], batch size: 102, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:21:16,148 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1671, 5.6275, 5.8113, 5.5094, 5.6247, 6.1022, 5.6037, 5.2371], device='cuda:0'), covar=tensor([0.0900, 0.1706, 0.1727, 0.1802, 0.1929, 0.0774, 0.1501, 0.2516], device='cuda:0'), in_proj_covar=tensor([0.0412, 0.0619, 0.0689, 0.0503, 0.0672, 0.0710, 0.0534, 0.0669], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 20:21:22,358 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294017.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:27,402 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294020.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:21:36,443 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7982, 3.0812, 3.2431, 1.6936, 2.7783, 1.8890, 3.2951, 3.3174], device='cuda:0'), covar=tensor([0.0268, 0.1004, 0.0679, 0.2850, 0.1075, 0.1404, 0.0730, 0.0987], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0165, 0.0167, 0.0154, 0.0144, 0.0130, 0.0142, 0.0178], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 20:22:29,329 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 29, batch 9850, loss[loss=0.1558, simple_loss=0.2558, pruned_loss=0.02788, over 16972.00 frames. ], tot_loss[loss=0.1618, simple_loss=0.2581, pruned_loss=0.03279, over 3041535.93 frames. ], batch size: 109, lr: 2.30e-03, grad_scale: 16.0 2023-05-02 20:23:02,393 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294065.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:23:09,660 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:23:20,704 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5107, 4.9533, 4.8288, 3.7695, 4.1424, 4.8204, 4.2646, 3.2520], device='cuda:0'), covar=tensor([0.0403, 0.0029, 0.0037, 0.0288, 0.0104, 0.0085, 0.0070, 0.0390], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0088, 0.0089, 0.0133, 0.0101, 0.0113, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-02 20:24:14,912 INFO [optim.py:368] (0/8) 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 20:24:32,438 INFO [train.py:904] (0/8) Epoch 29, batch 9900, loss[loss=0.1747, simple_loss=0.2773, pruned_loss=0.03608, over 16724.00 frames. ], tot_loss[loss=0.1623, simple_loss=0.2585, pruned_loss=0.0331, over 3018942.13 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:26:27,781 INFO [train.py:904] (0/8) Epoch 29, batch 9950, loss[loss=0.1518, simple_loss=0.2514, pruned_loss=0.02604, over 16704.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2607, pruned_loss=0.03366, over 3015775.10 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:27:06,796 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:28:08,145 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.560e+02 2.053e+02 2.347e+02 2.830e+02 4.908e+02, threshold=4.695e+02, percent-clipped=0.0 2023-05-02 20:28:27,673 INFO [train.py:904] (0/8) Epoch 29, batch 10000, loss[loss=0.1914, simple_loss=0.2961, pruned_loss=0.04334, over 16735.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2592, pruned_loss=0.03303, over 3048400.59 frames. ], batch size: 134, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:28:28,506 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.5755, 2.3717, 2.3154, 3.8012, 2.0765, 3.7651, 1.5432, 2.7639], device='cuda:0'), covar=tensor([0.1544, 0.0904, 0.1332, 0.0175, 0.0129, 0.0366, 0.1818, 0.0834], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0179, 0.0198, 0.0199, 0.0201, 0.0214, 0.0209, 0.0196], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 20:29:21,506 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294231.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:30:08,537 INFO [train.py:904] (0/8) Epoch 29, batch 10050, loss[loss=0.166, simple_loss=0.2632, pruned_loss=0.03438, over 15348.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2597, pruned_loss=0.03311, over 3069137.71 frames. ], batch size: 191, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:31:27,888 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.622e+02 2.082e+02 2.498e+02 3.055e+02 5.920e+02, threshold=4.997e+02, percent-clipped=2.0 2023-05-02 20:31:41,194 INFO [train.py:904] (0/8) Epoch 29, batch 10100, loss[loss=0.1681, simple_loss=0.2552, pruned_loss=0.04047, over 12933.00 frames. ], tot_loss[loss=0.1632, simple_loss=0.2599, pruned_loss=0.03325, over 3056233.23 frames. ], batch size: 248, lr: 2.30e-03, grad_scale: 8.0 2023-05-02 20:31:59,086 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.16 vs. limit=5.0 2023-05-02 20:32:06,752 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9155, 2.2017, 2.4671, 3.1891, 2.2300, 2.3777, 2.3947, 2.2989], device='cuda:0'), covar=tensor([0.1445, 0.3784, 0.2867, 0.0755, 0.4545, 0.2687, 0.3545, 0.3816], device='cuda:0'), in_proj_covar=tensor([0.0410, 0.0462, 0.0378, 0.0325, 0.0435, 0.0527, 0.0436, 0.0541], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:32:42,467 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294339.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:32:46,800 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294343.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:32:58,928 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-29.pt 2023-05-02 20:33:20,747 INFO [train.py:904] (0/8) Epoch 30, batch 0, loss[loss=0.1795, simple_loss=0.2667, pruned_loss=0.04615, over 17055.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2667, pruned_loss=0.04615, over 17055.00 frames. ], batch size: 53, lr: 2.26e-03, grad_scale: 8.0 2023-05-02 20:33:20,747 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 20:33:28,214 INFO [train.py:938] (0/8) Epoch 30, validation: loss=0.1426, simple_loss=0.2458, pruned_loss=0.01974, over 944034.00 frames. 2023-05-02 20:33:28,215 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 20:34:13,985 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6487, 3.7331, 3.9626, 2.0115, 3.1447, 2.2186, 4.0488, 4.0729], device='cuda:0'), covar=tensor([0.0254, 0.0918, 0.0598, 0.2607, 0.1024, 0.1301, 0.0611, 0.1071], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0164, 0.0166, 0.0153, 0.0144, 0.0129, 0.0141, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 20:34:28,813 INFO [optim.py:368] (0/8) 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 20:34:29,593 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294400.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:34:36,217 INFO [train.py:904] (0/8) Epoch 30, batch 50, loss[loss=0.1819, simple_loss=0.2662, pruned_loss=0.04883, over 16298.00 frames. ], tot_loss[loss=0.1728, simple_loss=0.2613, pruned_loss=0.04217, over 748085.41 frames. ], batch size: 165, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:34:37,758 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8537, 2.9003, 3.1995, 2.2117, 2.7400, 2.0783, 3.4049, 3.3644], device='cuda:0'), covar=tensor([0.0258, 0.1027, 0.0659, 0.1985, 0.0938, 0.1182, 0.0617, 0.0932], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0164, 0.0166, 0.0154, 0.0144, 0.0130, 0.0141, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-02 20:35:01,515 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294423.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:35:43,895 INFO [train.py:904] (0/8) Epoch 30, batch 100, loss[loss=0.1919, simple_loss=0.2722, pruned_loss=0.05583, over 16509.00 frames. ], tot_loss[loss=0.172, simple_loss=0.2606, pruned_loss=0.04168, over 1316182.00 frames. ], batch size: 75, lr: 2.26e-03, grad_scale: 1.0 2023-05-02 20:36:24,527 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294484.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:36:32,208 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.508e+02 2.392e+02 2.778e+02 3.445e+02 1.255e+03, threshold=5.557e+02, percent-clipped=5.0 2023-05-02 20:36:51,465 INFO [train.py:904] (0/8) Epoch 30, batch 150, loss[loss=0.1909, simple_loss=0.2662, pruned_loss=0.0578, over 16669.00 frames. ], tot_loss[loss=0.169, simple_loss=0.2577, pruned_loss=0.04013, over 1768788.98 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:36:55,928 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1073, 5.4815, 5.2080, 5.1839, 4.9581, 4.9415, 4.9093, 5.5544], device='cuda:0'), covar=tensor([0.1468, 0.1014, 0.1223, 0.1106, 0.0930, 0.0990, 0.1316, 0.0970], device='cuda:0'), in_proj_covar=tensor([0.0708, 0.0854, 0.0701, 0.0664, 0.0545, 0.0543, 0.0715, 0.0670], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:37:18,792 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9263, 4.4673, 3.1477, 2.4344, 2.7319, 2.7739, 4.7530, 3.6068], device='cuda:0'), covar=tensor([0.3024, 0.0636, 0.1989, 0.3383, 0.3156, 0.2292, 0.0383, 0.1673], device='cuda:0'), in_proj_covar=tensor([0.0333, 0.0271, 0.0310, 0.0324, 0.0301, 0.0276, 0.0300, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 20:37:20,788 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294526.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:52,708 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294549.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:37:59,965 INFO [train.py:904] (0/8) Epoch 30, batch 200, loss[loss=0.2051, simple_loss=0.2956, pruned_loss=0.05733, over 12477.00 frames. ], tot_loss[loss=0.1695, simple_loss=0.2577, pruned_loss=0.04071, over 2114878.86 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:38:15,915 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 20:38:22,870 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294571.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:38:47,937 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.3724, 5.7415, 5.5085, 5.5271, 5.1951, 5.2213, 5.1082, 5.8753], device='cuda:0'), covar=tensor([0.1431, 0.0998, 0.1410, 0.0930, 0.0918, 0.0793, 0.1356, 0.0945], device='cuda:0'), in_proj_covar=tensor([0.0714, 0.0860, 0.0708, 0.0670, 0.0550, 0.0547, 0.0721, 0.0676], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:39:01,687 INFO [optim.py:368] (0/8) 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 20:39:06,658 INFO [train.py:904] (0/8) Epoch 30, batch 250, loss[loss=0.1297, simple_loss=0.2219, pruned_loss=0.0188, over 17226.00 frames. ], tot_loss[loss=0.1689, simple_loss=0.2562, pruned_loss=0.04079, over 2382965.04 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:39:45,422 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=294632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:01,699 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294643.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:40:15,316 INFO [train.py:904] (0/8) Epoch 30, batch 300, loss[loss=0.1899, simple_loss=0.2639, pruned_loss=0.05796, over 16757.00 frames. ], tot_loss[loss=0.1676, simple_loss=0.2549, pruned_loss=0.04012, over 2600339.60 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:40:15,663 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0389, 4.7029, 5.0297, 5.2043, 5.4343, 4.7728, 5.4428, 5.4119], device='cuda:0'), covar=tensor([0.2065, 0.1682, 0.2014, 0.0918, 0.0629, 0.0997, 0.0557, 0.0716], device='cuda:0'), in_proj_covar=tensor([0.0667, 0.0816, 0.0937, 0.0831, 0.0632, 0.0652, 0.0692, 0.0805], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:41:06,781 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294691.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:12,186 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294695.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:41:19,995 INFO [optim.py:368] (0/8) 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 20:41:25,727 INFO [train.py:904] (0/8) Epoch 30, batch 350, loss[loss=0.1635, simple_loss=0.2555, pruned_loss=0.03568, over 17058.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.253, pruned_loss=0.03929, over 2766044.26 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 1.0 2023-05-02 20:41:27,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.71 vs. limit=5.0 2023-05-02 20:42:20,321 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1294, 5.7385, 5.8469, 5.5450, 5.6605, 6.2249, 5.7059, 5.4369], device='cuda:0'), covar=tensor([0.1098, 0.1978, 0.2356, 0.2183, 0.2479, 0.0930, 0.1632, 0.2434], device='cuda:0'), in_proj_covar=tensor([0.0425, 0.0639, 0.0712, 0.0517, 0.0691, 0.0732, 0.0548, 0.0687], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 20:42:22,883 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7189, 4.0244, 2.9819, 2.4032, 2.5802, 2.6284, 4.4082, 3.3272], device='cuda:0'), covar=tensor([0.3262, 0.0768, 0.2089, 0.3107, 0.3053, 0.2325, 0.0420, 0.1672], device='cuda:0'), in_proj_covar=tensor([0.0336, 0.0274, 0.0314, 0.0328, 0.0304, 0.0279, 0.0303, 0.0351], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 20:42:33,019 INFO [train.py:904] (0/8) Epoch 30, batch 400, loss[loss=0.1419, simple_loss=0.2324, pruned_loss=0.02571, over 17227.00 frames. ], tot_loss[loss=0.1643, simple_loss=0.2512, pruned_loss=0.03869, over 2896204.00 frames. ], batch size: 44, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:42:50,100 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8455, 2.9333, 2.5845, 2.9138, 3.1811, 2.9161, 3.5075, 3.4285], device='cuda:0'), covar=tensor([0.0184, 0.0457, 0.0559, 0.0420, 0.0336, 0.0449, 0.0265, 0.0322], device='cuda:0'), in_proj_covar=tensor([0.0232, 0.0246, 0.0234, 0.0235, 0.0244, 0.0243, 0.0240, 0.0243], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:43:06,910 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294779.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:43:11,812 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5944, 4.5588, 4.5336, 3.9824, 4.5338, 1.8495, 4.3023, 4.1590], device='cuda:0'), covar=tensor([0.0181, 0.0130, 0.0199, 0.0335, 0.0128, 0.2806, 0.0206, 0.0263], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0173, 0.0210, 0.0181, 0.0188, 0.0217, 0.0199, 0.0177], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:43:34,533 INFO [optim.py:368] (0/8) 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 20:43:41,862 INFO [train.py:904] (0/8) Epoch 30, batch 450, loss[loss=0.1563, simple_loss=0.2554, pruned_loss=0.02859, over 17262.00 frames. ], tot_loss[loss=0.1627, simple_loss=0.2497, pruned_loss=0.0378, over 3002007.12 frames. ], batch size: 52, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:44:12,172 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294826.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:35,486 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:44:49,603 INFO [train.py:904] (0/8) Epoch 30, batch 500, loss[loss=0.1332, simple_loss=0.22, pruned_loss=0.02323, over 15897.00 frames. ], tot_loss[loss=0.1613, simple_loss=0.2485, pruned_loss=0.03707, over 3068313.03 frames. ], batch size: 35, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:45:17,701 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=294874.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:45:52,586 INFO [optim.py:368] (0/8) 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 20:45:57,305 INFO [train.py:904] (0/8) Epoch 30, batch 550, loss[loss=0.1884, simple_loss=0.2656, pruned_loss=0.05558, over 16789.00 frames. ], tot_loss[loss=0.1604, simple_loss=0.2476, pruned_loss=0.03663, over 3126954.23 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:46:29,025 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=294927.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:47:00,876 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 20:47:05,470 INFO [train.py:904] (0/8) Epoch 30, batch 600, loss[loss=0.1891, simple_loss=0.2721, pruned_loss=0.05309, over 16999.00 frames. ], tot_loss[loss=0.161, simple_loss=0.2474, pruned_loss=0.03727, over 3172119.52 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:47:52,941 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=294989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:01,956 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=294995.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:48:07,377 INFO [optim.py:368] (0/8) 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 20:48:13,119 INFO [train.py:904] (0/8) Epoch 30, batch 650, loss[loss=0.1599, simple_loss=0.2576, pruned_loss=0.03108, over 17125.00 frames. ], tot_loss[loss=0.1601, simple_loss=0.2463, pruned_loss=0.03695, over 3210533.91 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:48:44,717 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 20:49:04,908 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295043.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:15,113 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295050.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 20:49:20,779 INFO [train.py:904] (0/8) Epoch 30, batch 700, loss[loss=0.1739, simple_loss=0.2511, pruned_loss=0.04836, over 16567.00 frames. ], tot_loss[loss=0.1594, simple_loss=0.2458, pruned_loss=0.03651, over 3234673.78 frames. ], batch size: 146, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:49:39,367 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295068.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:49:49,903 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.53 vs. limit=2.0 2023-05-02 20:49:54,495 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295079.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:50:23,318 INFO [optim.py:368] (0/8) 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 20:50:27,893 INFO [train.py:904] (0/8) Epoch 30, batch 750, loss[loss=0.1365, simple_loss=0.2324, pruned_loss=0.02034, over 17170.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2467, pruned_loss=0.03722, over 3251912.68 frames. ], batch size: 46, lr: 2.25e-03, grad_scale: 2.0 2023-05-02 20:50:31,112 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.87 vs. limit=5.0 2023-05-02 20:51:00,103 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295127.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:03,386 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295129.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:24,616 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295144.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:51:38,206 INFO [train.py:904] (0/8) Epoch 30, batch 800, loss[loss=0.1639, simple_loss=0.2476, pruned_loss=0.0401, over 16284.00 frames. ], tot_loss[loss=0.1602, simple_loss=0.2462, pruned_loss=0.03707, over 3263171.89 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:52:29,859 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295192.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:52:30,328 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.09 vs. limit=2.0 2023-05-02 20:52:42,513 INFO [optim.py:368] (0/8) 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 20:52:44,584 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4123, 4.3824, 4.3360, 3.7823, 4.3719, 1.8175, 4.1253, 3.8399], device='cuda:0'), covar=tensor([0.0177, 0.0115, 0.0202, 0.0298, 0.0123, 0.3043, 0.0159, 0.0299], device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0175, 0.0213, 0.0183, 0.0191, 0.0219, 0.0201, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:52:48,759 INFO [train.py:904] (0/8) Epoch 30, batch 850, loss[loss=0.1526, simple_loss=0.2326, pruned_loss=0.03632, over 16651.00 frames. ], tot_loss[loss=0.1595, simple_loss=0.2454, pruned_loss=0.03682, over 3262748.44 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:53:18,787 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295225.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:21,266 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295227.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:53:39,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4885, 5.8431, 5.5965, 5.7019, 5.2823, 5.3637, 5.2028, 5.9474], device='cuda:0'), covar=tensor([0.1476, 0.1017, 0.1186, 0.0899, 0.0956, 0.0727, 0.1377, 0.1054], device='cuda:0'), in_proj_covar=tensor([0.0730, 0.0878, 0.0724, 0.0684, 0.0561, 0.0559, 0.0741, 0.0692], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:53:48,687 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 20:53:56,733 INFO [train.py:904] (0/8) Epoch 30, batch 900, loss[loss=0.1523, simple_loss=0.231, pruned_loss=0.03675, over 15510.00 frames. ], tot_loss[loss=0.1586, simple_loss=0.2447, pruned_loss=0.03622, over 3275892.74 frames. ], batch size: 191, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:54:27,486 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295275.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:33,472 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295279.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:43,048 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295286.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:54:52,825 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1521, 4.0311, 4.2104, 4.3313, 4.3989, 3.9955, 4.2468, 4.4032], device='cuda:0'), covar=tensor([0.1673, 0.1168, 0.1291, 0.0687, 0.0628, 0.1410, 0.2348, 0.0890], device='cuda:0'), in_proj_covar=tensor([0.0694, 0.0845, 0.0975, 0.0861, 0.0654, 0.0677, 0.0717, 0.0834], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:55:02,823 INFO [optim.py:368] (0/8) 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 20:55:08,081 INFO [train.py:904] (0/8) Epoch 30, batch 950, loss[loss=0.1495, simple_loss=0.2236, pruned_loss=0.03769, over 16676.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2449, pruned_loss=0.03623, over 3279980.47 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:55:49,216 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295333.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:55:59,194 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295340.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:06,346 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295345.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 20:56:17,471 INFO [train.py:904] (0/8) Epoch 30, batch 1000, loss[loss=0.1536, simple_loss=0.2494, pruned_loss=0.0289, over 17126.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2446, pruned_loss=0.03619, over 3299416.50 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:56:24,400 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295359.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:56:33,027 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.27 vs. limit=5.0 2023-05-02 20:56:40,345 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4547, 3.5716, 3.6800, 3.6697, 3.6794, 3.5229, 3.5388, 3.5473], device='cuda:0'), covar=tensor([0.0433, 0.0904, 0.0575, 0.0486, 0.0580, 0.0643, 0.0790, 0.0576], device='cuda:0'), in_proj_covar=tensor([0.0443, 0.0506, 0.0484, 0.0448, 0.0530, 0.0513, 0.0588, 0.0412], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 20:57:12,955 INFO [zipformer.py:625] (0/8) 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] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.691e+02 2.023e+02 2.384e+02 2.929e+02 5.067e+02, threshold=4.769e+02, percent-clipped=1.0 2023-05-02 20:57:26,455 INFO [train.py:904] (0/8) Epoch 30, batch 1050, loss[loss=0.146, simple_loss=0.2274, pruned_loss=0.03229, over 16805.00 frames. ], tot_loss[loss=0.1587, simple_loss=0.2445, pruned_loss=0.03642, over 3306065.80 frames. ], batch size: 102, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:57:38,644 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5879, 4.5480, 4.5201, 3.9757, 4.5337, 1.8967, 4.2486, 4.0526], device='cuda:0'), covar=tensor([0.0158, 0.0132, 0.0183, 0.0346, 0.0125, 0.2836, 0.0180, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0184, 0.0177, 0.0214, 0.0185, 0.0192, 0.0220, 0.0203, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 20:57:47,750 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295420.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:57:53,571 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295424.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 20:58:36,221 INFO [train.py:904] (0/8) Epoch 30, batch 1100, loss[loss=0.1622, simple_loss=0.2567, pruned_loss=0.03379, over 16649.00 frames. ], tot_loss[loss=0.1581, simple_loss=0.2439, pruned_loss=0.03618, over 3314908.19 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 20:59:38,545 INFO [optim.py:368] (0/8) 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 20:59:43,316 INFO [train.py:904] (0/8) Epoch 30, batch 1150, loss[loss=0.1673, simple_loss=0.2647, pruned_loss=0.03492, over 17128.00 frames. ], tot_loss[loss=0.1573, simple_loss=0.2434, pruned_loss=0.03555, over 3314061.46 frames. ], batch size: 48, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:00:45,315 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6101, 2.6768, 2.3407, 2.4443, 2.9473, 2.6617, 3.1557, 3.1377], device='cuda:0'), covar=tensor([0.0223, 0.0530, 0.0662, 0.0593, 0.0382, 0.0497, 0.0338, 0.0373], device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0252, 0.0239, 0.0241, 0.0250, 0.0249, 0.0247, 0.0250], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:00:51,755 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6820, 3.6987, 2.9176, 2.2601, 2.3766, 2.4082, 3.8450, 3.2030], device='cuda:0'), covar=tensor([0.2788, 0.0617, 0.1744, 0.3375, 0.2899, 0.2342, 0.0573, 0.1731], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0278, 0.0316, 0.0331, 0.0308, 0.0282, 0.0307, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 21:00:52,296 INFO [train.py:904] (0/8) Epoch 30, batch 1200, loss[loss=0.1701, simple_loss=0.2441, pruned_loss=0.04806, over 16775.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2424, pruned_loss=0.03515, over 3315524.38 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:00:54,981 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:00:57,992 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295558.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:01:02,462 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.86 vs. limit=2.0 2023-05-02 21:01:29,927 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295581.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:01:55,322 INFO [optim.py:368] (0/8) 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 21:02:00,812 INFO [train.py:904] (0/8) Epoch 30, batch 1250, loss[loss=0.133, simple_loss=0.2232, pruned_loss=0.02139, over 15905.00 frames. ], tot_loss[loss=0.1565, simple_loss=0.2424, pruned_loss=0.03528, over 3317496.62 frames. ], batch size: 35, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:02:08,694 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4867, 3.5409, 4.0946, 2.1880, 3.2734, 2.5218, 3.8812, 3.7682], device='cuda:0'), covar=tensor([0.0276, 0.1037, 0.0475, 0.2182, 0.0865, 0.1045, 0.0649, 0.1119], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0158, 0.0149, 0.0133, 0.0146, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 21:02:18,629 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295617.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:02:21,363 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295619.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:02:34,273 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295629.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:02:42,477 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295635.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:02:58,039 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 30, batch 1300, loss[loss=0.1354, simple_loss=0.216, pruned_loss=0.02744, over 16786.00 frames. ], tot_loss[loss=0.156, simple_loss=0.2421, pruned_loss=0.03497, over 3311756.73 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:03:22,767 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2958, 4.1109, 4.3538, 4.4597, 4.5565, 4.1099, 4.3645, 4.5556], device='cuda:0'), covar=tensor([0.1584, 0.1198, 0.1322, 0.0700, 0.0599, 0.1365, 0.2625, 0.0807], device='cuda:0'), in_proj_covar=tensor([0.0703, 0.0855, 0.0988, 0.0873, 0.0663, 0.0685, 0.0724, 0.0845], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:03:52,617 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295685.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:57,714 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:03:59,030 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295690.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:04:02,284 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295693.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:11,930 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.484e+02 2.150e+02 2.443e+02 2.917e+02 6.484e+02, threshold=4.887e+02, percent-clipped=2.0 2023-05-02 21:04:18,063 INFO [train.py:904] (0/8) Epoch 30, batch 1350, loss[loss=0.1515, simple_loss=0.2408, pruned_loss=0.03107, over 16632.00 frames. ], tot_loss[loss=0.1566, simple_loss=0.243, pruned_loss=0.03505, over 3321146.78 frames. ], batch size: 62, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:04:33,104 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295715.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:46,349 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295724.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:04:57,202 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.82 vs. limit=2.0 2023-05-02 21:05:17,207 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295746.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:05:18,391 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1916, 5.1222, 4.9145, 4.3372, 5.0057, 1.8695, 4.7354, 4.5703], device='cuda:0'), covar=tensor([0.0120, 0.0118, 0.0250, 0.0487, 0.0140, 0.3189, 0.0177, 0.0349], device='cuda:0'), in_proj_covar=tensor([0.0185, 0.0177, 0.0215, 0.0186, 0.0193, 0.0221, 0.0205, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:05:26,449 INFO [train.py:904] (0/8) Epoch 30, batch 1400, loss[loss=0.1562, simple_loss=0.2332, pruned_loss=0.03962, over 16832.00 frames. ], tot_loss[loss=0.157, simple_loss=0.2433, pruned_loss=0.03533, over 3316268.00 frames. ], batch size: 124, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:05:52,092 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295772.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:06:30,029 INFO [optim.py:368] (0/8) 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 21:06:35,134 INFO [train.py:904] (0/8) Epoch 30, batch 1450, loss[loss=0.1582, simple_loss=0.2389, pruned_loss=0.03876, over 16859.00 frames. ], tot_loss[loss=0.1564, simple_loss=0.2425, pruned_loss=0.03518, over 3311796.52 frames. ], batch size: 96, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:06:56,753 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.6448, 6.0405, 5.7949, 5.9177, 5.4231, 5.5418, 5.4201, 6.1892], device='cuda:0'), covar=tensor([0.1575, 0.0956, 0.1141, 0.0948, 0.0989, 0.0649, 0.1369, 0.1044], device='cuda:0'), in_proj_covar=tensor([0.0742, 0.0892, 0.0735, 0.0697, 0.0571, 0.0568, 0.0751, 0.0703], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:06:56,819 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295820.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:07:44,322 INFO [train.py:904] (0/8) Epoch 30, batch 1500, loss[loss=0.1478, simple_loss=0.2284, pruned_loss=0.03365, over 16859.00 frames. ], tot_loss[loss=0.1561, simple_loss=0.2422, pruned_loss=0.03496, over 3319712.30 frames. ], batch size: 90, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:08:11,361 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=295873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:19,037 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4118, 2.3259, 2.3387, 4.1801, 2.3890, 2.7472, 2.4228, 2.5161], device='cuda:0'), covar=tensor([0.1476, 0.4134, 0.3562, 0.0617, 0.4413, 0.2936, 0.4032, 0.3885], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0483, 0.0394, 0.0344, 0.0451, 0.0553, 0.0456, 0.0567], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:08:22,038 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:22,081 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295881.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:08:47,916 INFO [optim.py:368] (0/8) 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 21:08:54,114 INFO [train.py:904] (0/8) Epoch 30, batch 1550, loss[loss=0.1849, simple_loss=0.2743, pruned_loss=0.04769, over 15368.00 frames. ], tot_loss[loss=0.1585, simple_loss=0.2441, pruned_loss=0.03647, over 3316430.23 frames. ], batch size: 190, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:09:05,642 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295912.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:09:08,349 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295914.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:09:29,569 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295929.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:09:36,333 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=295934.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:09:37,390 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295935.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:10:02,950 INFO [train.py:904] (0/8) Epoch 30, batch 1600, loss[loss=0.1604, simple_loss=0.2607, pruned_loss=0.03009, over 17129.00 frames. ], tot_loss[loss=0.159, simple_loss=0.2452, pruned_loss=0.03643, over 3310284.07 frames. ], batch size: 49, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:10:36,038 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9759, 2.8598, 2.5685, 4.8056, 3.8235, 4.2227, 1.7319, 3.2797], device='cuda:0'), covar=tensor([0.1366, 0.0824, 0.1362, 0.0207, 0.0225, 0.0405, 0.1683, 0.0784], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0184, 0.0203, 0.0209, 0.0208, 0.0221, 0.0214, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:10:42,974 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=295983.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:10:45,377 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=295985.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:10:50,592 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=295989.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:11:05,444 INFO [optim.py:368] (0/8) 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 21:11:05,820 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-296000.pt 2023-05-02 21:11:14,065 INFO [train.py:904] (0/8) Epoch 30, batch 1650, loss[loss=0.1615, simple_loss=0.2471, pruned_loss=0.03795, over 15915.00 frames. ], tot_loss[loss=0.16, simple_loss=0.2462, pruned_loss=0.03696, over 3307617.61 frames. ], batch size: 35, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:11:30,086 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296015.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:00,958 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296037.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:05,652 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296041.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:12:16,922 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 30, batch 1700, loss[loss=0.1411, simple_loss=0.227, pruned_loss=0.02762, over 16728.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2477, pruned_loss=0.03728, over 3309208.49 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:12:36,068 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296063.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:13:26,988 INFO [optim.py:368] (0/8) 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 21:13:32,346 INFO [train.py:904] (0/8) Epoch 30, batch 1750, loss[loss=0.1544, simple_loss=0.2465, pruned_loss=0.03115, over 17241.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.249, pruned_loss=0.03718, over 3313021.76 frames. ], batch size: 45, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:13:40,794 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296110.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:13:42,301 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.45 vs. limit=5.0 2023-05-02 21:14:24,215 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7968, 2.6576, 2.5571, 4.1654, 3.3128, 4.0495, 1.6892, 2.9377], device='cuda:0'), covar=tensor([0.1400, 0.0742, 0.1188, 0.0176, 0.0152, 0.0391, 0.1571, 0.0844], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0184, 0.0203, 0.0209, 0.0208, 0.0221, 0.0214, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:14:31,709 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3090, 3.5659, 3.5246, 2.0603, 2.9943, 2.2246, 3.7337, 3.8609], device='cuda:0'), covar=tensor([0.0284, 0.0887, 0.0742, 0.2496, 0.1057, 0.1282, 0.0596, 0.0960], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0171, 0.0171, 0.0158, 0.0148, 0.0133, 0.0146, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 21:14:41,104 INFO [train.py:904] (0/8) Epoch 30, batch 1800, loss[loss=0.176, simple_loss=0.2626, pruned_loss=0.04473, over 15571.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2502, pruned_loss=0.03737, over 3310607.84 frames. ], batch size: 190, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:15:12,953 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296176.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:15:18,294 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.24 vs. limit=2.0 2023-05-02 21:15:47,287 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.503e+02 2.165e+02 2.480e+02 2.953e+02 5.408e+02, threshold=4.959e+02, percent-clipped=1.0 2023-05-02 21:15:51,606 INFO [train.py:904] (0/8) Epoch 30, batch 1850, loss[loss=0.1467, simple_loss=0.2456, pruned_loss=0.02388, over 17034.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2517, pruned_loss=0.03776, over 3312418.20 frames. ], batch size: 50, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:16:03,791 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296212.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:16:06,026 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296214.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:16:27,213 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296229.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:16:56,267 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2844, 5.5964, 5.3702, 5.4122, 5.1163, 5.0886, 4.9867, 5.7333], device='cuda:0'), covar=tensor([0.1448, 0.0959, 0.1159, 0.0935, 0.0882, 0.0844, 0.1323, 0.0886], device='cuda:0'), in_proj_covar=tensor([0.0740, 0.0891, 0.0734, 0.0695, 0.0570, 0.0564, 0.0750, 0.0702], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:17:02,020 INFO [train.py:904] (0/8) Epoch 30, batch 1900, loss[loss=0.1612, simple_loss=0.2446, pruned_loss=0.03894, over 16690.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2514, pruned_loss=0.03726, over 3309492.05 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:17:05,235 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.94 vs. limit=5.0 2023-05-02 21:17:09,981 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296260.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:17:12,894 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296262.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:17:42,690 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.34 vs. limit=2.0 2023-05-02 21:17:45,574 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296285.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:17:55,609 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.25 vs. limit=2.0 2023-05-02 21:18:06,424 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.514e+02 2.062e+02 2.316e+02 2.814e+02 5.874e+02, threshold=4.633e+02, percent-clipped=2.0 2023-05-02 21:18:10,610 INFO [train.py:904] (0/8) Epoch 30, batch 1950, loss[loss=0.1608, simple_loss=0.2436, pruned_loss=0.03897, over 16821.00 frames. ], tot_loss[loss=0.1621, simple_loss=0.2507, pruned_loss=0.03677, over 3309810.59 frames. ], batch size: 96, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:18:27,001 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296315.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:18:51,919 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296333.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:19:02,334 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296341.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:19:10,870 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0347, 1.9864, 2.6048, 2.9083, 2.8663, 3.1723, 2.0167, 3.2971], device='cuda:0'), covar=tensor([0.0239, 0.0669, 0.0383, 0.0380, 0.0380, 0.0290, 0.0769, 0.0228], device='cuda:0'), in_proj_covar=tensor([0.0203, 0.0202, 0.0191, 0.0196, 0.0214, 0.0171, 0.0207, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 21:19:21,054 INFO [train.py:904] (0/8) Epoch 30, batch 2000, loss[loss=0.1798, simple_loss=0.2699, pruned_loss=0.04486, over 11951.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2518, pruned_loss=0.03685, over 3287274.96 frames. ], batch size: 246, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:19:37,456 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.44 vs. limit=2.0 2023-05-02 21:19:51,098 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296376.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:00,815 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6259, 3.7707, 2.6060, 4.4651, 3.1026, 4.3728, 2.4988, 3.2003], device='cuda:0'), covar=tensor([0.0379, 0.0519, 0.1472, 0.0319, 0.0851, 0.0600, 0.1631, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0180, 0.0186, 0.0198, 0.0179, 0.0183, 0.0226, 0.0209, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:20:09,536 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296389.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:20:23,011 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4052, 5.3553, 5.1135, 4.5230, 5.1921, 2.0579, 4.9065, 4.8754], device='cuda:0'), covar=tensor([0.0097, 0.0096, 0.0227, 0.0445, 0.0115, 0.2910, 0.0156, 0.0279], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0179, 0.0217, 0.0188, 0.0196, 0.0223, 0.0207, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:20:25,432 INFO [optim.py:368] (0/8) 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 21:20:29,066 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6993, 2.7135, 2.1578, 2.2863, 2.9079, 2.6314, 3.3255, 3.2253], device='cuda:0'), covar=tensor([0.0228, 0.0575, 0.0784, 0.0722, 0.0459, 0.0560, 0.0329, 0.0371], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0252, 0.0240, 0.0242, 0.0252, 0.0250, 0.0249, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:20:30,271 INFO [train.py:904] (0/8) Epoch 30, batch 2050, loss[loss=0.1551, simple_loss=0.2562, pruned_loss=0.02696, over 17141.00 frames. ], tot_loss[loss=0.1635, simple_loss=0.2526, pruned_loss=0.03716, over 3283823.57 frames. ], batch size: 49, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:20:32,298 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296405.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:11,780 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296434.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:21:39,728 INFO [train.py:904] (0/8) Epoch 30, batch 2100, loss[loss=0.183, simple_loss=0.2689, pruned_loss=0.04858, over 16300.00 frames. ], tot_loss[loss=0.1639, simple_loss=0.2527, pruned_loss=0.03755, over 3297346.95 frames. ], batch size: 165, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:21:43,053 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6884, 4.6090, 4.5346, 3.9673, 4.5984, 1.8682, 4.3867, 4.1962], device='cuda:0'), covar=tensor([0.0157, 0.0147, 0.0208, 0.0362, 0.0121, 0.2996, 0.0163, 0.0251], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0179, 0.0217, 0.0187, 0.0195, 0.0222, 0.0207, 0.0184], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:22:09,575 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296476.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:33,186 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296492.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:36,824 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296495.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:22:39,046 INFO [zipformer.py:625] (0/8) 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] (0/8) 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 21:22:48,899 INFO [train.py:904] (0/8) Epoch 30, batch 2150, loss[loss=0.1739, simple_loss=0.2758, pruned_loss=0.03597, over 17133.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2533, pruned_loss=0.03743, over 3303812.88 frames. ], batch size: 47, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:23:03,734 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296514.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:16,235 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296524.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:23:23,441 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296529.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:23:35,309 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4008, 4.4695, 4.7363, 4.7257, 4.7887, 4.4997, 4.4981, 4.3525], device='cuda:0'), covar=tensor([0.0375, 0.0664, 0.0432, 0.0430, 0.0568, 0.0437, 0.0895, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0448, 0.0510, 0.0487, 0.0451, 0.0534, 0.0517, 0.0593, 0.0415], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 21:23:43,120 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1470, 5.6065, 5.7433, 5.3942, 5.4873, 6.1298, 5.5799, 5.2726], device='cuda:0'), covar=tensor([0.0980, 0.2033, 0.2579, 0.2104, 0.2588, 0.0939, 0.1611, 0.2422], device='cuda:0'), in_proj_covar=tensor([0.0438, 0.0654, 0.0731, 0.0532, 0.0710, 0.0749, 0.0562, 0.0706], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 21:23:56,829 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296553.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 21:23:57,548 INFO [train.py:904] (0/8) Epoch 30, batch 2200, loss[loss=0.1849, simple_loss=0.2654, pruned_loss=0.05226, over 16871.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.253, pruned_loss=0.03766, over 3310712.79 frames. ], batch size: 96, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:24:04,167 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296558.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:13,100 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.1430, 2.3171, 2.4447, 3.9140, 2.2915, 2.6220, 2.3760, 2.4943], device='cuda:0'), covar=tensor([0.1698, 0.3811, 0.3136, 0.0736, 0.4179, 0.2707, 0.3927, 0.3253], device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0484, 0.0395, 0.0346, 0.0451, 0.0555, 0.0457, 0.0568], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:24:27,013 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296575.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:24:30,756 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:25:04,058 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.512e+02 2.151e+02 2.512e+02 3.134e+02 5.584e+02, threshold=5.024e+02, percent-clipped=2.0 2023-05-02 21:25:06,332 INFO [train.py:904] (0/8) Epoch 30, batch 2250, loss[loss=0.1685, simple_loss=0.2687, pruned_loss=0.03415, over 17043.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2541, pruned_loss=0.03855, over 3307189.57 frames. ], batch size: 53, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:25:20,072 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:25:31,110 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.75 vs. limit=5.0 2023-05-02 21:26:17,744 INFO [train.py:904] (0/8) Epoch 30, batch 2300, loss[loss=0.1369, simple_loss=0.2286, pruned_loss=0.02263, over 16845.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2536, pruned_loss=0.03869, over 3307222.39 frames. ], batch size: 39, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:26:41,399 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296671.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:26:45,154 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296674.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:27:05,041 INFO [zipformer.py:625] (0/8) 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] (0/8) 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] (0/8) Epoch 30, batch 2350, loss[loss=0.17, simple_loss=0.2626, pruned_loss=0.03867, over 16444.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2538, pruned_loss=0.0387, over 3315733.75 frames. ], batch size: 68, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:27:28,755 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296705.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:30,532 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9099, 4.4650, 4.4442, 3.0959, 3.6158, 4.4255, 3.8658, 2.6355], device='cuda:0'), covar=tensor([0.0496, 0.0079, 0.0050, 0.0387, 0.0179, 0.0105, 0.0120, 0.0494], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0136, 0.0105, 0.0117, 0.0100, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 21:28:30,571 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296749.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:36,424 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=296753.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:28:37,289 INFO [train.py:904] (0/8) Epoch 30, batch 2400, loss[loss=0.1663, simple_loss=0.2439, pruned_loss=0.04431, over 16828.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2542, pruned_loss=0.03864, over 3311911.91 frames. ], batch size: 116, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:29:27,118 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296790.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:29:43,968 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 30, batch 2450, loss[loss=0.1606, simple_loss=0.2426, pruned_loss=0.03929, over 16889.00 frames. ], tot_loss[loss=0.1656, simple_loss=0.2544, pruned_loss=0.03844, over 3317818.09 frames. ], batch size: 96, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:30:16,099 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8085, 2.7274, 2.6073, 4.9108, 3.5244, 4.1761, 1.7676, 3.0068], device='cuda:0'), covar=tensor([0.1508, 0.0967, 0.1460, 0.0250, 0.0246, 0.0477, 0.1804, 0.0950], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0185, 0.0204, 0.0210, 0.0209, 0.0222, 0.0214, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:30:48,235 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296848.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 21:30:55,422 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296853.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:30:56,389 INFO [train.py:904] (0/8) Epoch 30, batch 2500, loss[loss=0.1554, simple_loss=0.2386, pruned_loss=0.0361, over 16823.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2543, pruned_loss=0.03839, over 3324641.12 frames. ], batch size: 96, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:31:19,658 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296870.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:31:23,963 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296873.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:31:34,212 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0474, 5.0990, 5.4953, 5.4770, 5.5016, 5.1676, 5.1077, 4.9930], device='cuda:0'), covar=tensor([0.0388, 0.0595, 0.0416, 0.0416, 0.0508, 0.0442, 0.1056, 0.0445], device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0512, 0.0489, 0.0452, 0.0534, 0.0517, 0.0594, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 21:31:38,902 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6300, 3.7066, 2.8272, 2.2705, 2.3909, 2.3890, 3.7899, 3.2034], device='cuda:0'), covar=tensor([0.2846, 0.0557, 0.1831, 0.3078, 0.2862, 0.2264, 0.0513, 0.1592], device='cuda:0'), in_proj_covar=tensor([0.0339, 0.0278, 0.0316, 0.0330, 0.0309, 0.0282, 0.0307, 0.0357], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 21:31:44,986 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.4865, 5.8618, 5.4167, 5.7808, 5.3367, 5.1294, 5.4373, 5.9223], device='cuda:0'), covar=tensor([0.2409, 0.1560, 0.2679, 0.1486, 0.1757, 0.1400, 0.2281, 0.2181], device='cuda:0'), in_proj_covar=tensor([0.0739, 0.0890, 0.0733, 0.0693, 0.0569, 0.0563, 0.0748, 0.0701], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:32:04,584 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.84 vs. limit=2.0 2023-05-02 21:32:04,916 INFO [optim.py:368] (0/8) 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 21:32:07,805 INFO [train.py:904] (0/8) Epoch 30, batch 2550, loss[loss=0.1647, simple_loss=0.2458, pruned_loss=0.04184, over 16753.00 frames. ], tot_loss[loss=0.1653, simple_loss=0.2544, pruned_loss=0.0381, over 3335627.78 frames. ], batch size: 134, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:32:22,448 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 21:32:31,250 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 21:32:49,101 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296934.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:32:54,135 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=296938.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:15,498 INFO [train.py:904] (0/8) Epoch 30, batch 2600, loss[loss=0.1865, simple_loss=0.281, pruned_loss=0.04593, over 16660.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.254, pruned_loss=0.03816, over 3323324.26 frames. ], batch size: 57, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:33:36,669 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=296969.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:33:39,717 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=296971.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:17,834 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=296999.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:22,268 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.341e+02 2.007e+02 2.417e+02 2.925e+02 5.072e+02, threshold=4.833e+02, percent-clipped=0.0 2023-05-02 21:34:24,417 INFO [train.py:904] (0/8) Epoch 30, batch 2650, loss[loss=0.153, simple_loss=0.2609, pruned_loss=0.02253, over 17093.00 frames. ], tot_loss[loss=0.1661, simple_loss=0.2551, pruned_loss=0.03848, over 3322795.53 frames. ], batch size: 49, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:34:32,605 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.11 vs. limit=5.0 2023-05-02 21:34:46,006 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297019.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:34:58,076 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.91 vs. limit=2.0 2023-05-02 21:35:20,622 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297044.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:35:34,424 INFO [train.py:904] (0/8) Epoch 30, batch 2700, loss[loss=0.1829, simple_loss=0.2679, pruned_loss=0.04892, over 16656.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.255, pruned_loss=0.03806, over 3326440.58 frames. ], batch size: 89, lr: 2.25e-03, grad_scale: 8.0 2023-05-02 21:35:53,366 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7105, 4.0311, 4.1895, 4.1469, 4.2085, 3.9606, 3.7272, 3.9541], device='cuda:0'), covar=tensor([0.0735, 0.0778, 0.0680, 0.0718, 0.0756, 0.0747, 0.1380, 0.0774], device='cuda:0'), in_proj_covar=tensor([0.0450, 0.0513, 0.0490, 0.0453, 0.0534, 0.0517, 0.0595, 0.0416], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 21:36:23,935 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297090.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:36:42,013 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.416e+02 2.120e+02 2.454e+02 2.893e+02 5.534e+02, threshold=4.909e+02, percent-clipped=1.0 2023-05-02 21:36:43,154 INFO [train.py:904] (0/8) Epoch 30, batch 2750, loss[loss=0.1709, simple_loss=0.268, pruned_loss=0.03695, over 16558.00 frames. ], tot_loss[loss=0.1654, simple_loss=0.2551, pruned_loss=0.03782, over 3329054.45 frames. ], batch size: 68, lr: 2.25e-03, grad_scale: 4.0 2023-05-02 21:37:30,065 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:44,464 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297148.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:50,770 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297153.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:37:51,588 INFO [train.py:904] (0/8) Epoch 30, batch 2800, loss[loss=0.1632, simple_loss=0.2566, pruned_loss=0.03489, over 17247.00 frames. ], tot_loss[loss=0.1652, simple_loss=0.255, pruned_loss=0.03772, over 3330141.90 frames. ], batch size: 52, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:38:14,274 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297170.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:49,785 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297196.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:38:56,656 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:00,136 INFO [optim.py:368] (0/8) 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 21:39:01,281 INFO [train.py:904] (0/8) Epoch 30, batch 2850, loss[loss=0.1865, simple_loss=0.276, pruned_loss=0.04845, over 17104.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.254, pruned_loss=0.03753, over 3332491.74 frames. ], batch size: 53, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:39:21,042 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297218.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:26,982 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297222.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:39:34,094 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8713, 4.3060, 4.2958, 3.0850, 3.5801, 4.3269, 3.8276, 2.6402], device='cuda:0'), covar=tensor([0.0504, 0.0099, 0.0068, 0.0393, 0.0180, 0.0116, 0.0115, 0.0483], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0091, 0.0092, 0.0135, 0.0104, 0.0116, 0.0099, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 21:39:37,170 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297229.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 21:40:03,505 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8796, 5.1172, 5.2766, 5.0022, 5.0660, 5.7124, 5.1581, 4.8465], device='cuda:0'), covar=tensor([0.1520, 0.2433, 0.3003, 0.2378, 0.2827, 0.1206, 0.2071, 0.3009], device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0654, 0.0730, 0.0532, 0.0710, 0.0748, 0.0562, 0.0708], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 21:40:10,547 INFO [train.py:904] (0/8) Epoch 30, batch 2900, loss[loss=0.1829, simple_loss=0.257, pruned_loss=0.05443, over 15741.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2517, pruned_loss=0.03724, over 3335953.99 frames. ], batch size: 191, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:40:33,093 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297269.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:40:35,882 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-02 21:40:45,634 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.77 vs. limit=2.0 2023-05-02 21:40:50,565 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5535, 2.6482, 2.1004, 2.2878, 2.8958, 2.5931, 3.1564, 3.1402], device='cuda:0'), covar=tensor([0.0220, 0.0623, 0.0778, 0.0648, 0.0388, 0.0489, 0.0378, 0.0352], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0251, 0.0239, 0.0240, 0.0251, 0.0249, 0.0249, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:40:51,643 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=297283.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:08,334 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297294.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:41:22,344 INFO [optim.py:368] (0/8) 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] (0/8) Epoch 30, batch 2950, loss[loss=0.1492, simple_loss=0.2482, pruned_loss=0.0251, over 17128.00 frames. ], tot_loss[loss=0.163, simple_loss=0.251, pruned_loss=0.03746, over 3336525.36 frames. ], batch size: 49, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:41:41,496 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297317.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:18,969 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297344.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:42:33,298 INFO [train.py:904] (0/8) Epoch 30, batch 3000, loss[loss=0.1616, simple_loss=0.2536, pruned_loss=0.03482, over 16535.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2515, pruned_loss=0.038, over 3334087.28 frames. ], batch size: 68, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:42:33,299 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 21:42:42,093 INFO [train.py:938] (0/8) Epoch 30, validation: loss=0.1331, simple_loss=0.238, pruned_loss=0.01415, over 944034.00 frames. 2023-05-02 21:42:42,094 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 21:43:35,187 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297392.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:43:51,745 INFO [optim.py:368] (0/8) 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 21:43:51,761 INFO [train.py:904] (0/8) Epoch 30, batch 3050, loss[loss=0.1508, simple_loss=0.2376, pruned_loss=0.03201, over 16972.00 frames. ], tot_loss[loss=0.1634, simple_loss=0.2509, pruned_loss=0.03797, over 3332835.77 frames. ], batch size: 41, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:01,435 INFO [train.py:904] (0/8) Epoch 30, batch 3100, loss[loss=0.1525, simple_loss=0.2311, pruned_loss=0.03701, over 16313.00 frames. ], tot_loss[loss=0.164, simple_loss=0.2512, pruned_loss=0.03842, over 3330595.05 frames. ], batch size: 36, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:45:42,858 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6121, 3.6747, 2.4270, 4.0104, 2.9623, 3.9553, 2.5473, 3.0919], device='cuda:0'), covar=tensor([0.0358, 0.0508, 0.1594, 0.0448, 0.0811, 0.0869, 0.1389, 0.0729], device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0186, 0.0199, 0.0180, 0.0184, 0.0227, 0.0209, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:46:07,378 INFO [optim.py:368] (0/8) 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 21:46:07,394 INFO [train.py:904] (0/8) Epoch 30, batch 3150, loss[loss=0.1693, simple_loss=0.2675, pruned_loss=0.03554, over 17075.00 frames. ], tot_loss[loss=0.1638, simple_loss=0.2504, pruned_loss=0.03858, over 3331929.57 frames. ], batch size: 55, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:46:41,938 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297529.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 21:46:57,404 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8460, 2.9910, 2.9148, 5.1392, 4.1329, 4.5378, 1.8339, 3.2550], device='cuda:0'), covar=tensor([0.1378, 0.0817, 0.1136, 0.0203, 0.0209, 0.0360, 0.1632, 0.0766], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0185, 0.0204, 0.0211, 0.0210, 0.0222, 0.0215, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:47:17,219 INFO [train.py:904] (0/8) Epoch 30, batch 3200, loss[loss=0.1701, simple_loss=0.2619, pruned_loss=0.03913, over 16733.00 frames. ], tot_loss[loss=0.1637, simple_loss=0.2505, pruned_loss=0.03847, over 3327253.24 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:47:38,841 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9073, 3.1658, 2.8251, 5.1868, 4.2402, 4.5158, 1.8313, 3.3388], device='cuda:0'), covar=tensor([0.1391, 0.0759, 0.1220, 0.0207, 0.0254, 0.0363, 0.1663, 0.0779], device='cuda:0'), in_proj_covar=tensor([0.0176, 0.0185, 0.0204, 0.0211, 0.0210, 0.0222, 0.0215, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:47:49,841 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297577.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:47:51,101 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=297578.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:48:13,107 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297594.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:48:16,268 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.37 vs. limit=2.0 2023-05-02 21:48:27,242 INFO [optim.py:368] (0/8) 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 21:48:27,263 INFO [train.py:904] (0/8) Epoch 30, batch 3250, loss[loss=0.1632, simple_loss=0.2558, pruned_loss=0.03533, over 17082.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2513, pruned_loss=0.03855, over 3325855.31 frames. ], batch size: 53, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:49:20,245 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297642.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:49:36,049 INFO [train.py:904] (0/8) Epoch 30, batch 3300, loss[loss=0.1733, simple_loss=0.2576, pruned_loss=0.04453, over 16475.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2512, pruned_loss=0.03858, over 3333852.85 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:49:43,904 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1087, 5.5607, 5.6716, 5.3483, 5.4629, 6.0504, 5.4649, 5.1771], device='cuda:0'), covar=tensor([0.1015, 0.1910, 0.2386, 0.2272, 0.2586, 0.0976, 0.1580, 0.2626], device='cuda:0'), in_proj_covar=tensor([0.0440, 0.0657, 0.0731, 0.0531, 0.0711, 0.0750, 0.0560, 0.0710], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 21:50:25,809 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.22 vs. limit=2.0 2023-05-02 21:50:46,303 INFO [optim.py:368] (0/8) 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 21:50:46,319 INFO [train.py:904] (0/8) Epoch 30, batch 3350, loss[loss=0.1831, simple_loss=0.2814, pruned_loss=0.0424, over 17052.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2517, pruned_loss=0.03834, over 3332438.60 frames. ], batch size: 53, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:51:32,029 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.88 vs. limit=5.0 2023-05-02 21:51:56,074 INFO [train.py:904] (0/8) Epoch 30, batch 3400, loss[loss=0.1648, simple_loss=0.2543, pruned_loss=0.03765, over 16839.00 frames. ], tot_loss[loss=0.163, simple_loss=0.2507, pruned_loss=0.03768, over 3339361.84 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:53:07,136 INFO [optim.py:368] (0/8) 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 21:53:07,157 INFO [train.py:904] (0/8) Epoch 30, batch 3450, loss[loss=0.1693, simple_loss=0.2505, pruned_loss=0.04403, over 16508.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2499, pruned_loss=0.0378, over 3335441.65 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:53:48,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0664, 5.0968, 5.4887, 5.4921, 5.4930, 5.1476, 5.1037, 4.9239], device='cuda:0'), covar=tensor([0.0358, 0.0626, 0.0404, 0.0373, 0.0437, 0.0416, 0.0972, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0456, 0.0520, 0.0494, 0.0458, 0.0542, 0.0523, 0.0605, 0.0422], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 21:53:50,245 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8866, 2.4928, 2.0922, 2.2947, 2.8267, 2.6165, 2.8581, 2.9299], device='cuda:0'), covar=tensor([0.0254, 0.0468, 0.0575, 0.0494, 0.0265, 0.0399, 0.0217, 0.0331], device='cuda:0'), in_proj_covar=tensor([0.0243, 0.0252, 0.0240, 0.0242, 0.0253, 0.0250, 0.0251, 0.0253], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:54:17,197 INFO [train.py:904] (0/8) Epoch 30, batch 3500, loss[loss=0.1471, simple_loss=0.2424, pruned_loss=0.02589, over 17120.00 frames. ], tot_loss[loss=0.1615, simple_loss=0.2491, pruned_loss=0.03693, over 3341345.88 frames. ], batch size: 47, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:54:36,691 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3591, 4.7691, 4.7141, 3.6006, 3.9588, 4.7027, 4.1104, 3.0485], device='cuda:0'), covar=tensor([0.0443, 0.0070, 0.0047, 0.0348, 0.0159, 0.0093, 0.0100, 0.0467], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0093, 0.0095, 0.0139, 0.0106, 0.0119, 0.0102, 0.0134], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 21:54:51,989 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=297878.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:55:28,094 INFO [train.py:904] (0/8) Epoch 30, batch 3550, loss[loss=0.1707, simple_loss=0.2487, pruned_loss=0.04634, over 11449.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2484, pruned_loss=0.03664, over 3322199.04 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 4.0 2023-05-02 21:55:29,308 INFO [optim.py:368] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=297926.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:56:15,091 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2367, 5.6579, 5.2277, 5.5769, 5.1774, 5.0474, 5.1876, 5.7769], device='cuda:0'), covar=tensor([0.2767, 0.1752, 0.2606, 0.1569, 0.1784, 0.1362, 0.2595, 0.2137], device='cuda:0'), in_proj_covar=tensor([0.0749, 0.0903, 0.0743, 0.0703, 0.0580, 0.0569, 0.0759, 0.0710], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:56:16,963 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8355, 4.6094, 4.8718, 5.0484, 5.2317, 4.6277, 5.2266, 5.2264], device='cuda:0'), covar=tensor([0.1976, 0.1450, 0.1862, 0.0853, 0.0635, 0.1277, 0.0661, 0.0671], device='cuda:0'), in_proj_covar=tensor([0.0729, 0.0885, 0.1026, 0.0906, 0.0685, 0.0716, 0.0754, 0.0877], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 21:56:27,767 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.23 vs. limit=2.0 2023-05-02 21:56:38,339 INFO [train.py:904] (0/8) Epoch 30, batch 3600, loss[loss=0.1584, simple_loss=0.2498, pruned_loss=0.03349, over 16722.00 frames. ], tot_loss[loss=0.1603, simple_loss=0.2478, pruned_loss=0.03637, over 3325891.74 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:56:42,436 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=297957.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:57:05,005 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5981, 3.6030, 2.8202, 2.2620, 2.3535, 2.3449, 3.7463, 3.1501], device='cuda:0'), covar=tensor([0.2982, 0.0625, 0.1785, 0.3075, 0.2844, 0.2397, 0.0545, 0.1757], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0279, 0.0317, 0.0332, 0.0312, 0.0284, 0.0309, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 21:57:07,418 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6612, 3.8033, 2.8830, 2.3028, 2.4314, 2.4105, 3.9430, 3.3119], device='cuda:0'), covar=tensor([0.2918, 0.0573, 0.1833, 0.3233, 0.2998, 0.2302, 0.0486, 0.1564], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0279, 0.0317, 0.0332, 0.0312, 0.0284, 0.0309, 0.0359], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 21:57:26,453 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6276, 3.1785, 3.5715, 2.0515, 3.6661, 3.6500, 3.1708, 2.8594], device='cuda:0'), covar=tensor([0.0752, 0.0288, 0.0210, 0.1102, 0.0124, 0.0211, 0.0394, 0.0448], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0112, 0.0105, 0.0140, 0.0089, 0.0136, 0.0133, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 21:57:45,312 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-298000.pt 2023-05-02 21:57:53,333 INFO [train.py:904] (0/8) Epoch 30, batch 3650, loss[loss=0.1914, simple_loss=0.2669, pruned_loss=0.05798, over 16719.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2472, pruned_loss=0.03705, over 3320718.52 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 21:57:55,116 INFO [optim.py:368] (0/8) 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 21:58:05,665 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6107, 3.6193, 3.8217, 2.7157, 3.5076, 3.9190, 3.6090, 2.1721], device='cuda:0'), covar=tensor([0.0525, 0.0161, 0.0070, 0.0414, 0.0120, 0.0109, 0.0115, 0.0570], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0092, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 21:58:14,285 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298018.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 21:59:06,228 INFO [train.py:904] (0/8) Epoch 30, batch 3700, loss[loss=0.1712, simple_loss=0.2517, pruned_loss=0.04537, over 16416.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2458, pruned_loss=0.03801, over 3294614.60 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:15,401 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8670, 4.1071, 3.0475, 2.4516, 2.7100, 2.6872, 4.3153, 3.6020], device='cuda:0'), covar=tensor([0.2970, 0.0562, 0.2029, 0.3184, 0.2938, 0.2189, 0.0549, 0.1456], device='cuda:0'), in_proj_covar=tensor([0.0339, 0.0278, 0.0316, 0.0331, 0.0311, 0.0283, 0.0308, 0.0358], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 22:00:17,710 INFO [train.py:904] (0/8) Epoch 30, batch 3750, loss[loss=0.1792, simple_loss=0.2708, pruned_loss=0.04377, over 16644.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2464, pruned_loss=0.03928, over 3294171.05 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:00:19,705 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.710e+02 2.048e+02 2.384e+02 2.845e+02 4.773e+02, threshold=4.768e+02, percent-clipped=0.0 2023-05-02 22:00:50,040 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0792, 2.7581, 2.6538, 4.4884, 3.4467, 4.0580, 1.9581, 3.0770], device='cuda:0'), covar=tensor([0.1369, 0.0869, 0.1316, 0.0202, 0.0266, 0.0510, 0.1610, 0.0930], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0184, 0.0203, 0.0210, 0.0209, 0.0221, 0.0213, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:00:52,336 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9461, 2.7260, 2.5944, 1.9421, 2.6143, 2.7614, 2.6048, 1.9058], device='cuda:0'), covar=tensor([0.0484, 0.0127, 0.0109, 0.0416, 0.0155, 0.0168, 0.0150, 0.0456], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0092, 0.0094, 0.0137, 0.0106, 0.0119, 0.0101, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 22:01:30,456 INFO [train.py:904] (0/8) Epoch 30, batch 3800, loss[loss=0.1766, simple_loss=0.2503, pruned_loss=0.0515, over 16846.00 frames. ], tot_loss[loss=0.1645, simple_loss=0.2478, pruned_loss=0.04055, over 3285779.85 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:01:30,888 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2415, 5.3237, 5.6064, 5.5614, 5.6568, 5.2864, 5.1784, 4.9397], device='cuda:0'), covar=tensor([0.0332, 0.0494, 0.0346, 0.0431, 0.0578, 0.0371, 0.1213, 0.0541], device='cuda:0'), in_proj_covar=tensor([0.0455, 0.0521, 0.0496, 0.0457, 0.0542, 0.0523, 0.0605, 0.0422], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 22:01:35,894 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9472, 2.6543, 2.8476, 2.0930, 2.6740, 2.1092, 2.7535, 2.8844], device='cuda:0'), covar=tensor([0.0283, 0.0838, 0.0597, 0.1985, 0.0860, 0.1006, 0.0573, 0.0773], device='cuda:0'), in_proj_covar=tensor([0.0165, 0.0175, 0.0173, 0.0160, 0.0151, 0.0135, 0.0148, 0.0189], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 22:02:43,885 INFO [train.py:904] (0/8) Epoch 30, batch 3850, loss[loss=0.1633, simple_loss=0.237, pruned_loss=0.04474, over 16882.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2487, pruned_loss=0.0414, over 3272544.53 frames. ], batch size: 90, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:02:44,951 INFO [optim.py:368] (0/8) 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 22:03:30,426 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5002, 3.1510, 3.5782, 1.9966, 3.6308, 3.6397, 3.1038, 2.7839], device='cuda:0'), covar=tensor([0.0809, 0.0319, 0.0211, 0.1184, 0.0139, 0.0260, 0.0405, 0.0491], device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0113, 0.0104, 0.0140, 0.0089, 0.0136, 0.0133, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:03:53,267 INFO [train.py:904] (0/8) Epoch 30, batch 3900, loss[loss=0.2017, simple_loss=0.2758, pruned_loss=0.06374, over 12605.00 frames. ], tot_loss[loss=0.1663, simple_loss=0.2485, pruned_loss=0.04205, over 3264805.76 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:04:28,111 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.00 vs. limit=5.0 2023-05-02 22:04:42,295 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8126, 3.8650, 2.4554, 4.1073, 3.0911, 4.0872, 2.5725, 3.2158], device='cuda:0'), covar=tensor([0.0257, 0.0347, 0.1533, 0.0334, 0.0719, 0.0645, 0.1358, 0.0659], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0186, 0.0199, 0.0179, 0.0183, 0.0226, 0.0207, 0.0186], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:05:04,308 INFO [train.py:904] (0/8) Epoch 30, batch 3950, loss[loss=0.1507, simple_loss=0.2244, pruned_loss=0.03847, over 16785.00 frames. ], tot_loss[loss=0.1671, simple_loss=0.2481, pruned_loss=0.04303, over 3264387.18 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:05:05,534 INFO [optim.py:368] (0/8) 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 22:05:16,028 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.89 vs. limit=2.0 2023-05-02 22:05:17,515 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298313.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:06:02,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6512, 4.6963, 4.9632, 4.9333, 4.9920, 4.7074, 4.6927, 4.5339], device='cuda:0'), covar=tensor([0.0405, 0.0690, 0.0473, 0.0458, 0.0537, 0.0436, 0.0820, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0456, 0.0520, 0.0495, 0.0458, 0.0541, 0.0523, 0.0603, 0.0422], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 22:06:15,427 INFO [train.py:904] (0/8) Epoch 30, batch 4000, loss[loss=0.1648, simple_loss=0.251, pruned_loss=0.03926, over 16766.00 frames. ], tot_loss[loss=0.1672, simple_loss=0.2477, pruned_loss=0.04339, over 3275357.20 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:06:28,630 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 22:07:25,654 INFO [train.py:904] (0/8) Epoch 30, batch 4050, loss[loss=0.1531, simple_loss=0.2417, pruned_loss=0.03224, over 17133.00 frames. ], tot_loss[loss=0.1673, simple_loss=0.2489, pruned_loss=0.04287, over 3267847.85 frames. ], batch size: 49, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:07:27,601 INFO [optim.py:368] (0/8) 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 22:07:27,958 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0278, 4.3553, 4.5183, 4.4562, 4.4872, 4.2406, 3.9681, 4.1498], device='cuda:0'), covar=tensor([0.0585, 0.0745, 0.0561, 0.0645, 0.0734, 0.0688, 0.1383, 0.0688], device='cuda:0'), in_proj_covar=tensor([0.0455, 0.0519, 0.0493, 0.0457, 0.0540, 0.0522, 0.0601, 0.0421], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-02 22:07:36,199 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.71 vs. limit=2.0 2023-05-02 22:07:37,031 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5880, 4.5108, 4.6598, 4.8067, 4.9223, 4.4802, 4.8843, 4.9756], device='cuda:0'), covar=tensor([0.1741, 0.1207, 0.1540, 0.0690, 0.0518, 0.1071, 0.0724, 0.0567], device='cuda:0'), in_proj_covar=tensor([0.0719, 0.0871, 0.1006, 0.0890, 0.0676, 0.0703, 0.0739, 0.0862], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:08:02,607 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7924, 2.9519, 2.4973, 2.7248, 3.2526, 2.9314, 3.3154, 3.4244], device='cuda:0'), covar=tensor([0.0089, 0.0382, 0.0535, 0.0424, 0.0262, 0.0367, 0.0249, 0.0246], device='cuda:0'), in_proj_covar=tensor([0.0241, 0.0250, 0.0240, 0.0241, 0.0251, 0.0249, 0.0250, 0.0252], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:08:22,579 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.38 vs. limit=5.0 2023-05-02 22:08:37,718 INFO [train.py:904] (0/8) Epoch 30, batch 4100, loss[loss=0.1743, simple_loss=0.2717, pruned_loss=0.03845, over 16748.00 frames. ], tot_loss[loss=0.1677, simple_loss=0.2506, pruned_loss=0.04239, over 3272738.01 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:08:57,904 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.59 vs. limit=5.0 2023-05-02 22:09:19,659 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:09:53,969 INFO [train.py:904] (0/8) Epoch 30, batch 4150, loss[loss=0.1696, simple_loss=0.2683, pruned_loss=0.03539, over 16715.00 frames. ], tot_loss[loss=0.1725, simple_loss=0.2569, pruned_loss=0.04404, over 3230323.89 frames. ], batch size: 89, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:09:56,060 INFO [optim.py:368] (0/8) 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 22:10:35,422 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298530.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:10:55,802 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298543.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:11:12,262 INFO [train.py:904] (0/8) Epoch 30, batch 4200, loss[loss=0.189, simple_loss=0.2889, pruned_loss=0.04451, over 16787.00 frames. ], tot_loss[loss=0.1778, simple_loss=0.2639, pruned_loss=0.04579, over 3205595.19 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:08,241 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298591.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:12:26,985 INFO [train.py:904] (0/8) Epoch 30, batch 4250, loss[loss=0.1681, simple_loss=0.2666, pruned_loss=0.03484, over 16695.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2677, pruned_loss=0.04546, over 3195126.58 frames. ], batch size: 134, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:12:28,280 INFO [optim.py:368] (0/8) 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 22:12:40,770 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=298613.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:12:59,963 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.9467, 5.3874, 5.5492, 5.1629, 5.3358, 5.8771, 5.3385, 5.0814], device='cuda:0'), covar=tensor([0.0997, 0.1797, 0.1870, 0.2033, 0.2417, 0.0889, 0.1606, 0.2350], device='cuda:0'), in_proj_covar=tensor([0.0436, 0.0647, 0.0718, 0.0526, 0.0701, 0.0738, 0.0555, 0.0699], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 22:13:04,124 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298629.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:13:41,912 INFO [train.py:904] (0/8) Epoch 30, batch 4300, loss[loss=0.185, simple_loss=0.2795, pruned_loss=0.04523, over 15409.00 frames. ], tot_loss[loss=0.1795, simple_loss=0.2696, pruned_loss=0.04477, over 3208102.03 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:13:52,429 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=298661.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:17,867 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2610, 5.4521, 5.1452, 4.8771, 4.4437, 5.3719, 5.2604, 4.9154], device='cuda:0'), covar=tensor([0.0820, 0.0453, 0.0442, 0.0374, 0.1752, 0.0399, 0.0279, 0.0707], device='cuda:0'), in_proj_covar=tensor([0.0325, 0.0493, 0.0383, 0.0384, 0.0378, 0.0442, 0.0262, 0.0455], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-02 22:14:35,074 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=298689.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:14:36,359 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298690.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:14:47,931 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0070, 5.2307, 5.0346, 5.0424, 4.7757, 4.7190, 4.6719, 5.3537], device='cuda:0'), covar=tensor([0.1088, 0.0811, 0.0989, 0.0884, 0.0804, 0.0992, 0.1207, 0.0749], device='cuda:0'), in_proj_covar=tensor([0.0736, 0.0889, 0.0729, 0.0692, 0.0570, 0.0564, 0.0748, 0.0700], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:14:54,788 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.98 vs. limit=5.0 2023-05-02 22:14:56,162 INFO [train.py:904] (0/8) Epoch 30, batch 4350, loss[loss=0.195, simple_loss=0.2928, pruned_loss=0.04854, over 16715.00 frames. ], tot_loss[loss=0.182, simple_loss=0.2727, pruned_loss=0.04564, over 3211583.78 frames. ], batch size: 89, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:14:57,384 INFO [optim.py:368] (0/8) 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 22:16:04,016 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=298750.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:16:09,614 INFO [train.py:904] (0/8) Epoch 30, batch 4400, loss[loss=0.1777, simple_loss=0.2657, pruned_loss=0.04484, over 16673.00 frames. ], tot_loss[loss=0.1838, simple_loss=0.2743, pruned_loss=0.04659, over 3228538.09 frames. ], batch size: 57, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:21,883 INFO [train.py:904] (0/8) Epoch 30, batch 4450, loss[loss=0.1963, simple_loss=0.2742, pruned_loss=0.05926, over 12054.00 frames. ], tot_loss[loss=0.1872, simple_loss=0.2776, pruned_loss=0.04839, over 3208201.41 frames. ], batch size: 247, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:17:23,560 INFO [optim.py:368] (0/8) 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 22:17:24,113 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9969, 3.1111, 2.7043, 2.8366, 3.3946, 2.9539, 3.5041, 3.5588], device='cuda:0'), covar=tensor([0.0088, 0.0358, 0.0489, 0.0408, 0.0233, 0.0376, 0.0203, 0.0217], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0246, 0.0236, 0.0237, 0.0247, 0.0245, 0.0246, 0.0247], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:18:09,771 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298838.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:18:32,507 INFO [train.py:904] (0/8) Epoch 30, batch 4500, loss[loss=0.2039, simple_loss=0.2871, pruned_loss=0.06037, over 17121.00 frames. ], tot_loss[loss=0.1878, simple_loss=0.2777, pruned_loss=0.04899, over 3218555.54 frames. ], batch size: 49, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:18:34,337 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7358, 2.8971, 2.5051, 2.5279, 3.1884, 2.7298, 3.2912, 3.3511], device='cuda:0'), covar=tensor([0.0101, 0.0333, 0.0465, 0.0462, 0.0232, 0.0363, 0.0241, 0.0215], device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0247, 0.0236, 0.0238, 0.0247, 0.0245, 0.0246, 0.0248], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:18:48,234 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.28 vs. limit=2.0 2023-05-02 22:19:06,123 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 22:19:19,732 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298886.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:19:44,883 INFO [train.py:904] (0/8) Epoch 30, batch 4550, loss[loss=0.2039, simple_loss=0.2891, pruned_loss=0.05939, over 16710.00 frames. ], tot_loss[loss=0.1896, simple_loss=0.2791, pruned_loss=0.05006, over 3226085.12 frames. ], batch size: 62, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:19:45,304 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6424, 4.3916, 4.2252, 2.9371, 3.8255, 4.3300, 3.7734, 2.6627], device='cuda:0'), covar=tensor([0.0551, 0.0036, 0.0053, 0.0403, 0.0102, 0.0084, 0.0108, 0.0424], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0091, 0.0093, 0.0136, 0.0105, 0.0117, 0.0100, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 22:19:46,103 INFO [optim.py:368] (0/8) 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 22:20:57,304 INFO [train.py:904] (0/8) Epoch 30, batch 4600, loss[loss=0.1925, simple_loss=0.2798, pruned_loss=0.0526, over 16284.00 frames. ], tot_loss[loss=0.1914, simple_loss=0.2805, pruned_loss=0.05113, over 3211392.09 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:21:14,564 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7610, 4.7724, 4.5192, 3.7474, 4.6543, 1.7992, 4.3899, 3.9955], device='cuda:0'), covar=tensor([0.0076, 0.0068, 0.0164, 0.0344, 0.0065, 0.3037, 0.0105, 0.0268], device='cuda:0'), in_proj_covar=tensor([0.0188, 0.0179, 0.0218, 0.0190, 0.0196, 0.0222, 0.0207, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:21:42,907 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=298985.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 22:21:50,164 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.46 vs. limit=2.0 2023-05-02 22:21:59,693 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.15 vs. limit=2.0 2023-05-02 22:22:09,582 INFO [train.py:904] (0/8) Epoch 30, batch 4650, loss[loss=0.181, simple_loss=0.2726, pruned_loss=0.04473, over 16272.00 frames. ], tot_loss[loss=0.1912, simple_loss=0.2799, pruned_loss=0.05121, over 3226481.52 frames. ], batch size: 165, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:22:10,055 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.6628, 4.5550, 4.6900, 4.8379, 4.9726, 4.5130, 4.9688, 5.0220], device='cuda:0'), covar=tensor([0.1538, 0.1090, 0.1374, 0.0643, 0.0442, 0.1104, 0.0596, 0.0486], device='cuda:0'), in_proj_covar=tensor([0.0696, 0.0843, 0.0977, 0.0862, 0.0654, 0.0681, 0.0715, 0.0832], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:22:10,878 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.373e+02 1.831e+02 2.100e+02 2.427e+02 4.549e+02, threshold=4.199e+02, percent-clipped=1.0 2023-05-02 22:23:09,783 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299045.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:23:22,504 INFO [train.py:904] (0/8) Epoch 30, batch 4700, loss[loss=0.1739, simple_loss=0.2582, pruned_loss=0.04485, over 11544.00 frames. ], tot_loss[loss=0.1885, simple_loss=0.2768, pruned_loss=0.05007, over 3229949.39 frames. ], batch size: 248, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:23:47,783 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1918, 5.4319, 5.2698, 5.2884, 5.0110, 4.9241, 4.8452, 5.5607], device='cuda:0'), covar=tensor([0.1139, 0.0793, 0.0870, 0.0684, 0.0724, 0.0851, 0.1109, 0.0770], device='cuda:0'), in_proj_covar=tensor([0.0728, 0.0877, 0.0721, 0.0683, 0.0563, 0.0558, 0.0736, 0.0691], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:24:36,521 INFO [train.py:904] (0/8) Epoch 30, batch 4750, loss[loss=0.1527, simple_loss=0.2462, pruned_loss=0.02958, over 16745.00 frames. ], tot_loss[loss=0.1836, simple_loss=0.2721, pruned_loss=0.04751, over 3243932.40 frames. ], batch size: 89, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:24:37,720 INFO [optim.py:368] (0/8) 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 22:25:07,700 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299126.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:24,807 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6595, 2.6392, 2.5824, 4.4219, 2.9941, 3.9354, 1.6570, 2.8566], device='cuda:0'), covar=tensor([0.1410, 0.0837, 0.1241, 0.0178, 0.0214, 0.0385, 0.1649, 0.0870], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0208, 0.0208, 0.0219, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:25:25,749 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299138.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:25:49,342 INFO [train.py:904] (0/8) Epoch 30, batch 4800, loss[loss=0.1816, simple_loss=0.2736, pruned_loss=0.04483, over 16794.00 frames. ], tot_loss[loss=0.1798, simple_loss=0.2686, pruned_loss=0.04546, over 3239550.37 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:26:37,232 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:37,344 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299186.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:38,555 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299187.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:26:45,020 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299191.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:27:04,362 INFO [train.py:904] (0/8) Epoch 30, batch 4850, loss[loss=0.2338, simple_loss=0.305, pruned_loss=0.08127, over 11697.00 frames. ], tot_loss[loss=0.1793, simple_loss=0.2691, pruned_loss=0.04475, over 3226103.44 frames. ], batch size: 246, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:27:06,460 INFO [optim.py:368] (0/8) 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 22:27:49,337 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299234.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:02,963 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299243.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:16,461 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299252.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:28:18,997 INFO [train.py:904] (0/8) Epoch 30, batch 4900, loss[loss=0.1623, simple_loss=0.2611, pruned_loss=0.03178, over 16817.00 frames. ], tot_loss[loss=0.1775, simple_loss=0.2684, pruned_loss=0.04332, over 3216187.48 frames. ], batch size: 102, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:28:48,088 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2224, 3.4379, 3.6342, 2.3010, 2.9899, 2.4881, 3.5972, 3.7726], device='cuda:0'), covar=tensor([0.0293, 0.0852, 0.0671, 0.2088, 0.0952, 0.0983, 0.0648, 0.0929], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0173, 0.0172, 0.0159, 0.0149, 0.0134, 0.0147, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 22:29:06,069 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299285.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:29:33,779 INFO [train.py:904] (0/8) Epoch 30, batch 4950, loss[loss=0.1805, simple_loss=0.2742, pruned_loss=0.04337, over 16737.00 frames. ], tot_loss[loss=0.1763, simple_loss=0.2674, pruned_loss=0.04261, over 3199164.55 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:29:34,214 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299304.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:29:34,849 INFO [optim.py:368] (0/8) 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 22:30:02,182 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299323.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:30:17,901 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299333.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:30:36,100 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299345.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:30:47,910 INFO [train.py:904] (0/8) Epoch 30, batch 5000, loss[loss=0.1988, simple_loss=0.2919, pruned_loss=0.05285, over 16598.00 frames. ], tot_loss[loss=0.1774, simple_loss=0.2691, pruned_loss=0.0428, over 3197366.73 frames. ], batch size: 75, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:31:32,954 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299384.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:31:45,864 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299393.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:32:01,706 INFO [train.py:904] (0/8) Epoch 30, batch 5050, loss[loss=0.1762, simple_loss=0.2733, pruned_loss=0.03955, over 15444.00 frames. ], tot_loss[loss=0.1782, simple_loss=0.2706, pruned_loss=0.04293, over 3201146.18 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:32:02,879 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.469e+02 2.023e+02 2.400e+02 2.818e+02 3.992e+02, threshold=4.801e+02, percent-clipped=0.0 2023-05-02 22:32:21,511 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 22:33:13,775 INFO [train.py:904] (0/8) Epoch 30, batch 5100, loss[loss=0.1844, simple_loss=0.2794, pruned_loss=0.04472, over 16715.00 frames. ], tot_loss[loss=0.1771, simple_loss=0.2691, pruned_loss=0.04256, over 3192951.77 frames. ], batch size: 124, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:33:47,136 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0280, 3.2669, 3.5184, 2.0435, 2.9022, 2.4422, 3.5301, 3.4794], device='cuda:0'), covar=tensor([0.0261, 0.0801, 0.0617, 0.2123, 0.0888, 0.0919, 0.0538, 0.0926], device='cuda:0'), in_proj_covar=tensor([0.0163, 0.0174, 0.0173, 0.0159, 0.0150, 0.0134, 0.0147, 0.0187], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 22:33:57,982 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299482.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:34:30,234 INFO [train.py:904] (0/8) Epoch 30, batch 5150, loss[loss=0.156, simple_loss=0.2574, pruned_loss=0.02732, over 16854.00 frames. ], tot_loss[loss=0.176, simple_loss=0.2686, pruned_loss=0.04167, over 3189237.99 frames. ], batch size: 96, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:34:31,342 INFO [optim.py:368] (0/8) 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 22:35:33,157 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299547.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:35:40,564 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8245, 2.7704, 2.7170, 1.9301, 2.6361, 2.7786, 2.6476, 1.8575], device='cuda:0'), covar=tensor([0.0569, 0.0092, 0.0096, 0.0426, 0.0132, 0.0145, 0.0141, 0.0489], device='cuda:0'), in_proj_covar=tensor([0.0142, 0.0093, 0.0094, 0.0138, 0.0105, 0.0119, 0.0101, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 22:35:41,671 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.0982, 1.5787, 1.9348, 2.1378, 2.2512, 2.3770, 1.7858, 2.3315], device='cuda:0'), covar=tensor([0.0294, 0.0640, 0.0371, 0.0438, 0.0428, 0.0291, 0.0710, 0.0199], device='cuda:0'), in_proj_covar=tensor([0.0200, 0.0200, 0.0189, 0.0196, 0.0214, 0.0170, 0.0206, 0.0171], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 22:35:43,072 INFO [train.py:904] (0/8) Epoch 30, batch 5200, loss[loss=0.1654, simple_loss=0.2647, pruned_loss=0.03309, over 15402.00 frames. ], tot_loss[loss=0.1749, simple_loss=0.2675, pruned_loss=0.04121, over 3187592.67 frames. ], batch size: 191, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:36:48,435 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299599.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:36:55,753 INFO [train.py:904] (0/8) Epoch 30, batch 5250, loss[loss=0.1568, simple_loss=0.2525, pruned_loss=0.03061, over 16760.00 frames. ], tot_loss[loss=0.1736, simple_loss=0.2653, pruned_loss=0.04089, over 3176565.14 frames. ], batch size: 83, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:36:56,956 INFO [optim.py:368] (0/8) 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 22:37:20,748 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-02 22:37:41,429 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.2165, 5.2860, 5.5453, 5.5119, 5.5901, 5.2545, 5.1550, 4.9286], device='cuda:0'), covar=tensor([0.0255, 0.0482, 0.0312, 0.0345, 0.0379, 0.0305, 0.0935, 0.0528], device='cuda:0'), in_proj_covar=tensor([0.0441, 0.0503, 0.0482, 0.0447, 0.0527, 0.0509, 0.0588, 0.0412], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 22:38:06,904 INFO [train.py:904] (0/8) Epoch 30, batch 5300, loss[loss=0.159, simple_loss=0.2504, pruned_loss=0.03377, over 16429.00 frames. ], tot_loss[loss=0.1702, simple_loss=0.2613, pruned_loss=0.03953, over 3197132.61 frames. ], batch size: 146, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:38:42,892 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=299679.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:39:18,187 INFO [train.py:904] (0/8) Epoch 30, batch 5350, loss[loss=0.1695, simple_loss=0.2646, pruned_loss=0.03725, over 15299.00 frames. ], tot_loss[loss=0.1693, simple_loss=0.2599, pruned_loss=0.03938, over 3194963.37 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:39:19,347 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.376e+02 1.953e+02 2.222e+02 2.592e+02 5.137e+02, threshold=4.444e+02, percent-clipped=1.0 2023-05-02 22:39:22,932 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3276, 3.6212, 3.7678, 2.2047, 3.0676, 2.6685, 3.6113, 3.9509], device='cuda:0'), covar=tensor([0.0326, 0.0827, 0.0615, 0.2155, 0.0951, 0.0913, 0.0786, 0.0951], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0172, 0.0171, 0.0157, 0.0149, 0.0133, 0.0146, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 22:40:11,692 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 22:40:29,718 INFO [train.py:904] (0/8) Epoch 30, batch 5400, loss[loss=0.1762, simple_loss=0.2743, pruned_loss=0.03907, over 15441.00 frames. ], tot_loss[loss=0.1712, simple_loss=0.2625, pruned_loss=0.03991, over 3198374.65 frames. ], batch size: 190, lr: 2.24e-03, grad_scale: 8.0 2023-05-02 22:40:32,480 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.91 vs. limit=5.0 2023-05-02 22:40:51,575 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299769.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:09,370 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299782.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:41:34,570 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=299798.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:41:46,649 INFO [train.py:904] (0/8) Epoch 30, batch 5450, loss[loss=0.1825, simple_loss=0.2635, pruned_loss=0.05079, over 11847.00 frames. ], tot_loss[loss=0.174, simple_loss=0.2657, pruned_loss=0.04118, over 3192348.72 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:41:47,798 INFO [optim.py:368] (0/8) 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 22:42:27,033 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:27,206 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299830.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:42:52,073 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299847.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:43:02,914 INFO [train.py:904] (0/8) Epoch 30, batch 5500, loss[loss=0.2228, simple_loss=0.3016, pruned_loss=0.07198, over 15328.00 frames. ], tot_loss[loss=0.1817, simple_loss=0.2723, pruned_loss=0.04551, over 3160461.16 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:43:11,052 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=299859.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 22:44:09,164 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299895.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:14,104 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:44:21,858 INFO [train.py:904] (0/8) Epoch 30, batch 5550, loss[loss=0.19, simple_loss=0.2852, pruned_loss=0.04738, over 16633.00 frames. ], tot_loss[loss=0.1897, simple_loss=0.279, pruned_loss=0.05018, over 3134829.34 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:44:23,805 INFO [optim.py:368] (0/8) 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 22:44:36,119 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 22:45:31,925 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=299947.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:45:43,006 INFO [train.py:904] (0/8) Epoch 30, batch 5600, loss[loss=0.1892, simple_loss=0.2735, pruned_loss=0.05248, over 16006.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.2833, pruned_loss=0.05387, over 3092796.72 frames. ], batch size: 35, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:45:47,263 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9336, 2.6403, 2.5245, 1.9327, 2.4867, 2.6823, 2.5740, 1.9214], device='cuda:0'), covar=tensor([0.0465, 0.0126, 0.0118, 0.0408, 0.0174, 0.0163, 0.0147, 0.0427], device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0092, 0.0093, 0.0136, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 22:46:21,553 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7461, 3.8597, 2.4175, 4.7250, 3.0674, 4.5018, 2.5530, 3.1521], device='cuda:0'), covar=tensor([0.0338, 0.0434, 0.1862, 0.0201, 0.0905, 0.0579, 0.1658, 0.0899], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0183, 0.0196, 0.0175, 0.0181, 0.0222, 0.0205, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:46:26,621 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=299979.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:47:02,053 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-300000.pt 2023-05-02 22:47:11,378 INFO [train.py:904] (0/8) Epoch 30, batch 5650, loss[loss=0.2146, simple_loss=0.2986, pruned_loss=0.06529, over 16420.00 frames. ], tot_loss[loss=0.2015, simple_loss=0.2881, pruned_loss=0.05752, over 3063818.50 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:47:13,273 INFO [optim.py:368] (0/8) 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 22:47:37,587 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.96 vs. limit=5.0 2023-05-02 22:47:48,228 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300027.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:48:20,818 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7301, 2.4620, 2.3076, 3.3546, 2.1429, 3.5271, 1.6109, 2.7063], device='cuda:0'), covar=tensor([0.1440, 0.0860, 0.1366, 0.0259, 0.0208, 0.0428, 0.1871, 0.0910], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0207, 0.0207, 0.0218, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:48:27,604 INFO [train.py:904] (0/8) Epoch 30, batch 5700, loss[loss=0.217, simple_loss=0.3029, pruned_loss=0.06557, over 15362.00 frames. ], tot_loss[loss=0.2049, simple_loss=0.2903, pruned_loss=0.0597, over 3047351.26 frames. ], batch size: 190, lr: 2.23e-03, grad_scale: 16.0 2023-05-02 22:48:57,389 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300073.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:49:45,350 INFO [train.py:904] (0/8) Epoch 30, batch 5750, loss[loss=0.224, simple_loss=0.2929, pruned_loss=0.07754, over 10741.00 frames. ], tot_loss[loss=0.2083, simple_loss=0.2932, pruned_loss=0.06173, over 3007649.29 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:49:49,229 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.972e+02 3.119e+02 3.894e+02 5.149e+02 1.112e+03, threshold=7.788e+02, percent-clipped=3.0 2023-05-02 22:50:18,716 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300125.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:50:35,036 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:51:06,738 INFO [train.py:904] (0/8) Epoch 30, batch 5800, loss[loss=0.2179, simple_loss=0.2898, pruned_loss=0.07303, over 11672.00 frames. ], tot_loss[loss=0.2058, simple_loss=0.2923, pruned_loss=0.05959, over 3020220.17 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:51:07,286 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300154.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 22:51:34,572 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7704, 3.8419, 2.4388, 4.5046, 3.0131, 4.3816, 2.5764, 3.1251], device='cuda:0'), covar=tensor([0.0307, 0.0409, 0.1685, 0.0271, 0.0801, 0.0601, 0.1535, 0.0877], device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0183, 0.0197, 0.0175, 0.0181, 0.0222, 0.0206, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:52:25,625 INFO [train.py:904] (0/8) Epoch 30, batch 5850, loss[loss=0.1984, simple_loss=0.2913, pruned_loss=0.05277, over 16757.00 frames. ], tot_loss[loss=0.203, simple_loss=0.29, pruned_loss=0.05795, over 3042308.90 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:52:28,951 INFO [optim.py:368] (0/8) 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 22:52:36,271 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5127, 3.5776, 3.3707, 3.0281, 3.1528, 3.4818, 3.3351, 3.2952], device='cuda:0'), covar=tensor([0.0691, 0.0681, 0.0362, 0.0344, 0.0507, 0.0583, 0.1415, 0.0506], device='cuda:0'), in_proj_covar=tensor([0.0317, 0.0482, 0.0373, 0.0374, 0.0369, 0.0431, 0.0255, 0.0443], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:52:53,934 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5045, 2.5954, 2.2424, 2.3419, 2.9943, 2.6515, 2.9573, 3.1853], device='cuda:0'), covar=tensor([0.0147, 0.0482, 0.0585, 0.0501, 0.0284, 0.0394, 0.0252, 0.0294], device='cuda:0'), in_proj_covar=tensor([0.0233, 0.0243, 0.0233, 0.0234, 0.0244, 0.0241, 0.0242, 0.0243], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:52:56,858 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2771, 3.9215, 3.9060, 2.4411, 3.5929, 3.9955, 3.6040, 2.0012], device='cuda:0'), covar=tensor([0.0713, 0.0081, 0.0082, 0.0551, 0.0123, 0.0144, 0.0126, 0.0624], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0093, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 22:53:48,217 INFO [train.py:904] (0/8) Epoch 30, batch 5900, loss[loss=0.1847, simple_loss=0.2741, pruned_loss=0.0477, over 16447.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.289, pruned_loss=0.05742, over 3052203.15 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:54:47,622 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.81 vs. limit=5.0 2023-05-02 22:55:08,082 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 22:55:09,241 INFO [train.py:904] (0/8) Epoch 30, batch 5950, loss[loss=0.198, simple_loss=0.2907, pruned_loss=0.05263, over 16784.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2898, pruned_loss=0.05616, over 3069838.59 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:55:12,898 INFO [optim.py:368] (0/8) 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 22:55:37,942 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8533, 2.8032, 2.5843, 4.8258, 3.5589, 4.0709, 1.6591, 2.9558], device='cuda:0'), covar=tensor([0.1352, 0.0833, 0.1329, 0.0157, 0.0322, 0.0461, 0.1701, 0.0917], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0208, 0.0208, 0.0219, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 22:56:29,066 INFO [train.py:904] (0/8) Epoch 30, batch 6000, loss[loss=0.213, simple_loss=0.2922, pruned_loss=0.06687, over 11368.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.2882, pruned_loss=0.0553, over 3082061.04 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:56:29,067 INFO [train.py:929] (0/8) Computing validation loss 2023-05-02 22:56:39,791 INFO [train.py:938] (0/8) Epoch 30, validation: loss=0.1471, simple_loss=0.2591, pruned_loss=0.01755, over 944034.00 frames. 2023-05-02 22:56:39,792 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-02 22:57:02,248 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.64 vs. limit=2.0 2023-05-02 22:57:57,808 INFO [train.py:904] (0/8) Epoch 30, batch 6050, loss[loss=0.1825, simple_loss=0.2834, pruned_loss=0.04083, over 16755.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2871, pruned_loss=0.05469, over 3092950.67 frames. ], batch size: 83, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:58:01,300 INFO [optim.py:368] (0/8) 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 22:58:29,375 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300425.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:58:35,432 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300429.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 22:58:53,843 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4817, 1.7706, 2.1796, 2.4046, 2.4755, 2.7293, 1.9092, 2.6714], device='cuda:0'), covar=tensor([0.0254, 0.0569, 0.0361, 0.0370, 0.0383, 0.0224, 0.0628, 0.0178], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0199, 0.0187, 0.0195, 0.0211, 0.0169, 0.0205, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 22:59:15,666 INFO [train.py:904] (0/8) Epoch 30, batch 6100, loss[loss=0.2031, simple_loss=0.2949, pruned_loss=0.05561, over 16885.00 frames. ], tot_loss[loss=0.1974, simple_loss=0.2867, pruned_loss=0.05411, over 3104413.58 frames. ], batch size: 116, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 22:59:16,250 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300454.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 22:59:20,448 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3371, 3.8102, 3.7508, 2.4091, 3.5324, 3.8649, 3.5457, 2.1662], device='cuda:0'), covar=tensor([0.0664, 0.0070, 0.0092, 0.0549, 0.0127, 0.0137, 0.0122, 0.0568], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0093, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 22:59:46,871 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300473.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:00:11,106 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.66 vs. limit=5.0 2023-05-02 23:00:32,803 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300502.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 23:00:34,973 INFO [train.py:904] (0/8) Epoch 30, batch 6150, loss[loss=0.1644, simple_loss=0.2653, pruned_loss=0.03176, over 16839.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2847, pruned_loss=0.05366, over 3097735.22 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:00:38,228 INFO [optim.py:368] (0/8) 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 23:01:53,681 INFO [train.py:904] (0/8) Epoch 30, batch 6200, loss[loss=0.1736, simple_loss=0.2654, pruned_loss=0.0409, over 16608.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2828, pruned_loss=0.0534, over 3086809.76 frames. ], batch size: 62, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:13,169 INFO [train.py:904] (0/8) Epoch 30, batch 6250, loss[loss=0.197, simple_loss=0.2853, pruned_loss=0.05429, over 16542.00 frames. ], tot_loss[loss=0.1948, simple_loss=0.2827, pruned_loss=0.05341, over 3086379.30 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:03:16,360 INFO [optim.py:368] (0/8) 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 23:03:34,775 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300617.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:04:31,374 INFO [train.py:904] (0/8) Epoch 30, batch 6300, loss[loss=0.2084, simple_loss=0.2964, pruned_loss=0.06025, over 16892.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.282, pruned_loss=0.05227, over 3087768.18 frames. ], batch size: 109, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:10,880 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300678.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:05:23,252 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.1681, 5.1797, 4.9844, 4.2353, 5.0881, 1.7472, 4.7990, 4.6329], device='cuda:0'), covar=tensor([0.0167, 0.0143, 0.0235, 0.0464, 0.0137, 0.3168, 0.0168, 0.0298], device='cuda:0'), in_proj_covar=tensor([0.0187, 0.0178, 0.0216, 0.0189, 0.0194, 0.0220, 0.0206, 0.0182], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:05:50,451 INFO [train.py:904] (0/8) Epoch 30, batch 6350, loss[loss=0.2021, simple_loss=0.2896, pruned_loss=0.05729, over 15452.00 frames. ], tot_loss[loss=0.1962, simple_loss=0.2837, pruned_loss=0.05436, over 3050529.19 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:05:53,985 INFO [optim.py:368] (0/8) 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 23:06:29,155 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=300729.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:07:07,476 INFO [train.py:904] (0/8) Epoch 30, batch 6400, loss[loss=0.1976, simple_loss=0.2879, pruned_loss=0.05363, over 16745.00 frames. ], tot_loss[loss=0.1955, simple_loss=0.283, pruned_loss=0.05406, over 3092359.03 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:07:42,793 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=300777.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:07:59,881 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6428, 2.5989, 2.4224, 3.7206, 2.5250, 3.8177, 1.4587, 2.8570], device='cuda:0'), covar=tensor([0.1488, 0.0828, 0.1303, 0.0180, 0.0227, 0.0527, 0.1923, 0.0833], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0183, 0.0202, 0.0207, 0.0208, 0.0220, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 23:08:18,875 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=300802.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:08:21,343 INFO [train.py:904] (0/8) Epoch 30, batch 6450, loss[loss=0.1936, simple_loss=0.2791, pruned_loss=0.05401, over 15311.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2825, pruned_loss=0.05318, over 3096970.63 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:08:24,287 INFO [optim.py:368] (0/8) 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 23:08:26,151 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9399, 2.7183, 2.5621, 1.8917, 2.5740, 2.7674, 2.6479, 1.9173], device='cuda:0'), covar=tensor([0.0485, 0.0131, 0.0129, 0.0445, 0.0171, 0.0159, 0.0143, 0.0478], device='cuda:0'), in_proj_covar=tensor([0.0141, 0.0093, 0.0094, 0.0137, 0.0105, 0.0118, 0.0101, 0.0133], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 23:09:38,125 INFO [train.py:904] (0/8) Epoch 30, batch 6500, loss[loss=0.1878, simple_loss=0.2757, pruned_loss=0.04994, over 16689.00 frames. ], tot_loss[loss=0.193, simple_loss=0.2808, pruned_loss=0.0526, over 3120647.06 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:09:52,580 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=300863.0, num_to_drop=1, layers_to_drop={2} 2023-05-02 23:10:03,571 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8833, 3.1258, 2.6587, 5.0905, 3.8104, 4.3082, 1.7271, 3.2070], device='cuda:0'), covar=tensor([0.1366, 0.0769, 0.1331, 0.0154, 0.0354, 0.0422, 0.1775, 0.0833], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0182, 0.0202, 0.0207, 0.0208, 0.0219, 0.0212, 0.0201], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 23:10:12,610 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5247, 2.2662, 1.8988, 2.0365, 2.5267, 2.2124, 2.2559, 2.6488], device='cuda:0'), covar=tensor([0.0265, 0.0446, 0.0597, 0.0494, 0.0293, 0.0398, 0.0227, 0.0286], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0241, 0.0232, 0.0232, 0.0243, 0.0239, 0.0239, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:10:58,885 INFO [train.py:904] (0/8) Epoch 30, batch 6550, loss[loss=0.2246, simple_loss=0.3098, pruned_loss=0.06967, over 16344.00 frames. ], tot_loss[loss=0.1944, simple_loss=0.2828, pruned_loss=0.05296, over 3124301.92 frames. ], batch size: 35, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:11:01,656 INFO [optim.py:368] (0/8) 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 23:11:04,387 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2188, 2.0567, 1.8039, 1.7766, 2.2944, 1.9639, 1.8868, 2.3619], device='cuda:0'), covar=tensor([0.0304, 0.0511, 0.0603, 0.0558, 0.0319, 0.0445, 0.0260, 0.0334], device='cuda:0'), in_proj_covar=tensor([0.0231, 0.0241, 0.0232, 0.0232, 0.0243, 0.0239, 0.0239, 0.0242], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:12:19,432 INFO [train.py:904] (0/8) Epoch 30, batch 6600, loss[loss=0.1879, simple_loss=0.2816, pruned_loss=0.04714, over 16392.00 frames. ], tot_loss[loss=0.196, simple_loss=0.2853, pruned_loss=0.05336, over 3120368.24 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:12:42,097 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8492, 3.9271, 4.1178, 4.0988, 4.1327, 3.9192, 3.9204, 3.9088], device='cuda:0'), covar=tensor([0.0384, 0.0641, 0.0473, 0.0456, 0.0486, 0.0487, 0.0799, 0.0574], device='cuda:0'), in_proj_covar=tensor([0.0442, 0.0506, 0.0483, 0.0448, 0.0529, 0.0512, 0.0587, 0.0412], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 23:12:48,787 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=300973.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:13:38,998 INFO [train.py:904] (0/8) Epoch 30, batch 6650, loss[loss=0.1996, simple_loss=0.2859, pruned_loss=0.05663, over 16689.00 frames. ], tot_loss[loss=0.1965, simple_loss=0.2849, pruned_loss=0.05405, over 3122038.14 frames. ], batch size: 134, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:13:41,670 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.69 vs. limit=2.0 2023-05-02 23:13:43,934 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.884e+02 2.674e+02 3.483e+02 3.945e+02 7.489e+02, threshold=6.965e+02, percent-clipped=2.0 2023-05-02 23:13:49,618 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.33 vs. limit=2.0 2023-05-02 23:14:06,381 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.20 vs. limit=2.0 2023-05-02 23:14:55,127 INFO [train.py:904] (0/8) Epoch 30, batch 6700, loss[loss=0.2184, simple_loss=0.2954, pruned_loss=0.07069, over 11442.00 frames. ], tot_loss[loss=0.1973, simple_loss=0.2849, pruned_loss=0.05485, over 3107793.07 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:15:44,008 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5261, 3.5859, 3.3867, 3.0126, 3.1952, 3.4942, 3.3424, 3.3161], device='cuda:0'), covar=tensor([0.0556, 0.0572, 0.0318, 0.0290, 0.0502, 0.0498, 0.1155, 0.0475], device='cuda:0'), in_proj_covar=tensor([0.0313, 0.0477, 0.0367, 0.0370, 0.0365, 0.0426, 0.0253, 0.0438], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:16:07,684 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1658, 3.2533, 3.7195, 2.1158, 3.1798, 2.3837, 3.5638, 3.6099], device='cuda:0'), covar=tensor([0.0252, 0.0956, 0.0579, 0.2228, 0.0835, 0.1001, 0.0595, 0.1081], device='cuda:0'), in_proj_covar=tensor([0.0161, 0.0172, 0.0172, 0.0158, 0.0149, 0.0134, 0.0146, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 23:16:11,413 INFO [train.py:904] (0/8) Epoch 30, batch 6750, loss[loss=0.1695, simple_loss=0.2678, pruned_loss=0.03563, over 16842.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.284, pruned_loss=0.05512, over 3099065.04 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:16:15,736 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.912e+02 2.866e+02 3.351e+02 4.005e+02 5.751e+02, threshold=6.702e+02, percent-clipped=0.0 2023-05-02 23:17:07,589 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301140.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:17:29,024 INFO [train.py:904] (0/8) Epoch 30, batch 6800, loss[loss=0.25, simple_loss=0.3166, pruned_loss=0.09164, over 11671.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.284, pruned_loss=0.05508, over 3104212.02 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:17:35,474 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301158.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:17:57,700 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9196, 2.7175, 2.5921, 1.9151, 2.5682, 2.7369, 2.6217, 1.9335], device='cuda:0'), covar=tensor([0.0498, 0.0108, 0.0118, 0.0426, 0.0158, 0.0157, 0.0146, 0.0434], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0092, 0.0093, 0.0136, 0.0104, 0.0117, 0.0100, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 23:18:43,923 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301201.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:18:47,474 INFO [train.py:904] (0/8) Epoch 30, batch 6850, loss[loss=0.1867, simple_loss=0.2868, pruned_loss=0.04327, over 16500.00 frames. ], tot_loss[loss=0.1994, simple_loss=0.286, pruned_loss=0.05637, over 3072578.45 frames. ], batch size: 75, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:18:51,709 INFO [optim.py:368] (0/8) 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 23:20:04,231 INFO [train.py:904] (0/8) Epoch 30, batch 6900, loss[loss=0.192, simple_loss=0.2836, pruned_loss=0.05018, over 16409.00 frames. ], tot_loss[loss=0.1999, simple_loss=0.2881, pruned_loss=0.05586, over 3079450.48 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:20:21,874 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5303, 3.5403, 2.6604, 2.2015, 2.3357, 2.4158, 3.7407, 3.1870], device='cuda:0'), covar=tensor([0.3176, 0.0718, 0.1989, 0.3085, 0.2717, 0.2200, 0.0507, 0.1415], device='cuda:0'), in_proj_covar=tensor([0.0339, 0.0277, 0.0316, 0.0330, 0.0307, 0.0281, 0.0305, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 23:20:34,867 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301273.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:20:50,634 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5099, 2.2221, 1.9349, 1.9678, 2.5050, 2.1656, 2.2880, 2.6545], device='cuda:0'), covar=tensor([0.0285, 0.0510, 0.0597, 0.0559, 0.0321, 0.0432, 0.0251, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0234, 0.0244, 0.0235, 0.0236, 0.0246, 0.0242, 0.0242, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:21:22,707 INFO [train.py:904] (0/8) Epoch 30, batch 6950, loss[loss=0.1827, simple_loss=0.2691, pruned_loss=0.04814, over 16424.00 frames. ], tot_loss[loss=0.2019, simple_loss=0.2894, pruned_loss=0.05716, over 3076686.36 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:21:26,941 INFO [optim.py:368] (0/8) 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 23:21:46,999 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.1236, 4.1884, 4.4375, 4.4221, 4.4406, 4.1876, 4.1921, 4.1808], device='cuda:0'), covar=tensor([0.0389, 0.0630, 0.0442, 0.0410, 0.0501, 0.0476, 0.0949, 0.0531], device='cuda:0'), in_proj_covar=tensor([0.0445, 0.0508, 0.0485, 0.0449, 0.0530, 0.0514, 0.0591, 0.0413], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 23:21:50,651 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301321.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:22:38,944 INFO [train.py:904] (0/8) Epoch 30, batch 7000, loss[loss=0.1966, simple_loss=0.3028, pruned_loss=0.04522, over 16530.00 frames. ], tot_loss[loss=0.2011, simple_loss=0.2895, pruned_loss=0.05629, over 3085430.78 frames. ], batch size: 68, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:22:52,629 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9459, 2.7170, 2.8640, 2.1645, 2.7011, 2.1797, 2.6977, 2.9239], device='cuda:0'), covar=tensor([0.0304, 0.0913, 0.0514, 0.1840, 0.0842, 0.0996, 0.0578, 0.0781], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0173, 0.0172, 0.0158, 0.0149, 0.0134, 0.0146, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 23:22:55,905 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5476, 2.9207, 3.1469, 1.9051, 2.7889, 2.1041, 3.0618, 3.2325], device='cuda:0'), covar=tensor([0.0315, 0.0902, 0.0661, 0.2335, 0.0952, 0.1141, 0.0725, 0.0991], device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0173, 0.0172, 0.0158, 0.0149, 0.0134, 0.0146, 0.0185], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0005], device='cuda:0') 2023-05-02 23:23:39,590 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.11 vs. limit=2.0 2023-05-02 23:23:55,574 INFO [train.py:904] (0/8) Epoch 30, batch 7050, loss[loss=0.1963, simple_loss=0.281, pruned_loss=0.05577, over 16379.00 frames. ], tot_loss[loss=0.2008, simple_loss=0.2901, pruned_loss=0.0558, over 3083780.22 frames. ], batch size: 146, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:24:00,751 INFO [optim.py:368] (0/8) 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 23:24:38,212 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301430.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:25:14,090 INFO [train.py:904] (0/8) Epoch 30, batch 7100, loss[loss=0.197, simple_loss=0.2851, pruned_loss=0.05446, over 16786.00 frames. ], tot_loss[loss=0.1989, simple_loss=0.288, pruned_loss=0.05487, over 3107774.93 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:25:21,020 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301458.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:25:39,956 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.44 vs. limit=2.0 2023-05-02 23:25:58,174 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.93 vs. limit=5.0 2023-05-02 23:25:58,455 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.38 vs. limit=5.0 2023-05-02 23:26:03,698 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.7043, 1.8537, 1.6921, 1.5222, 1.9677, 1.5932, 1.5826, 1.9018], device='cuda:0'), covar=tensor([0.0230, 0.0293, 0.0459, 0.0408, 0.0230, 0.0310, 0.0177, 0.0230], device='cuda:0'), in_proj_covar=tensor([0.0235, 0.0245, 0.0235, 0.0236, 0.0247, 0.0243, 0.0242, 0.0245], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:26:09,741 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5643, 2.7146, 2.7558, 4.4057, 2.5712, 3.0386, 2.7123, 2.8577], device='cuda:0'), covar=tensor([0.1319, 0.3118, 0.2749, 0.0509, 0.3746, 0.2251, 0.3201, 0.3066], device='cuda:0'), in_proj_covar=tensor([0.0427, 0.0479, 0.0391, 0.0340, 0.0449, 0.0551, 0.0453, 0.0562], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:26:14,409 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301491.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:21,069 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301496.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:33,763 INFO [train.py:904] (0/8) Epoch 30, batch 7150, loss[loss=0.191, simple_loss=0.284, pruned_loss=0.04896, over 16808.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2867, pruned_loss=0.05481, over 3108164.04 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:26:36,533 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301506.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:26:37,392 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.858e+02 2.724e+02 3.109e+02 4.036e+02 9.536e+02, threshold=6.218e+02, percent-clipped=1.0 2023-05-02 23:27:49,038 INFO [train.py:904] (0/8) Epoch 30, batch 7200, loss[loss=0.1779, simple_loss=0.2684, pruned_loss=0.04373, over 16697.00 frames. ], tot_loss[loss=0.1957, simple_loss=0.2845, pruned_loss=0.05341, over 3099976.96 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:10,986 INFO [train.py:904] (0/8) Epoch 30, batch 7250, loss[loss=0.1693, simple_loss=0.2594, pruned_loss=0.03964, over 16869.00 frames. ], tot_loss[loss=0.1933, simple_loss=0.2821, pruned_loss=0.05223, over 3114090.45 frames. ], batch size: 96, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:29:15,149 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.563e+02 2.328e+02 2.684e+02 3.448e+02 5.554e+02, threshold=5.368e+02, percent-clipped=0.0 2023-05-02 23:29:21,954 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.94 vs. limit=5.0 2023-05-02 23:29:52,601 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5655, 4.6462, 4.4506, 4.1522, 4.1231, 4.5615, 4.3055, 4.3041], device='cuda:0'), covar=tensor([0.0598, 0.0494, 0.0301, 0.0304, 0.0854, 0.0440, 0.0586, 0.0580], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0472, 0.0364, 0.0365, 0.0361, 0.0421, 0.0251, 0.0433], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:29:57,328 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=2.01 vs. limit=2.0 2023-05-02 23:30:01,827 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 23:30:27,551 INFO [train.py:904] (0/8) Epoch 30, batch 7300, loss[loss=0.2284, simple_loss=0.3045, pruned_loss=0.07616, over 11119.00 frames. ], tot_loss[loss=0.1943, simple_loss=0.2822, pruned_loss=0.05317, over 3075714.44 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:30:43,295 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-02 23:31:44,860 INFO [train.py:904] (0/8) Epoch 30, batch 7350, loss[loss=0.1896, simple_loss=0.275, pruned_loss=0.05206, over 16658.00 frames. ], tot_loss[loss=0.1958, simple_loss=0.2836, pruned_loss=0.05404, over 3077018.65 frames. ], batch size: 57, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:31:46,546 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1757, 1.5740, 1.9587, 2.1004, 2.2162, 2.4162, 1.7446, 2.3124], device='cuda:0'), covar=tensor([0.0281, 0.0590, 0.0339, 0.0421, 0.0379, 0.0260, 0.0585, 0.0177], device='cuda:0'), in_proj_covar=tensor([0.0199, 0.0200, 0.0188, 0.0195, 0.0212, 0.0169, 0.0205, 0.0169], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002], device='cuda:0') 2023-05-02 23:31:50,967 INFO [optim.py:368] (0/8) 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 23:31:55,561 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4986, 3.5669, 3.3275, 2.9353, 3.1793, 3.4723, 3.3217, 3.3222], device='cuda:0'), covar=tensor([0.0585, 0.0626, 0.0292, 0.0268, 0.0463, 0.0493, 0.1323, 0.0469], device='cuda:0'), in_proj_covar=tensor([0.0309, 0.0470, 0.0362, 0.0364, 0.0360, 0.0419, 0.0250, 0.0432], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:32:15,053 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.9460, 4.1560, 3.9670, 4.0078, 3.7502, 3.8138, 3.8053, 4.1315], device='cuda:0'), covar=tensor([0.1078, 0.0861, 0.1061, 0.0872, 0.0749, 0.1616, 0.0967, 0.1064], device='cuda:0'), in_proj_covar=tensor([0.0721, 0.0869, 0.0717, 0.0679, 0.0554, 0.0555, 0.0730, 0.0683], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:32:44,217 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301743.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:33:00,871 INFO [train.py:904] (0/8) Epoch 30, batch 7400, loss[loss=0.2042, simple_loss=0.294, pruned_loss=0.05717, over 16231.00 frames. ], tot_loss[loss=0.1978, simple_loss=0.285, pruned_loss=0.05528, over 3059324.20 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:33:53,669 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=301786.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:10,821 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=301796.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:23,478 INFO [train.py:904] (0/8) Epoch 30, batch 7450, loss[loss=0.2146, simple_loss=0.2824, pruned_loss=0.07344, over 11078.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2864, pruned_loss=0.05612, over 3056590.79 frames. ], batch size: 248, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:34:24,270 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301804.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:34:29,850 INFO [optim.py:368] (0/8) 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 23:35:23,117 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.6414, 3.8868, 4.0666, 4.0214, 4.0299, 3.8496, 3.6103, 3.8586], device='cuda:0'), covar=tensor([0.0658, 0.0882, 0.0609, 0.0695, 0.0783, 0.0747, 0.1383, 0.0666], device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0501, 0.0478, 0.0444, 0.0522, 0.0507, 0.0582, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 23:35:30,254 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=301844.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:35:45,214 INFO [train.py:904] (0/8) Epoch 30, batch 7500, loss[loss=0.2415, simple_loss=0.3141, pruned_loss=0.08441, over 11485.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2868, pruned_loss=0.05576, over 3050018.93 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:36:57,889 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301899.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:05,281 INFO [train.py:904] (0/8) Epoch 30, batch 7550, loss[loss=0.1753, simple_loss=0.2676, pruned_loss=0.04147, over 16855.00 frames. ], tot_loss[loss=0.1987, simple_loss=0.2858, pruned_loss=0.05576, over 3050228.94 frames. ], batch size: 102, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:37:07,781 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=301905.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:37:11,511 INFO [optim.py:368] (0/8) 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 23:37:43,941 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8602, 2.2187, 2.4698, 3.1314, 2.2717, 2.4320, 2.3684, 2.3630], device='cuda:0'), covar=tensor([0.1480, 0.3175, 0.2487, 0.0823, 0.3992, 0.2270, 0.3126, 0.3182], device='cuda:0'), in_proj_covar=tensor([0.0427, 0.0481, 0.0391, 0.0341, 0.0450, 0.0552, 0.0454, 0.0563], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:38:23,962 INFO [train.py:904] (0/8) Epoch 30, batch 7600, loss[loss=0.2179, simple_loss=0.2887, pruned_loss=0.07353, over 11200.00 frames. ], tot_loss[loss=0.198, simple_loss=0.2846, pruned_loss=0.05575, over 3055108.23 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:38:34,149 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301960.0, num_to_drop=1, layers_to_drop={1} 2023-05-02 23:38:41,004 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8190, 2.8850, 2.4939, 4.5927, 3.2378, 3.9734, 1.7666, 2.9906], device='cuda:0'), covar=tensor([0.1492, 0.0875, 0.1464, 0.0179, 0.0330, 0.0480, 0.1844, 0.0925], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0183, 0.0203, 0.0209, 0.0209, 0.0221, 0.0213, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 23:38:43,721 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=301966.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:39:37,963 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-302000.pt 2023-05-02 23:39:47,155 INFO [train.py:904] (0/8) Epoch 30, batch 7650, loss[loss=0.2063, simple_loss=0.2949, pruned_loss=0.05884, over 15317.00 frames. ], tot_loss[loss=0.1993, simple_loss=0.2856, pruned_loss=0.05652, over 3041111.87 frames. ], batch size: 191, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:39:53,139 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 2.050e+02 2.820e+02 3.410e+02 4.459e+02 8.175e+02, threshold=6.820e+02, percent-clipped=3.0 2023-05-02 23:41:06,728 INFO [train.py:904] (0/8) Epoch 30, batch 7700, loss[loss=0.1702, simple_loss=0.2591, pruned_loss=0.04069, over 16236.00 frames. ], tot_loss[loss=0.1982, simple_loss=0.2847, pruned_loss=0.0558, over 3060988.47 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:41:24,992 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4095, 3.5075, 3.6341, 3.6179, 3.6387, 3.4633, 3.5039, 3.5165], device='cuda:0'), covar=tensor([0.0397, 0.0716, 0.0477, 0.0455, 0.0491, 0.0580, 0.0835, 0.0606], device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0501, 0.0479, 0.0444, 0.0522, 0.0507, 0.0585, 0.0409], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 23:41:57,528 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302086.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:18,123 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302099.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:42:26,401 INFO [train.py:904] (0/8) Epoch 30, batch 7750, loss[loss=0.2096, simple_loss=0.3062, pruned_loss=0.05651, over 16335.00 frames. ], tot_loss[loss=0.1977, simple_loss=0.2846, pruned_loss=0.0554, over 3075080.44 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:42:33,840 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.870e+02 2.636e+02 3.091e+02 3.789e+02 8.720e+02, threshold=6.183e+02, percent-clipped=2.0 2023-05-02 23:43:12,805 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302134.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:43:41,544 INFO [train.py:904] (0/8) Epoch 30, batch 7800, loss[loss=0.194, simple_loss=0.2796, pruned_loss=0.05422, over 16737.00 frames. ], tot_loss[loss=0.1983, simple_loss=0.2854, pruned_loss=0.05563, over 3101320.16 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:43:47,603 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.18 vs. limit=2.0 2023-05-02 23:44:55,043 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302200.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:45:00,020 INFO [train.py:904] (0/8) Epoch 30, batch 7850, loss[loss=0.1989, simple_loss=0.2872, pruned_loss=0.05524, over 16875.00 frames. ], tot_loss[loss=0.199, simple_loss=0.2865, pruned_loss=0.05572, over 3093447.41 frames. ], batch size: 42, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:45:06,746 INFO [optim.py:368] (0/8) 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 23:45:19,092 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3836, 3.5075, 3.6370, 3.6096, 3.6273, 3.4497, 3.4808, 3.5144], device='cuda:0'), covar=tensor([0.0460, 0.0727, 0.0542, 0.0492, 0.0577, 0.0595, 0.0896, 0.0615], device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0500, 0.0478, 0.0443, 0.0522, 0.0506, 0.0583, 0.0408], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 23:46:07,399 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8036, 4.3299, 3.1123, 2.4445, 2.8790, 2.7783, 4.7843, 3.7238], device='cuda:0'), covar=tensor([0.3273, 0.0629, 0.1931, 0.3075, 0.2905, 0.2153, 0.0399, 0.1362], device='cuda:0'), in_proj_covar=tensor([0.0341, 0.0277, 0.0318, 0.0332, 0.0310, 0.0283, 0.0306, 0.0355], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 23:46:15,236 INFO [train.py:904] (0/8) Epoch 30, batch 7900, loss[loss=0.2149, simple_loss=0.3026, pruned_loss=0.06362, over 16807.00 frames. ], tot_loss[loss=0.1986, simple_loss=0.2862, pruned_loss=0.05547, over 3095790.79 frames. ], batch size: 124, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:46:16,930 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302255.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 23:46:25,222 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:46:25,382 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302261.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:47:05,166 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.8496, 4.6842, 4.8869, 5.0313, 5.2155, 4.6794, 5.2048, 5.2168], device='cuda:0'), covar=tensor([0.2132, 0.1413, 0.1694, 0.0806, 0.0617, 0.1077, 0.0674, 0.0719], device='cuda:0'), in_proj_covar=tensor([0.0676, 0.0820, 0.0948, 0.0842, 0.0638, 0.0659, 0.0700, 0.0810], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:47:22,374 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302296.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:47:32,982 INFO [train.py:904] (0/8) Epoch 30, batch 7950, loss[loss=0.1932, simple_loss=0.2869, pruned_loss=0.04978, over 16395.00 frames. ], tot_loss[loss=0.1998, simple_loss=0.2867, pruned_loss=0.05642, over 3080557.11 frames. ], batch size: 35, lr: 2.23e-03, grad_scale: 4.0 2023-05-02 23:47:36,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.7568, 4.0374, 4.2380, 4.1890, 4.2040, 3.9947, 3.7219, 3.9955], device='cuda:0'), covar=tensor([0.0618, 0.0823, 0.0579, 0.0652, 0.0761, 0.0713, 0.1489, 0.0702], device='cuda:0'), in_proj_covar=tensor([0.0439, 0.0501, 0.0479, 0.0444, 0.0523, 0.0507, 0.0583, 0.0409], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-02 23:47:41,025 INFO [optim.py:368] (0/8) 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 23:47:53,459 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8959, 4.3183, 3.0848, 2.4807, 2.9068, 2.6879, 4.7579, 3.8524], device='cuda:0'), covar=tensor([0.3045, 0.0649, 0.2009, 0.2981, 0.2999, 0.2191, 0.0415, 0.1259], device='cuda:0'), in_proj_covar=tensor([0.0342, 0.0278, 0.0318, 0.0332, 0.0310, 0.0283, 0.0307, 0.0356], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 23:48:48,638 INFO [train.py:904] (0/8) Epoch 30, batch 8000, loss[loss=0.2595, simple_loss=0.3239, pruned_loss=0.09755, over 11462.00 frames. ], tot_loss[loss=0.2002, simple_loss=0.287, pruned_loss=0.05672, over 3083373.88 frames. ], batch size: 247, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:48:54,506 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302357.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:49:48,592 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5369, 3.6042, 3.3518, 3.0641, 3.2093, 3.5038, 3.2904, 3.3246], device='cuda:0'), covar=tensor([0.0595, 0.0775, 0.0312, 0.0280, 0.0507, 0.0531, 0.1556, 0.0527], device='cuda:0'), in_proj_covar=tensor([0.0310, 0.0470, 0.0362, 0.0364, 0.0359, 0.0418, 0.0251, 0.0432], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:49:57,567 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302399.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:50:03,897 INFO [train.py:904] (0/8) Epoch 30, batch 8050, loss[loss=0.2141, simple_loss=0.2817, pruned_loss=0.0733, over 11505.00 frames. ], tot_loss[loss=0.1991, simple_loss=0.2861, pruned_loss=0.05603, over 3078021.74 frames. ], batch size: 246, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:50:11,665 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.866e+02 2.982e+02 3.316e+02 3.990e+02 1.007e+03, threshold=6.633e+02, percent-clipped=4.0 2023-05-02 23:50:29,778 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4187, 3.5351, 2.5846, 2.1245, 2.2586, 2.2398, 3.6949, 3.0812], device='cuda:0'), covar=tensor([0.3469, 0.0691, 0.2197, 0.3240, 0.3084, 0.2553, 0.0625, 0.1541], device='cuda:0'), in_proj_covar=tensor([0.0340, 0.0276, 0.0316, 0.0330, 0.0308, 0.0282, 0.0305, 0.0353], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-02 23:51:10,082 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302447.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:51:13,956 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8773, 2.7541, 2.6794, 1.9560, 2.6170, 2.7682, 2.6201, 1.9054], device='cuda:0'), covar=tensor([0.0519, 0.0118, 0.0108, 0.0409, 0.0152, 0.0166, 0.0147, 0.0474], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0092, 0.0093, 0.0136, 0.0103, 0.0117, 0.0100, 0.0132], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-02 23:51:21,040 INFO [train.py:904] (0/8) Epoch 30, batch 8100, loss[loss=0.1993, simple_loss=0.2886, pruned_loss=0.05499, over 16208.00 frames. ], tot_loss[loss=0.199, simple_loss=0.286, pruned_loss=0.05601, over 3052265.18 frames. ], batch size: 165, lr: 2.23e-03, grad_scale: 8.0 2023-05-02 23:51:56,151 INFO [zipformer.py:625] (0/8) 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] (0/8) Epoch 30, batch 8150, loss[loss=0.1655, simple_loss=0.2476, pruned_loss=0.04172, over 16628.00 frames. ], tot_loss[loss=0.1971, simple_loss=0.2839, pruned_loss=0.05521, over 3063011.68 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:52:43,486 INFO [optim.py:368] (0/8) 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 23:53:28,213 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302538.0, num_to_drop=1, layers_to_drop={0} 2023-05-02 23:53:52,245 INFO [train.py:904] (0/8) Epoch 30, batch 8200, loss[loss=0.1831, simple_loss=0.2744, pruned_loss=0.04593, over 16379.00 frames. ], tot_loss[loss=0.1945, simple_loss=0.2806, pruned_loss=0.05417, over 3083302.04 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:53:53,891 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302555.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:53:55,109 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302556.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:54:03,204 INFO [zipformer.py:625] (0/8) 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] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302603.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:13,853 INFO [train.py:904] (0/8) Epoch 30, batch 8250, loss[loss=0.1762, simple_loss=0.2775, pruned_loss=0.03747, over 16431.00 frames. ], tot_loss[loss=0.191, simple_loss=0.2793, pruned_loss=0.05137, over 3069043.18 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:55:22,072 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.591e+02 2.554e+02 2.948e+02 3.684e+02 6.863e+02, threshold=5.896e+02, percent-clipped=1.0 2023-05-02 23:55:22,472 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302609.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:55:28,007 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.2395, 2.9838, 3.0984, 1.9317, 3.2440, 3.2823, 2.8127, 2.7582], device='cuda:0'), covar=tensor([0.0793, 0.0258, 0.0238, 0.1181, 0.0111, 0.0249, 0.0472, 0.0450], device='cuda:0'), in_proj_covar=tensor([0.0150, 0.0112, 0.0104, 0.0139, 0.0088, 0.0134, 0.0132, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 23:55:50,647 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.17 vs. limit=2.0 2023-05-02 23:56:01,981 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302632.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:35,858 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302652.0, num_to_drop=0, layers_to_drop=set() 2023-05-02 23:56:38,713 INFO [train.py:904] (0/8) Epoch 30, batch 8300, loss[loss=0.1729, simple_loss=0.2736, pruned_loss=0.03609, over 16361.00 frames. ], tot_loss[loss=0.1868, simple_loss=0.277, pruned_loss=0.04829, over 3077634.07 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:56:46,481 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8072, 2.4825, 2.4605, 3.3979, 1.9202, 3.6219, 1.5988, 2.8621], device='cuda:0'), covar=tensor([0.1403, 0.0762, 0.1119, 0.0184, 0.0102, 0.0344, 0.1740, 0.0734], device='cuda:0'), in_proj_covar=tensor([0.0175, 0.0183, 0.0203, 0.0208, 0.0209, 0.0220, 0.0213, 0.0202], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-02 23:57:43,556 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302693.0, num_to_drop=1, layers_to_drop={3} 2023-05-02 23:58:01,687 INFO [train.py:904] (0/8) Epoch 30, batch 8350, loss[loss=0.1654, simple_loss=0.266, pruned_loss=0.0324, over 16469.00 frames. ], tot_loss[loss=0.1849, simple_loss=0.2769, pruned_loss=0.04642, over 3084967.36 frames. ], batch size: 75, lr: 2.22e-03, grad_scale: 8.0 2023-05-02 23:58:09,439 INFO [optim.py:368] (0/8) 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 23:58:16,615 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0209, 2.1202, 2.2129, 3.4534, 2.0269, 2.3807, 2.2267, 2.2476], device='cuda:0'), covar=tensor([0.1383, 0.3786, 0.3313, 0.0707, 0.4779, 0.2820, 0.4118, 0.3616], device='cuda:0'), in_proj_covar=tensor([0.0420, 0.0473, 0.0385, 0.0334, 0.0442, 0.0541, 0.0446, 0.0553], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-02 23:59:22,786 INFO [train.py:904] (0/8) Epoch 30, batch 8400, loss[loss=0.1471, simple_loss=0.2504, pruned_loss=0.02191, over 16862.00 frames. ], tot_loss[loss=0.1814, simple_loss=0.2741, pruned_loss=0.04429, over 3078402.63 frames. ], batch size: 96, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:44,418 INFO [train.py:904] (0/8) Epoch 30, batch 8450, loss[loss=0.1579, simple_loss=0.2616, pruned_loss=0.02708, over 16360.00 frames. ], tot_loss[loss=0.1784, simple_loss=0.2718, pruned_loss=0.04249, over 3066816.77 frames. ], batch size: 165, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:00:45,007 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.7327, 4.9282, 5.1467, 4.9007, 5.0431, 5.5209, 5.0213, 4.7593], device='cuda:0'), covar=tensor([0.1093, 0.1825, 0.1979, 0.1816, 0.2013, 0.0789, 0.1445, 0.2154], device='cuda:0'), in_proj_covar=tensor([0.0426, 0.0633, 0.0706, 0.0517, 0.0685, 0.0725, 0.0548, 0.0689], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0003, 0.0003, 0.0002, 0.0003], device='cuda:0') 2023-05-03 00:00:52,082 INFO [optim.py:368] (0/8) 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 00:00:54,445 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6162, 3.8708, 3.9432, 2.7102, 3.4770, 3.9083, 3.6212, 2.3957], device='cuda:0'), covar=tensor([0.0500, 0.0078, 0.0053, 0.0401, 0.0138, 0.0113, 0.0098, 0.0490], device='cuda:0'), in_proj_covar=tensor([0.0139, 0.0092, 0.0093, 0.0136, 0.0103, 0.0117, 0.0099, 0.0131], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0004], device='cuda:0') 2023-05-03 00:01:16,727 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.31 vs. limit=2.0 2023-05-03 00:01:31,128 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302833.0, num_to_drop=1, layers_to_drop={2} 2023-05-03 00:02:04,579 INFO [train.py:904] (0/8) Epoch 30, batch 8500, loss[loss=0.1552, simple_loss=0.2353, pruned_loss=0.03758, over 11735.00 frames. ], tot_loss[loss=0.175, simple_loss=0.2685, pruned_loss=0.04073, over 3054345.32 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:02:08,485 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302856.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:02:48,221 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=302881.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:02:55,745 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.52 vs. limit=2.0 2023-05-03 00:03:07,407 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4317, 3.3455, 3.4658, 3.5318, 3.5784, 3.3333, 3.5608, 3.6239], device='cuda:0'), covar=tensor([0.1342, 0.1005, 0.1114, 0.0680, 0.0707, 0.2063, 0.0939, 0.0949], device='cuda:0'), in_proj_covar=tensor([0.0666, 0.0808, 0.0935, 0.0832, 0.0630, 0.0651, 0.0689, 0.0800], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:03:13,609 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0719, 3.0836, 1.8909, 3.2949, 2.2987, 3.3413, 2.1903, 2.6611], device='cuda:0'), covar=tensor([0.0301, 0.0378, 0.1585, 0.0289, 0.0884, 0.0488, 0.1362, 0.0684], device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0179, 0.0193, 0.0170, 0.0176, 0.0216, 0.0200, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-03 00:03:26,607 INFO [train.py:904] (0/8) Epoch 30, batch 8550, loss[loss=0.1586, simple_loss=0.2475, pruned_loss=0.03484, over 11775.00 frames. ], tot_loss[loss=0.1731, simple_loss=0.2663, pruned_loss=0.03997, over 3037143.36 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:03:27,652 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=302904.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:03:38,834 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.520e+02 2.175e+02 2.493e+02 2.958e+02 5.835e+02, threshold=4.986e+02, percent-clipped=2.0 2023-05-03 00:03:39,571 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.8028, 3.4618, 3.8617, 2.0872, 3.9830, 4.0416, 3.1403, 3.1372], device='cuda:0'), covar=tensor([0.0665, 0.0269, 0.0203, 0.1068, 0.0082, 0.0182, 0.0394, 0.0422], device='cuda:0'), in_proj_covar=tensor([0.0148, 0.0110, 0.0101, 0.0137, 0.0086, 0.0132, 0.0130, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-03 00:03:59,964 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=2.85 vs. limit=5.0 2023-05-03 00:04:11,689 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=4.91 vs. limit=5.0 2023-05-03 00:04:45,093 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=302942.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:04:56,149 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.6286, 2.6108, 2.4164, 4.0672, 2.3689, 3.9193, 1.5309, 2.9204], device='cuda:0'), covar=tensor([0.1452, 0.0823, 0.1274, 0.0177, 0.0133, 0.0373, 0.1753, 0.0775], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0181, 0.0201, 0.0206, 0.0206, 0.0218, 0.0211, 0.0200], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-03 00:05:04,468 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=302952.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:05:08,036 INFO [train.py:904] (0/8) Epoch 30, batch 8600, loss[loss=0.1617, simple_loss=0.2487, pruned_loss=0.03733, over 12270.00 frames. ], tot_loss[loss=0.1721, simple_loss=0.2663, pruned_loss=0.03891, over 3029652.97 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:05:37,143 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.16 vs. limit=2.0 2023-05-03 00:05:52,928 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.85 vs. limit=2.0 2023-05-03 00:06:16,020 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=302988.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:06:39,321 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303000.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:06:46,611 INFO [train.py:904] (0/8) Epoch 30, batch 8650, loss[loss=0.1726, simple_loss=0.2664, pruned_loss=0.03939, over 16630.00 frames. ], tot_loss[loss=0.1696, simple_loss=0.2641, pruned_loss=0.03755, over 3025674.28 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:06:58,845 INFO [optim.py:368] (0/8) 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 00:07:05,245 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1123, 2.5590, 2.6499, 1.9509, 2.7796, 2.7773, 2.5108, 2.5192], device='cuda:0'), covar=tensor([0.0706, 0.0278, 0.0249, 0.0990, 0.0134, 0.0266, 0.0489, 0.0419], device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0109, 0.0101, 0.0137, 0.0086, 0.0131, 0.0130, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0003, 0.0003, 0.0004, 0.0002, 0.0004, 0.0004, 0.0003], device='cuda:0') 2023-05-03 00:08:31,135 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303052.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:08:34,586 INFO [train.py:904] (0/8) Epoch 30, batch 8700, loss[loss=0.1413, simple_loss=0.233, pruned_loss=0.02477, over 12045.00 frames. ], tot_loss[loss=0.1666, simple_loss=0.261, pruned_loss=0.03608, over 3022258.65 frames. ], batch size: 248, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:14,442 INFO [train.py:904] (0/8) Epoch 30, batch 8750, loss[loss=0.1807, simple_loss=0.2815, pruned_loss=0.03999, over 16696.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2607, pruned_loss=0.03549, over 3032165.46 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:10:25,190 INFO [optim.py:368] (0/8) 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 00:10:38,659 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303113.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:11:05,914 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5107, 3.7573, 3.7964, 2.6397, 3.3799, 3.8228, 3.5473, 2.1712], device='cuda:0'), covar=tensor([0.0498, 0.0077, 0.0060, 0.0401, 0.0133, 0.0102, 0.0101, 0.0526], device='cuda:0'), in_proj_covar=tensor([0.0138, 0.0090, 0.0092, 0.0134, 0.0103, 0.0115, 0.0098, 0.0130], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-03 00:11:23,592 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303133.0, num_to_drop=1, layers_to_drop={2} 2023-05-03 00:12:06,345 INFO [train.py:904] (0/8) Epoch 30, batch 8800, loss[loss=0.171, simple_loss=0.2688, pruned_loss=0.03662, over 16770.00 frames. ], tot_loss[loss=0.1642, simple_loss=0.2592, pruned_loss=0.03458, over 3039902.39 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:13:02,852 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303181.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:13:09,903 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.5169, 3.6824, 3.3698, 3.1077, 3.0663, 3.5539, 3.2815, 3.3763], device='cuda:0'), covar=tensor([0.0689, 0.0587, 0.0425, 0.0365, 0.0900, 0.0458, 0.2272, 0.0552], device='cuda:0'), in_proj_covar=tensor([0.0308, 0.0467, 0.0360, 0.0361, 0.0355, 0.0415, 0.0249, 0.0429], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:13:50,930 INFO [train.py:904] (0/8) Epoch 30, batch 8850, loss[loss=0.1393, simple_loss=0.2322, pruned_loss=0.02318, over 12345.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2622, pruned_loss=0.03436, over 3034052.03 frames. ], batch size: 246, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:14:00,696 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.436e+02 2.249e+02 2.609e+02 3.163e+02 6.486e+02, threshold=5.217e+02, percent-clipped=3.0 2023-05-03 00:14:12,182 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.3992, 2.8814, 2.8587, 1.6891, 2.5925, 1.9100, 2.9628, 3.1242], device='cuda:0'), covar=tensor([0.0340, 0.1112, 0.0975, 0.3059, 0.1339, 0.1512, 0.0902, 0.1095], device='cuda:0'), in_proj_covar=tensor([0.0157, 0.0167, 0.0167, 0.0155, 0.0146, 0.0131, 0.0143, 0.0180], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004], device='cuda:0') 2023-05-03 00:15:03,375 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303237.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:15:38,687 INFO [train.py:904] (0/8) Epoch 30, batch 8900, loss[loss=0.1809, simple_loss=0.2797, pruned_loss=0.04105, over 16770.00 frames. ], tot_loss[loss=0.1655, simple_loss=0.2632, pruned_loss=0.03383, over 3052188.13 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:15:49,555 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303258.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:17:03,856 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303288.0, num_to_drop=1, layers_to_drop={1} 2023-05-03 00:17:25,116 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4922, 3.3917, 2.7115, 2.1474, 2.1566, 2.2680, 3.4605, 2.9080], device='cuda:0'), covar=tensor([0.2994, 0.0632, 0.1878, 0.3065, 0.2921, 0.2362, 0.0466, 0.1562], device='cuda:0'), in_proj_covar=tensor([0.0334, 0.0271, 0.0312, 0.0326, 0.0302, 0.0278, 0.0300, 0.0347], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-03 00:17:43,161 INFO [train.py:904] (0/8) Epoch 30, batch 8950, loss[loss=0.1462, simple_loss=0.2403, pruned_loss=0.02603, over 17243.00 frames. ], tot_loss[loss=0.1658, simple_loss=0.2629, pruned_loss=0.03436, over 3061762.35 frames. ], batch size: 52, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:17:53,313 INFO [optim.py:368] (0/8) 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 00:18:16,133 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303319.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:18:53,979 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303336.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:19:10,837 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.7408, 2.2598, 1.9300, 2.0052, 2.5551, 2.1890, 2.1380, 2.6448], device='cuda:0'), covar=tensor([0.0205, 0.0514, 0.0630, 0.0536, 0.0344, 0.0464, 0.0257, 0.0336], device='cuda:0'), in_proj_covar=tensor([0.0226, 0.0239, 0.0230, 0.0231, 0.0241, 0.0237, 0.0235, 0.0238], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:19:32,199 INFO [train.py:904] (0/8) Epoch 30, batch 9000, loss[loss=0.1422, simple_loss=0.2294, pruned_loss=0.02752, over 16450.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2598, pruned_loss=0.03318, over 3075010.09 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:19:32,200 INFO [train.py:929] (0/8) Computing validation loss 2023-05-03 00:19:42,081 INFO [train.py:938] (0/8) Epoch 30, validation: loss=0.1431, simple_loss=0.2465, pruned_loss=0.01984, over 944034.00 frames. 2023-05-03 00:19:42,082 INFO [train.py:939] (0/8) Maximum memory allocated so far is 17814MB 2023-05-03 00:20:14,547 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.48 vs. limit=2.0 2023-05-03 00:21:27,819 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303403.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:29,175 INFO [train.py:904] (0/8) Epoch 30, batch 9050, loss[loss=0.1589, simple_loss=0.2541, pruned_loss=0.03184, over 15377.00 frames. ], tot_loss[loss=0.1636, simple_loss=0.2604, pruned_loss=0.03343, over 3065894.35 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:21:38,023 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303408.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:21:39,556 INFO [optim.py:368] (0/8) 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 00:21:48,917 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.3686, 4.7159, 4.5211, 4.5562, 4.2666, 4.2343, 4.1970, 4.7405], device='cuda:0'), covar=tensor([0.1280, 0.0797, 0.1000, 0.0728, 0.0780, 0.1474, 0.1143, 0.0852], device='cuda:0'), in_proj_covar=tensor([0.0713, 0.0855, 0.0705, 0.0667, 0.0545, 0.0546, 0.0716, 0.0673], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:23:13,940 INFO [train.py:904] (0/8) Epoch 30, batch 9100, loss[loss=0.182, simple_loss=0.2757, pruned_loss=0.04411, over 16968.00 frames. ], tot_loss[loss=0.1641, simple_loss=0.2601, pruned_loss=0.03409, over 3067496.70 frames. ], batch size: 116, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:23:17,787 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9425, 2.1050, 2.1303, 3.4013, 2.0495, 2.3306, 2.2455, 2.2046], device='cuda:0'), covar=tensor([0.1484, 0.3976, 0.3592, 0.0727, 0.4751, 0.2891, 0.3948, 0.3840], device='cuda:0'), in_proj_covar=tensor([0.0418, 0.0470, 0.0385, 0.0331, 0.0440, 0.0539, 0.0445, 0.0550], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:23:33,651 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303464.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:24:58,979 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.4565, 3.0371, 2.7096, 2.2471, 2.1957, 2.3377, 2.9909, 2.7723], device='cuda:0'), covar=tensor([0.2673, 0.0726, 0.1748, 0.2994, 0.2887, 0.2399, 0.0521, 0.1640], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0270, 0.0310, 0.0324, 0.0300, 0.0276, 0.0299, 0.0345], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-03 00:25:08,033 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.5029, 3.7478, 2.7594, 2.1521, 2.2336, 2.4389, 3.9164, 3.1094], device='cuda:0'), covar=tensor([0.3259, 0.0603, 0.2020, 0.3176, 0.3055, 0.2278, 0.0416, 0.1572], device='cuda:0'), in_proj_covar=tensor([0.0332, 0.0270, 0.0310, 0.0324, 0.0300, 0.0276, 0.0299, 0.0346], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003], device='cuda:0') 2023-05-03 00:25:12,858 INFO [train.py:904] (0/8) Epoch 30, batch 9150, loss[loss=0.1563, simple_loss=0.2492, pruned_loss=0.03169, over 16327.00 frames. ], tot_loss[loss=0.1646, simple_loss=0.2612, pruned_loss=0.03405, over 3073280.43 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:25:25,339 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.438e+02 2.268e+02 2.639e+02 3.152e+02 6.290e+02, threshold=5.277e+02, percent-clipped=4.0 2023-05-03 00:25:41,505 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.0513, 4.0852, 4.3183, 4.3048, 4.3222, 4.1297, 4.1209, 4.1406], device='cuda:0'), covar=tensor([0.0354, 0.0837, 0.0483, 0.0406, 0.0441, 0.0473, 0.0748, 0.0472], device='cuda:0'), in_proj_covar=tensor([0.0431, 0.0489, 0.0471, 0.0436, 0.0515, 0.0497, 0.0569, 0.0401], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-03 00:26:22,496 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303537.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:26:54,954 INFO [train.py:904] (0/8) Epoch 30, batch 9200, loss[loss=0.1583, simple_loss=0.24, pruned_loss=0.03826, over 12086.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2577, pruned_loss=0.03353, over 3068784.59 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:27:31,983 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303574.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:27:51,756 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303585.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:28:13,174 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.8951, 6.2951, 5.9786, 6.1223, 5.6495, 5.7116, 5.6360, 6.3520], device='cuda:0'), covar=tensor([0.1228, 0.0796, 0.0915, 0.0722, 0.0759, 0.0551, 0.1319, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0711, 0.0855, 0.0704, 0.0667, 0.0545, 0.0546, 0.0717, 0.0673], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0001, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:28:29,163 INFO [train.py:904] (0/8) Epoch 30, batch 9250, loss[loss=0.1564, simple_loss=0.2513, pruned_loss=0.03077, over 16668.00 frames. ], tot_loss[loss=0.1625, simple_loss=0.2576, pruned_loss=0.03373, over 3061703.35 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:28:41,902 INFO [optim.py:368] (0/8) 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 00:28:49,560 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303614.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:29:39,005 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303635.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:30:18,347 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4518, 4.5478, 4.3787, 4.0546, 4.1082, 4.4786, 4.1998, 4.2106], device='cuda:0'), covar=tensor([0.0594, 0.0604, 0.0317, 0.0309, 0.0754, 0.0581, 0.0656, 0.0665], device='cuda:0'), in_proj_covar=tensor([0.0307, 0.0465, 0.0360, 0.0359, 0.0353, 0.0413, 0.0247, 0.0427], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:30:19,439 INFO [train.py:904] (0/8) Epoch 30, batch 9300, loss[loss=0.139, simple_loss=0.2334, pruned_loss=0.02233, over 16746.00 frames. ], tot_loss[loss=0.1606, simple_loss=0.2555, pruned_loss=0.03288, over 3057547.57 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:30:33,999 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.47 vs. limit=2.0 2023-05-03 00:31:31,844 INFO [scaling.py:679] (0/8) Whitening: num_groups=1, num_channels=384, metric=3.99 vs. limit=5.0 2023-05-03 00:32:05,577 INFO [train.py:904] (0/8) Epoch 30, batch 9350, loss[loss=0.1569, simple_loss=0.2557, pruned_loss=0.02904, over 16554.00 frames. ], tot_loss[loss=0.1598, simple_loss=0.2545, pruned_loss=0.03256, over 3052889.40 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:32:13,799 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303708.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:32:16,693 INFO [optim.py:368] (0/8) 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 00:33:31,735 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.9013, 2.0650, 2.4058, 3.1928, 2.1923, 2.2422, 2.3234, 2.2269], device='cuda:0'), covar=tensor([0.1600, 0.4139, 0.3196, 0.0895, 0.5117, 0.3261, 0.3763, 0.4136], device='cuda:0'), in_proj_covar=tensor([0.0419, 0.0471, 0.0386, 0.0332, 0.0441, 0.0540, 0.0446, 0.0551], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:33:46,278 INFO [train.py:904] (0/8) Epoch 30, batch 9400, loss[loss=0.1756, simple_loss=0.2768, pruned_loss=0.03719, over 16810.00 frames. ], tot_loss[loss=0.1596, simple_loss=0.2546, pruned_loss=0.0323, over 3060616.74 frames. ], batch size: 124, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:33:50,141 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303756.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:33:56,173 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303759.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:34:23,972 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.6912, 2.1285, 1.8610, 1.9273, 2.4118, 2.0534, 2.0039, 2.5182], device='cuda:0'), covar=tensor([0.0209, 0.0525, 0.0645, 0.0531, 0.0363, 0.0486, 0.0218, 0.0346], device='cuda:0'), in_proj_covar=tensor([0.0227, 0.0241, 0.0231, 0.0232, 0.0243, 0.0240, 0.0235, 0.0239], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:35:00,693 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3216, 3.5358, 3.6057, 2.4526, 3.2597, 3.6377, 3.4455, 2.0125], device='cuda:0'), covar=tensor([0.0552, 0.0084, 0.0071, 0.0458, 0.0148, 0.0105, 0.0105, 0.0581], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0134, 0.0103, 0.0114, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-03 00:35:11,646 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.0578, 3.0957, 2.0125, 3.3315, 2.3702, 3.3100, 2.1473, 2.6093], device='cuda:0'), covar=tensor([0.0388, 0.0439, 0.1585, 0.0324, 0.0857, 0.0652, 0.1581, 0.0835], device='cuda:0'), in_proj_covar=tensor([0.0174, 0.0178, 0.0191, 0.0168, 0.0176, 0.0215, 0.0200, 0.0179], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-03 00:35:20,301 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4799, 3.4241, 3.5277, 3.5940, 3.6310, 3.3376, 3.6025, 3.6800], device='cuda:0'), covar=tensor([0.1382, 0.0908, 0.1074, 0.0695, 0.0645, 0.2567, 0.0934, 0.0737], device='cuda:0'), in_proj_covar=tensor([0.0654, 0.0791, 0.0913, 0.0819, 0.0618, 0.0641, 0.0678, 0.0784], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:35:26,329 INFO [train.py:904] (0/8) Epoch 30, batch 9450, loss[loss=0.1539, simple_loss=0.2503, pruned_loss=0.02877, over 16888.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2567, pruned_loss=0.03259, over 3052220.80 frames. ], batch size: 116, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:35:37,000 INFO [optim.py:368] (0/8) Clipping_scale=2.0, grad-norm quartiles 1.278e+02 2.120e+02 2.504e+02 3.073e+02 7.133e+02, threshold=5.009e+02, percent-clipped=4.0 2023-05-03 00:35:37,968 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303810.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:37:08,357 INFO [train.py:904] (0/8) Epoch 30, batch 9500, loss[loss=0.1584, simple_loss=0.2586, pruned_loss=0.02911, over 16375.00 frames. ], tot_loss[loss=0.1607, simple_loss=0.2563, pruned_loss=0.03256, over 3060707.11 frames. ], batch size: 146, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:37:43,299 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303871.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:21,261 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.8671, 1.3924, 1.7573, 1.8095, 1.8975, 1.9453, 1.7665, 1.9212], device='cuda:0'), covar=tensor([0.0289, 0.0515, 0.0260, 0.0363, 0.0371, 0.0228, 0.0514, 0.0172], device='cuda:0'), in_proj_covar=tensor([0.0195, 0.0197, 0.0185, 0.0190, 0.0208, 0.0165, 0.0203, 0.0166], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:38:36,408 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303896.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:38:54,498 INFO [train.py:904] (0/8) Epoch 30, batch 9550, loss[loss=0.1691, simple_loss=0.2665, pruned_loss=0.03582, over 16686.00 frames. ], tot_loss[loss=0.1609, simple_loss=0.2562, pruned_loss=0.03281, over 3071399.78 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:39:08,525 INFO [optim.py:368] (0/8) 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 00:39:17,620 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=303914.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:39:40,470 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([1.9177, 2.2973, 2.3562, 3.1208, 1.8383, 3.3389, 1.7562, 2.7734], device='cuda:0'), covar=tensor([0.1391, 0.0791, 0.1203, 0.0178, 0.0101, 0.0369, 0.1703, 0.0779], device='cuda:0'), in_proj_covar=tensor([0.0172, 0.0179, 0.0198, 0.0201, 0.0200, 0.0214, 0.0208, 0.0197], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004, 0.0004], device='cuda:0') 2023-05-03 00:39:49,599 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=303930.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:40:36,089 INFO [train.py:904] (0/8) Epoch 30, batch 9600, loss[loss=0.1601, simple_loss=0.2631, pruned_loss=0.02858, over 16664.00 frames. ], tot_loss[loss=0.1619, simple_loss=0.2571, pruned_loss=0.03336, over 3054824.29 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:40:43,409 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=303957.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:40:52,684 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=303962.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:41:15,718 INFO [zipformer.py:625] (0/8) warmup_begin=2666.7, warmup_end=3333.3, batch_count=303974.0, num_to_drop=1, layers_to_drop={0} 2023-05-03 00:42:10,761 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.30 vs. limit=2.0 2023-05-03 00:42:16,393 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/checkpoint-304000.pt 2023-05-03 00:42:31,195 INFO [train.py:904] (0/8) Epoch 30, batch 9650, loss[loss=0.1623, simple_loss=0.2608, pruned_loss=0.03186, over 15408.00 frames. ], tot_loss[loss=0.1631, simple_loss=0.2592, pruned_loss=0.03353, over 3061251.14 frames. ], batch size: 191, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:42:48,098 INFO [optim.py:368] (0/8) 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 00:43:23,326 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.4759, 3.4037, 3.5292, 3.6006, 3.6313, 3.3568, 3.6083, 3.6798], device='cuda:0'), covar=tensor([0.1373, 0.0982, 0.1218, 0.0732, 0.0626, 0.2144, 0.0951, 0.0818], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0783, 0.0902, 0.0811, 0.0611, 0.0633, 0.0671, 0.0775], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:43:27,053 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([3.8476, 3.7307, 3.9073, 3.9968, 4.0928, 3.7050, 4.0618, 4.1302], device='cuda:0'), covar=tensor([0.1705, 0.1099, 0.1458, 0.0790, 0.0557, 0.1823, 0.0720, 0.0823], device='cuda:0'), in_proj_covar=tensor([0.0647, 0.0783, 0.0902, 0.0811, 0.0611, 0.0633, 0.0671, 0.0776], device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:43:40,611 INFO [zipformer.py:625] (0/8) warmup_begin=3333.3, warmup_end=4000.0, batch_count=304035.0, num_to_drop=1, layers_to_drop={2} 2023-05-03 00:43:58,817 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.88 vs. limit=2.0 2023-05-03 00:44:20,850 INFO [train.py:904] (0/8) Epoch 30, batch 9700, loss[loss=0.1557, simple_loss=0.2484, pruned_loss=0.03154, over 17051.00 frames. ], tot_loss[loss=0.1624, simple_loss=0.2579, pruned_loss=0.03343, over 3055866.57 frames. ], batch size: 55, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:44:21,563 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.2856, 4.2717, 4.1396, 3.3205, 4.1899, 1.8140, 3.9803, 3.7328], device='cuda:0'), covar=tensor([0.0148, 0.0127, 0.0211, 0.0273, 0.0114, 0.2856, 0.0149, 0.0315], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0172, 0.0209, 0.0180, 0.0187, 0.0215, 0.0198, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:44:30,054 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304059.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:46:04,100 INFO [train.py:904] (0/8) Epoch 30, batch 9750, loss[loss=0.1837, simple_loss=0.2814, pruned_loss=0.04297, over 16894.00 frames. ], tot_loss[loss=0.1626, simple_loss=0.2576, pruned_loss=0.03378, over 3050201.23 frames. ], batch size: 116, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:46:09,340 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=304107.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:46:17,047 INFO [optim.py:368] (0/8) 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 00:46:26,289 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.5264, 4.5087, 4.8534, 4.8076, 4.8524, 4.5964, 4.5467, 4.5017], device='cuda:0'), covar=tensor([0.0338, 0.0638, 0.0441, 0.0483, 0.0483, 0.0414, 0.0960, 0.0452], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0489, 0.0469, 0.0436, 0.0513, 0.0496, 0.0567, 0.0400], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-03 00:47:41,843 INFO [train.py:904] (0/8) Epoch 30, batch 9800, loss[loss=0.1696, simple_loss=0.271, pruned_loss=0.03407, over 16630.00 frames. ], tot_loss[loss=0.1614, simple_loss=0.257, pruned_loss=0.03286, over 3053608.64 frames. ], batch size: 134, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:47:56,370 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0219, 5.0139, 5.4054, 5.3863, 5.4214, 5.1283, 5.0526, 4.9525], device='cuda:0'), covar=tensor([0.0370, 0.0741, 0.0498, 0.0461, 0.0436, 0.0509, 0.1094, 0.0416], device='cuda:0'), in_proj_covar=tensor([0.0430, 0.0489, 0.0470, 0.0436, 0.0513, 0.0496, 0.0567, 0.0400], device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0003], device='cuda:0') 2023-05-03 00:48:04,644 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304166.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:49:14,304 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([4.4067, 4.4025, 4.1723, 3.5092, 4.3294, 1.6923, 4.0895, 3.8432], device='cuda:0'), covar=tensor([0.0094, 0.0081, 0.0204, 0.0253, 0.0095, 0.3055, 0.0125, 0.0305], device='cuda:0'), in_proj_covar=tensor([0.0181, 0.0172, 0.0209, 0.0179, 0.0187, 0.0215, 0.0198, 0.0175], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:49:24,997 INFO [train.py:904] (0/8) Epoch 30, batch 9850, loss[loss=0.1502, simple_loss=0.2451, pruned_loss=0.0276, over 16504.00 frames. ], tot_loss[loss=0.1611, simple_loss=0.2578, pruned_loss=0.03216, over 3058534.06 frames. ], batch size: 68, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:49:39,351 INFO [optim.py:368] (0/8) 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 00:49:41,833 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.1991, 3.6150, 3.6246, 2.3835, 3.2612, 3.6418, 3.4264, 1.9137], device='cuda:0'), covar=tensor([0.0642, 0.0065, 0.0067, 0.0493, 0.0145, 0.0102, 0.0098, 0.0636], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0089, 0.0091, 0.0134, 0.0102, 0.0113, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-03 00:50:19,510 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304230.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:50:27,183 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.63 vs. limit=2.0 2023-05-03 00:51:13,333 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304252.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:51:16,031 INFO [train.py:904] (0/8) Epoch 30, batch 9900, loss[loss=0.1685, simple_loss=0.2677, pruned_loss=0.03467, over 15241.00 frames. ], tot_loss[loss=0.1608, simple_loss=0.2579, pruned_loss=0.03185, over 3057319.34 frames. ], batch size: 190, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:52:13,220 INFO [zipformer.py:625] (0/8) warmup_begin=666.7, warmup_end=1333.3, batch_count=304278.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:53:14,783 INFO [train.py:904] (0/8) Epoch 30, batch 9950, loss[loss=0.1615, simple_loss=0.257, pruned_loss=0.03294, over 16988.00 frames. ], tot_loss[loss=0.1628, simple_loss=0.2603, pruned_loss=0.0326, over 3053879.88 frames. ], batch size: 109, lr: 2.22e-03, grad_scale: 4.0 2023-05-03 00:53:31,713 INFO [optim.py:368] (0/8) 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 00:54:00,511 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([5.0190, 5.0572, 4.8707, 4.3798, 4.5097, 4.9464, 4.8764, 4.5988], device='cuda:0'), covar=tensor([0.0626, 0.0680, 0.0396, 0.0405, 0.1128, 0.0562, 0.0315, 0.0799], device='cuda:0'), in_proj_covar=tensor([0.0302, 0.0457, 0.0353, 0.0354, 0.0347, 0.0405, 0.0243, 0.0419], device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002], device='cuda:0') 2023-05-03 00:54:20,736 INFO [zipformer.py:625] (0/8) warmup_begin=1333.3, warmup_end=2000.0, batch_count=304330.0, num_to_drop=1, layers_to_drop={2} 2023-05-03 00:55:15,646 INFO [train.py:904] (0/8) Epoch 30, batch 10000, loss[loss=0.1579, simple_loss=0.2526, pruned_loss=0.03161, over 17154.00 frames. ], tot_loss[loss=0.1622, simple_loss=0.2596, pruned_loss=0.03241, over 3070950.34 frames. ], batch size: 48, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:55:40,301 INFO [zipformer.py:1454] (0/8) attn_weights_entropy = tensor([2.3835, 3.6204, 3.6511, 2.4948, 3.2935, 3.6977, 3.5664, 1.8909], device='cuda:0'), covar=tensor([0.0623, 0.0100, 0.0087, 0.0505, 0.0164, 0.0133, 0.0089, 0.0727], device='cuda:0'), in_proj_covar=tensor([0.0137, 0.0090, 0.0091, 0.0133, 0.0102, 0.0113, 0.0097, 0.0129], device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0002, 0.0002, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003], device='cuda:0') 2023-05-03 00:56:57,648 INFO [train.py:904] (0/8) Epoch 30, batch 10050, loss[loss=0.1742, simple_loss=0.2648, pruned_loss=0.04181, over 12146.00 frames. ], tot_loss[loss=0.1617, simple_loss=0.2592, pruned_loss=0.03216, over 3064310.45 frames. ], batch size: 247, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:57:10,564 INFO [optim.py:368] (0/8) 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 00:58:34,285 INFO [train.py:904] (0/8) Epoch 30, batch 10100, loss[loss=0.1391, simple_loss=0.2389, pruned_loss=0.01969, over 16856.00 frames. ], tot_loss[loss=0.162, simple_loss=0.2595, pruned_loss=0.03228, over 3061939.98 frames. ], batch size: 96, lr: 2.22e-03, grad_scale: 8.0 2023-05-03 00:58:37,383 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=192, metric=1.72 vs. limit=2.0 2023-05-03 00:58:57,200 INFO [zipformer.py:625] (0/8) warmup_begin=2000.0, warmup_end=2666.7, batch_count=304466.0, num_to_drop=0, layers_to_drop=set() 2023-05-03 00:59:20,303 INFO [scaling.py:679] (0/8) Whitening: num_groups=8, num_channels=96, metric=1.19 vs. limit=2.0 2023-05-03 00:59:55,833 INFO [checkpoint.py:75] (0/8) Saving checkpoint to pruned_transducer_stateless7/exp_multidataset/epoch-30.pt 2023-05-03 01:00:01,616 INFO [train.py:1169] (0/8) Done!